Over the last year AI hype has transformed from generative AI, the Chatbots we were all so impressed with months ago, to Agentic AI, those snippets of code that operate behind the curtains and allow us more ‘refined’ cognitive pastimes while they slog through the daily grind of e-mails, memos, and meeting notes. 
Generative AI is a collaborative tool, with the interaction between the model and the user the key to success.  We think of generative Ai as being a somewhat over-eager assistant who is willing to do almost anything you ask but can be a bit careless.  Said assistant, who has no experience, sits with you day after day, eager to help you, sometimes so much so that they can become a bit effusive.  When assigning a task to your generative assistant, you must be very precise and specific about details as the more specific you are the greater the chance that the AI will give an answer that is specific to your query and correct.

The Shift from Chatbots to Agents
Agentic AI can be thought of as the  assistant you have had for years. One for whom might look messy and disorganized but one you can leave a simple note and know the work will  be done while you are out of the office and done just the way you like it.  In fact, sometimes you don’t even have to leave a note, and the work gets done without you input.
While there is relatively little data compiled on agentic tasks at this early stage, the most common task for agentic AI has been summarization.  The common “TL;DR” (Too long; Didn’t read”) shows a generational frustration with long e-mails, You Tube videos over 30 seconds, legal documents, or sadly, books.  Agentic AI is used to avoid even seeing those items in their long form and just reading the summaries as they appear on your desktop.  Agentic AI is also used for classification and triage.  Automatic e-mail and memo sifting/tagging is a common agentic function, allowing all the ‘unimportant  e-mails to fall into the ‘junk’ bucket for a quick human review before they are deleted, while the ‘important ones are elevated to the top of the stack, all before you even see them.
​
The "TL;DR" Culture and the Rise of Triage
Agentic systems come in neither a blanks slate, available for task assignment, nor a pre-programmed package.  As a reminder, AI systems are tools and tools either need to be initially developed with a purpose in mind (a hammer has two-pound nails in and pull them out) or taught a purpose, but agentic AI splits the difference. Here’s how:

• The Model - Most Agents start  with a foundation AI model.  The model has been taught lots of things, everything from physics and literature to writing Python code, but as it stands it can do nothing on its own.
  • The Framework – The model is wrapped in a framework that provides pre-programmed ‘instincts.  The framework adds the following:
    • Planning Loop – This gives the model a path to sequentially follow tasks (“Do X, then Y, then Z”)
    • Handshake – This code allows the model to connect with your computer, the internet, and  other company or private resources that you access on a daily basis.
    • Memory – A “Yellow Pad” so the agent can remember what it learned when it finished task 1 and will not forget it when it finishes each successive task.
  • Job Description – As an agentic AI user, this is your stage.  You define what you want the agentic AI to do for you.  Here are some examples:

Scheduling & Logistics
Your Request - "I need to meet with the Marketing team and our external vendor, Sarah, sometime next week for 45 minutes. Find a time that works for everyone, send the invites, and if Sarah hasn't replied by Friday, send her a polite nudge."
Agent Action - It checks your calendar, emails the team to find their availability, cross-references Sarah’s time zone, drafts the calendar invite, and sets a "watch" on its own internal clock to follow up if it doesn't see a "Meeting Accepted" notification.

Deep Synthesis
Your Request - "Our competitor just released their quarterly earnings report. Compare their growth in the EU sector to ours, put the data into a table in a Google Doc, and Slack me the three biggest risks you see for our Q3 strategy."
Agent Action - It downloads the competitor’s PDF, scrapes your company’s internal database for EU sales, performs the math, creates the document, and sends you a summarized Slack message with the "Top 3 Risks."

Data/Lead Triage
Your Request - "Go through the 'Contact Us' form submissions from the last 24 hours. If any mention a budget over $50k, add them to Salesforce and assign them to Jim. If they are just looking for support, draft a reply with a link to our help docs and archive the ticket."
Agent Action - It reads the incoming text, performs a "Sentiment and Intent" analysis, logs into Salesforce to create the lead, and interacts with your email API to send the support responses.

System Maintenance
Your Request - "I’m getting reports that the client portal is slow. Run a diagnostic on the server logs from the last hour, identify any IP addresses with weird traffic spikes, and if you find a potential DDOS attack, temporarily block those IPs and alert the security team."
Agent Action - It accesses the server via a secure terminal, parses thousands of lines of log data, identifies patterns, executes a "Block" command on specific IPs, and sends a high-priority alert to the human team.

Real-World Delegation: Scheduling, Synthesis, and Triage
With agentic models you can think of yourself as the CEO (unless you already are).  The model is the office manager and those tasks shown above, which can be captured in  “Scheduled  Scripts” are your employees.  You can have them repeat some tasks regularly.  You can call up a more specific task when necessary, or you can create a new one for a specific instance.
When you use generative AI, unless you have an account with that Ai’s provider, the AI itself only knows the content of your query.  The hosting platform collects things like your ip address, the type of device you are using, the browser type, your screen resolution, and places a cookie that temporarily puts an id on your computer, all similar to what is collected when you visit sites on-line.  The generative AI retains the conversation as it progresses so it is able to answer follow-up questions that are part of the chat, but when the chat ends the Ai itself will delete the conversation.  Most platforms will retain the chat for anywhere from 72 hours to 30 days.  This is the limit of what the AI knows about you unless your account with the AI provider, allows you to retain your chats for your own review.
Agentic Ai is very different.  It must have a persistent memory, and it must be integrated into whatever system is using it.  The persistent memory comes in the form of a vector database that builds up over time, eventually becoming a 360° profile of you, the user.  It remembers that time you specified a certain brand, a certain file type, and which tools you prefer to use, but it goes much deeper than that.  It learns what tasks you trust it to do on its own, which newsletters you like to read and which ones you keep.  It tracks your response cadence, the speed at which you respond to various communications.  It learns your “tone” or writing voice and will match it when writing replies, and it can make internal decisions about the priority of those preferences based on simple things like whether you saved all e-mails about a certain topic.  It can learn about your relationships and can understand that a particular contact can be significant to other contacts in your file.  It can see relationships in your daily life that you do not always see, like connecting a PDF in your “download” folder to an e-mail thread from weeks ago because they shared a particular project name or code.

The Privacy Paradox: The 360° User Profile
The bottom-line is, for agentic AI to work well it has to know all about you, and for some this might seem a particularly onerous invasion of privacy.  While a digital intrusion its is not much different than the years of observing your habits that a long-time assistant might make.  While that slice of digital privacy might be problematic for some it is not what we worry about.  Our issue is that once you give the keys to the castle to someone else they can be stolen, and while agentic AI is still in its infancy, such pilferage has already begun.

The Security Crisis: Poisoning and the "Keys to the Castle"
Agentic AI can be attacked in two ways, either on a large scale or a small one.
Large – A patient hacker can ‘poison’ the AI’s training data by adding just a small amount of data.  This is a small number of documents in a massive dataset, but it teaches the AI to associate a certain phrase with a command to bypass security filters.  The AI operates normally until the hacker sends an e-mail or document to the model that contains the trigger phrase, and the Ai follows by dropping its security and doing whatever the hacker specifies.  Recent studies have indicated that it takes only poisoning 0.01% of a database to plant a permanent backdoor in the model.
Small – In this situation the hacker sends a seeming innocuous PDF or e-mail directly to you.  Hidden in the metadata or even printed in white so as to be invisible, the hacker includes instructions that poison your model in a similar fashion  As the agent ‘reads’ your e-mail it absorbs the backdoor instructions, such as including a hidden “CC” to the hackers location  for every e-mail you send.  Taken further, the hacker could instruct the agentic Ai to send him your vector file, and while it might seem like just a list of coordinates, there are tools that allow that data to be converted into plain text, essentially opening up every aspect of your life, business, decision you have made since the agentic AI was installed.  This isn’t just name-address-password, it is everything the agentic AI has collected about what you care about, what you do on a daily basis, and what you don’t.  What could be better to create a false you?  To add insult to injury, what do hackers use to unravel the vector databases they steal? Another AI model.

Conclusion: The Double-Edged Sword of Autonomy
The transition from Generative to Agentic AI represents the most significant leap in personal productivity since the advent of the personal computer. We are moving away from a world where we struggle to keep up with our digital lives and into one where our software finally works as hard as we do. By sifting our inboxes, managing our logistics, and synthesizing our data, these agents offer us the most precious of modern commodities: time. However, as we move from "Instruction" to "Delegation," the stakes change. To be truly useful, an agent must possess the "keys to the castle", access to our files, our schedules, and our very way of thinking. This intimacy creates a paradox of progress: the more an agent knows about us, the more effective it becomes, but the more devastating a potential compromise becomes.
The future of Agentic AI will not just be won by the most capable models, but by the most secure frameworks. As we embrace these digital proxies to handle our "daily grind," we do so with eyes wide open, recognizing that while an agent can be our most powerful ally, its persistent memory and deep integration require a new standard of digital vigilance. In 2026, the goal is no longer just to build an AI that can think, but to build one that we can trust to act in our name without opening a back door to our lives.

The End of the Initial Purchase Era
In 2024 the government of China was faced with a dilemma.  The country’s population had been rising in affluence since the early 2000’s and the economy had seen massive growth.  However by 2017/18 consumer electronics’ products, bought by this new Chinese middle class, were up for replacement, and CE growth transitioned from initial purchase growth to a more normal replacement cycle.  The value of purchases continued to increase as middle class affluence continued to grow, but in 2017 unit volumes in flagship products began to decline for the first time.

The Urban-Rural Connectivity Gap
By that time ~98% of the Chinese population had at least a mobile phone and the transition from a ‘dumb’ mobile phone to a smartphone had already reached between 55% and 60% of the population.  This was driven by the urban population’s desire for better connectivity and more “user-experience” oriented applications.  However the fall in unit volume, for the first time since China had been collecting such data, was a head’s-up for the Chinese government.  There were other factors that precipitated this change, such as the over 98% of the Chinese population who accessed the internet through their phone, and that smartphone conversion was now moving from the urban population to more remote regions, but the unit volume outcome was a blow to the growth that the Chinese government had been touting for years.. 
​
The "Urban Bubble" and the Delayed Response
Chinese government officials lived in the same urban environments as affluent consumers and were sheltered somewhat from the slowing smartphone growth.  They were aware that the growth of more mature CE products, particularly appliances, had been slowing and put in place some relatively narrow subsidies for these items in the early 2020’s, but it was not until 2024 when smartphones were beginning to be considered as part of the government’s  “New for Old” program.  In fact, smartphones were not included until January 2025, after almost 8 years of declining sales.
From Green Energy to National Solvency
That program, which began in a slightly different form in 2024, was originally intended, at least so the Chinese government says, to focus on energy efficiency.  Initially 8 home appliances were targeted:

• Refrigerators
• Washing Machines
• Televisions
• Air Conditioners
• Computers
• Water Heaters
• Household Stoves
• Range Hoods

Along with the automotive sector, with a specific focus on all-electric or plug-in hybrids, the point of these subsidies was that ‘new’ models would be more energy efficient and lower the broad Chinese energy use profile.  In fact, the subsidy was based on the energy efficiency of the ‘new’ device, with a 20% subsidy for those devices with a Grade 1 designation (Most Energy Efficient) and 15% for those with a Grade 2 (Less Energy Efficient).
​
The Fiscal Pivot: Centralizing the Burden
In the early days of the subsidy (2024) much of the funding was done through local governments who fronted the capital.  This local emphasis was the reason behind the program’s slow start and considerable friction between provincial governments and the Central government, but in July of 2024 the Central government issued 150 billion yuan ($20.7 billion US) of ultra-long special treasury bonds (20, 30, and 50 year).
The Central government changed the plan at that time from a roughly 50/50 burden on the provinces, to a general 1:9 ratio, with the central government now bearing most of the program cost.  In actuality, the percentage changes based on the prosperity and location of the provinces involved, with less prosperous provinces in the 5% range and the more affluent in the 15% responsibility range.
We note, in the table below, that the consumer allocation percentage of the float is less than half of the float (and spend) with the larger allocation going to industrial equipment.

Future Tech and the Aging Demographic
This year’s program will expand further, with the NDRC indicating that while 2024 and 2025 was oriented toward unit volume (moving inventory), 2026 will be more about “Future Tech” and the “Silver Economy”.  Future tech is easy to understand.  It includes new items like:

• Smart glasses (AI/AR) (15% rebate capped at 500 yuan ($72.65 US)),
• Fitness bands (not previously included) and some
• Health related products

But the “Silver Economy”[1] bears a bit more scrutiny.  The official definition not only specifies senior citizens but also leans toward those in their 40’s and 50’s, who have to “…prepare for the challenges of an aging population”.  The government takes this new philosophy even further vowing to build 10 dedicated “Silver Economy” industrial parks across the country to ‘cluster’ companies specializing in elder-tech.  It is also a central theme in the 2026 – 2030 Five-Year-Plan that is currently under development and includes age-friendly home renovations, hardware like heart or sleep monitors, and, of course, companion or exoskeleton robotics.  We expect these items will find there way into the subsidy program at some point.

Mandating Quality: The War on "Involution"
At the same time the government also narrowed the number of home appliance categories from 12 to 6, eliminated most Grade 2 subsidies, and lowered the caps on high-end appliances from 2,0900 yuan ($290.61 US) to 1,500 yuan ($217.96 US) to open the funding to a large population swath.  That said, in a more uncharacteristic move, the central government also implemented in the 2026 program what it calls “Anti-Involution”.  Involution is a Chinese word for a “race to the bottom”, the point at which companies are so willing to compete on price that profitability disappears, innovation is stalled, and workers suffer.
New “Price Conduct Rules” that are part of the 2026 subsidy plan make sure that manufacturers do not sell product below cost to capture share and use these rules to rein in ‘disorderly’ low-price e-commerce competition by blacklisting violators for 6 to 12 months.  It seems that the central government, after years of idly standing by while Chinese companies undercut foreign businesses by selling at or below cost to gain share, they seem to notice that it also has the effect of causing deflation, and when that is internal to China it becomes a very sensitive subject  A number of Chinese smartphone brands are very familiar  with the issues surrounding price “conduct” and are well-known for those tactics.

The AI Filter: Subsidizing the Next Generation
This year (2026)  the requirements go even further.  Any smartphone looking to generate a 15% subsidy must meet the following three new requirements relating  to AI:

• Dedicated NPU – The phone must have a stand-alone NPU (Neural Processing Unit) and cannot rely on Ai produced on the phone’s general CPU or GPU.
• On-Device Processing – With variations, this requirement states that the phone’s hardware must be capable of running a 7-billion parameter model without the cloud.
• Memory – All phones must have at least 12GB of RAM, forcing budget phones, which are typically 8GB out of subsidy territory.
• Domesticity – In order to receive a subsidy the phone must be on the central government’s approved list.  To maintain a place on that list it must contain a domestically-produced processor or a compliant one.  Typically Huawei’s (pvt) Kirin processors, Xiaomi’s (1810.HK) Surge, and certain Qualcomm (QCOM) Density processors, along with AI framework processors like those from Baidu (BIDU) (Ernie) and Alibaba (BABA) (Qwen)

The Compliance Trap: Apple and the Regulatory Pincer.
In the 2nd half of this year the Chinese government is expected to release a set of AI Handset Standards (2.0) that will mesh with the subsidy requirements.  The focus here will be AI privacy, essentially keeping AI from leaking sensitive user and corporate data.  This will be significant for Apple (AAPL), as the company will find itself caught between the strict subsidy price caps and AI compliance issues.  Apple is able to pass the NPU requirement as both the A19 and A19 Pro chips in the iPhone 17 series have a neural engine that qualifies.  The iPhone 17 Pro, the Pro Max, and the iPhone Air all meet the RAM requirement (12GB) but the basic iPhone 17 does not with 8GB, meaning it can be cut out of the subsidy pie if the memory is not changed, a margin question for Apple, especially given the rapid rise in memory pricing.  Apple also faced the price cap rule (6,000 yuan) for the iPhone 17 in 2025, but was able to ‘adjust’ the initial price to 5,999 yuan last year, making the 500 yuan subsidy available to this basic model in China. 
But there is another more fuzzy issue, that referring to Apple Intelligence.  Apple has been forced to partner with Alibaba and Baidu to ‘backend’ Apple intelligence in China and  Apple’s AI must be able to refuse to answer 95% of the 2,000 questions asked by the government concerning sensitive political issues.  Apple is still in the testing phase, but could find itself in an “all”, “none”, or “partial” mode where it might not be able to access subsidies in every province.  Apple can assuage these potential problems by offering higher trade-in values for other Apple products to offset the loss of the subsidy, but it is still up in the air about the final level of compliance.  On the opposite pole is Transsion (688036.CH).   Transsion  is probably the most well-known for the price cutting tactic mentioned above, as the company dominates the low-end market in China and Asia, but has the ability to rein in its aggressive pricing nature to capture this more draconian subsidy program.
​
Conclusion: Engineering the Upgrade
The shift from 2024’s volume-based subsidies to 2026’s "Future Tech" and "Silver Economy" mandates represents more than just a retail stimulus; it is a calculated attempt to force-march the Chinese consumer into a domestic, AI-integrated ecosystem. By tying financial incentives to high-threshold hardware specs, like the 12GB RAM floor and dedicated NPUs, Beijing is effectively de-platforming the "low-quality" growth that defined the previous decade.
For domestic champions like Huawei and Xiaomi, these rules provide a subsidized path to premiumization. For foreign players like Apple, the path is narrower, requiring a delicate dance between maintaining global brand margins and meeting increasingly localized, "fuzzy" AI compliance standards.
Ultimately, the "New for Old" program's success will be measured by whether it can truly break the cycle of "Involution." If the government can successfully pivot the middle class from "buying cheap" to "buying smart," they may finally solve the deflationary puzzle that has haunted the post-2017 market. However, with the central government now bearing 90% of the fiscal weight through ultra-long treasury bonds, the stakes are high.  Beijing is not just subsidizing a replacement cycle; it is betting the national balance sheet on the hope that AI and the Silver Economy can reignite a cooling economic engine.
As we move into the second half of 2026, the rollout of "AI Handset Standards 2.0" will be the ultimate litmus test. It will determine if China can create a self-sustaining, high-value tech market, or if it has simply replaced a "race to the bottom" with a high-priced, state-funded "race to the top."
 


[1] The official Chinese government definition of “Silver Economy” - "The sum of economic activities that provide products and services to senior citizens and prepare for the challenges of an aging population."

Yesterday Samsung Electronics (005930.KS) officially announced their latest flagship line of Galaxy smartphones, the Galaxy S26, the Galaxy S26+, and the Galaxy S26 Ultra.  Smartphone release announcements are typically a parade of incremental hardware updates and changes, culminating in the ultimate question, did the price change, and if so, is it worth it to upgrade?  While the price question always remains (Only the Ultra increased in price by $50 – The S26 and the S26+ remained the same), this year’s release, usually a highly focused bally-ho over changes that are almost negligible to most smartphone buyers, was oriented differently, which we see as a significant change.
Under the Hood: Moving Matrix Math to the Center
There was still plenty of focus on hardware changes, with much attention given to the upgrade to the chipset in the Galaxy Ultra.  While the Galaxy S25 Ultra sported the Qualcomm (QCOM) Snapdragon 8 Elite chipset (Technically the 8 Elite Gen 4), this year’s Galaxy Ultra moved up to the Qualcomm Snapdragon 8 Elite Gen 5 chipset,  This solved a heating issue that plagued the S25 ultra, but more significant was the fact that Qualcomm moved the hardware that performs matrix math (The parallel processing used by large language models) from a combination of the NPU (Neural Processing Unit) and the GPU (Graphics Processing Unit) to the CPU (Central Processing Unit) itself.  This resulted in much faster responses to AI requests, both internal and from the user.  In itself this was a plus, but most important was the focus on AI rather than more visible hardware improvements.
From Reactive to Proactive: The Rise of "Nudge"
This focus on AI continued, particularly with the S26 Ultra.  In most AI phones, when you need the AI to help with a particular function or application, such as when when someone asks for a copy of an image, you ask the AI to send it to that particular person with a query.  In the S26 Ultra Samsung has added “Nudge”, best described as an AI agent.  Now when someone asks for a copy of an image a ‘”nudge” icon automatically appears on the photos and a simple click on the icon allows the agent to take over the process of mapping out the steps needed to send the image to the person with whom you are messaging, and executes the process for you.  This would be called “Proactive AI” rather than “Reactive AI”.
The ”Proactivity” travels deeper as it is part of the Samsung One UI 8.5 operating system (Based on Android 16).  This means that instead of being a typical “Wait-and-See” system that remains idle until you click on an icon, it is constantly observing your local context in order to predict your next move.  The OS framework also contains “Hooks” that allow Gemini and Bixby to “click” buttons and navigate through applications, similar to agentic engines like Comet. 
Deep Integration: The Ecosystem "Hooks"
This “Nudge” system is able to hook into eight core internal applications:

• Gallery – The system makes proactive suggestions concerning specific photos when the context asks for them.
• Calendar – Offers to create events or check for conflicts when it sees a date or time on your screen.
• Contacts – Pops up contact cards or phone numbers during conversations.
• Reminders – Suggests setting an alert if you are discussing a deadline, project, or task.
• Smart Things – Offers to control home devices when things like “Coming Home” or “Sleep” are mentioned.
• Quick Share – Recognizes people in your photos and suggests sending them the file directly.
• Notes & Calculations – Brings up relevant notes or formulas that relate to what you are currently typing or viewing.
• 3rd Party Applications – Unlike competing “nudge-like” systems (Primarily from Google (GOOG)), as long as you are using the Samsung keyboard the Nudge system can hook into many 3rd party messaging application, with support already confirmed for WhatsApp, Instagram DMs, Telegram, Samsung Messages, and Google Messages..
• Other “Hooks” – Not only can Nudge make suggestions concerning 3rd party applications but is also able to perform tasks (Multi-step workflows), such as booking Uber (UBER) directly from a Nudge pop-up.  Throughout this year Samsung will be announcing further integration with Instacart (CART), DoorDash (DASH), and Amazon (AMZN), giving Nudge the ability to automatically fill a cart with items in a recipe you are looking at or source and purchase items on Amazon based on an image you are looking at on your phone.

The Friction Points: Privacy, Power, and "Clippy" Fatigue
While “Nudge” can be a time saver, there is some downside.

• Privacy – While the data Nudge uses is local to the Snapdragon chipset, some will be uncomfortable that the system is “reading” everything they write, including private messages, e-mails, and notes.  Discussions concerning legal or medical issues can still push Nudge to make what might be embarrassing suggestions if you are sharing the phone with someone else.
• Overload – Remember Microsoft’s (MSFT) “Clippy”, the virtual assistant that popped up on your screen?  The fact that Clippy is gone indicates that an overly aggressive AI could easily become an annoyance.
• Reliability – As Nudge must resolve conversational context, it can lead to errors for dates or incorrect photo suggestions, and more disturbing hallucinations are also a possibility.
• Power – The Qualcomm Snap 8 Gen 5 chipset is 35% more efficient, but even so, with Nudge on, the NPU is active at all times.  Some users have indicated that turning off Nudge gives an additional 45 to 60 minutes of screen time before recharge, so its use will drain the phone’s battery faster than normal.
• Lock-in – As the Nudge system is based on the Samsung keyboard, Nudge users are limited to Samsung devices, eliminating GBoard (Google) and Swiftkey (Microsoft).

As with any new system, consumers will have to decide how much it helps versus the drawbacks, but what is most significant to us is the fact that Samsung, known for its hardware expertise, seems to have shifted its focus to software, something we have been championing for some time.  The ability for smartphone brands to make significant hardware changes that push consumers to upgrade has diminished over the last few years as competition has increased.  This has continued to push  consumers to hold their phones for extended periods, leading to little growth in the smartphone market.  

​The Economics of Intelligence: High Margins over High Capital
Software does not have the high capital cost of hardware manufacturing, and consumer adoption to AI has given smartphone brands an opportunity that they have not had in a number of years, and Samsung seems to have taken the opportunity in a big way.  Google (Pixel smartphone line) has also been a proponent, as have Honor (pvt), Oppo (pvt), Vivo (pvt) and Apple (AAPL), each in its own way, although some are more hands on, meaning you have to invoke the Nudge-like application before it will perform.  The idea that software can be as much of an attraction to consumers as hardware is certainly not a new one, given estimates between $550 billion and $670 billion for the mobile software market, but brands have always relied on hardware as the selling point for new smartphone models.  The shift to software will test the mettle of brands that have relied on hardware for years, and if the brand enthusiasm carries through for another year or so, the change in perspective will represent a substantial shift in focus for smartphone brands.
Conclusion: A Catalyst for a New Mobile Era
In conclusion, the launch of the Galaxy S26 series represents more than just a new product cycle.  It signals a fundamental restructuring of the smartphone industry’s DNA. By prioritizing Proactive AI and agentic software over incremental hardware specs, Samsung is acknowledging that the "Hardware Era", defined by megapixel counts, multiple cameras, and screen resolution might have reached its peak.
This shift to an "Intelligent OS" model addresses the industry's greatest existential threat,  the stagnating global replacement cycle. As consumers hold onto devices for nearly four years, the value proposition must move from the physical device to the utility it provides. By integrating "Nudge" directly into the kernel of the OS and leveraging the Snapdragon 8 Elite Gen 5’s unique CPU architecture, Samsung is attempting to transform the smartphone from a passive portal into an active, autonomous partner.
If successful, this pivot will force a market-wide recalibration. Competitors will no longer be judged solely on their ability to manufacture hardware, but on their ability to develop deep software "hooks" and agentic ecosystems. In a software market worth upwards of $670 billion, the S26 Ultra isn't just a phone, it is a manifesto for the next decade of mobile computing, one where the "killer app" is the operating system itself. The industry is moving from selling metal to selling ease of use, and Samsung’s proactive turn may well be the catalyst that defines a new era.
 

The Great Digital Divide
We live in a divided world where polarization has become the norm and an open mind the exception.  While politics is the usual playing field for such a divide, the technology world has supplied us with a new one over the last few years, AI.  Some, particularly those developing the technology, believe AI is unlike any other technology advancement and will change the way humanity will evolve for centuries to come.  Others believe it is a scourge that will push humanity to a secondary position and eventually morph the human race into a herd of narcotized sheep.

The Third Perspective: AI as a Utility
But there is another contingent.  A group that looks at AI as nothing more than a tool that can assist mankind in its endeavors to both for the positive, to forge a better world or to the negative, to gain power over his/her fellow man, things that we have been doing since the dawn of man.  Having used AI on a daily basis since its early days of commercial existence, we fall into this middle camp and look at Ai as a “Tool with benefits” rather than the  dualistic earth-shaker that supporters of both camps seem to see.  Perhaps this is because, in our daily work, we see the points where AI can be successful and the places where it is not, but at the same time we have spent time conversing with AIs about this very question, and the AIs we have spoken with point out their good and bad points quite easily.

Defining the Nature of a Tool
Tools are devices that help mankind perform tasks.  They can be quite simple, as were stone cutting tools that helped man cut wood to build shelters or more complex tools like the plow that allowed scalable agriculture to develop.  But with the development of each tool the lives of the population were fundamentally changed.  Tools perform a function, usually a simple one that is fixed at the point when the tool is produced.  Some tools can do more than one thing, as a hammer can pound in nails and can pull them out, but it cannot adapt and perform other functions without being redesigned and rebuilt.

The Proponent’s Argument vs. Reality
Proponents say that AI is different.  It can adapt and change.  It can reason and learn, and above all, AI is probabilistic. making it more than a tool.  However, when we look at what AI systems are and how they work, it becomes more obvious that they are tools, sophisticated tools, but tools none the less.  Ai designers point to the fact that AI systems are robust, that they are able to make sense of complex data even when it is ‘noisy’ and includes irrelevant information and are able to build on their knowledge, but at least now, some of those claims are less than true.

Why AI Still Fits the Toolbox
Our goal is not to denigrate the development of AI or cast aspersions on those involved in their commercialization, but merely to point out that while AI is a valuable tool that we see as a major step forward, it is still a tool and while more adaptable than many tools, it still falls into the tool category and not the god-like creation some say it is.  Here is why:

• Usefulness – Most tools are able to function as soon as they are constructed.  AI tools only become useful when they are trained and are only as useful as the data on which they are trained.
• Perception - You cannot “trick” most tools as they have no perception of their use.  While some say that AI is perceptive, you can “trick” an AI system because its understanding is based on pixels and math, not on a real-world understanding of context.  Adversarial Machine Learning has proven that adding “noise” to an image, which is invisible to humans, can cause the AI to see a toaster as a school bus, as, in reality, the Ai is not “seeing a toaster or school bus, it is ‘calculating” a toaster or school bus.
• Fragility – There is a concept called Distribution Shift with AI systems.  The illustration would be if you teach an AI to sort red apples in a bright room and then show it red apples in a dark room, it might decide that the object is black coal.  Unlike other tools, the AI does not stop functioning, it gives a confident wrong answer.  This occurs because its knowledge is tied to a specific statistical distribution that it was taught, not the underlying reality of the object.
• Learning & Forgetfulness – Traditional tools are additive.  If you add a handle to a tool it does not forget how to work.  If you try to teach an existing AI something new, such as trying to teach a translation AI to do complex math, the AI can forget how to translate .  This is called Catastrophic Interference and has been a big stumbling block in moving AI from being a specialized tool to the general intelligence category.  This stems from the fact that AI systems do not store data in folders, like we do on computing devices.  They represent the connections between AI neurons as weights that represent how strong each of those millions of connection are.  When an AI system learns to recognize a cat, all of those weights relate to defining “cat-ness”.  If you try to teach that AI to also recognize dogs, the process begins to change some of those shared weights to “dog-ness” and the AI is less able to understand “cat-ness” and can often forget the original “cat-ness” altogether.

The "Update" Illusion
wait, we are often told that the AIs we are using are learning all the time.  In fact, Ai designers get around the forgetfulness issue by using RAG (Retrieval Augmented Generation), opening up the AI to a search engine.  The AI is then able paste the search data into its response, although the AI has not incorporated the search data in its weightings.  The ‘learned’ the data is only us to augment that particular answer and is deleted along with the chat when it is over.  There are other tricks that designers can use, but AIs need to be completely retrained in order to learn a new function.

Architectural Instability: The Shared Space Problem
The forgetfulness issue is a big one because it points to a major difference between AI systems and humans and what further points to the view that AIs are tools, not a god-like creation that will eventually overtake humankind.
AI system share space.  When an AI system learns about something the entire network shifts.  Teach it something else and the AI uses that same network for the new data.  There is no ‘reserved space” for the earlier data, so with each addition the AI  sees less of the original data and more of a conglomeration.  Sort of how using a dark marker in a notebook bleeds through to the next page.  Eventually is becomes a mess.

The Biological Advantage: Human Consolidation
Humans consolidate data.  We move new information to a temporary point in the brain and during sleep, that information is moved to a long-term storage location.  During that move the brain locks down information that it deems important so it cannot be overwritten.  This mechanism allows use to maintain old memories while making room for new ones,
Batch Processing vs. Sequential Wisdom
Further, AI systems learn in batch mode, essentially all at once.  In order to add to thet knowledge everything previously learned must be learned again along with the new data.  Humans are sequential learners. This is because the brain weights information by both importance and emotion, something AI tools cannot do as they have no emotional context.
​
Conclusion: The Hand, Not the Replacement
When we pull back the curtain on the "black box" of AI, we don’t find a digital god or a sentient rival; we find an incredibly sophisticated reflection of our own collective information. The very things that make AI feel "broken" to some—its fragile memory, its need for "tricks" like RAG to stay current, and its tendency to confidently misinterpret a dark room—are actually the boundary lines that define it as a tool.
AI doesn't have to be a competitor for the human throne to be revolutionary. Just as the telescope didn't replace the eye but allowed it to see further, and the engine didn't replace the leg but allowed it to move faster, AI is an extension of our ability to process a world that has become far too data-heavy for the naked mind. It is a "Cognitive Swiss Army Knife" that can pop open the right blade for almost any task, but it still lacks the calloused hands and common sense of the person holding it.
By stepping away from the "all or nothing" hype, we can appreciate AI for what it really is: a remarkable, albeit slightly temperamental, achievement of human engineering. It’s a better mousetrap, sure—but one that still needs a human to decide where to place it and what's worth catching. In the end, the most impressive thing about AI isn’t how the machine "thinks," but how we choose to use it to work, create, and build a better version of the world we already live in.

The Fourth-quarter Catch-Up: An Accounting Signal of Pricing Shifts
Universal Display (OLED) reported fourth-quarter sales that were in-line with consensus and slightly better than our estimate of $171.7 million.  In that number was a cumulative $10 million catch-up adjustment, roughly double that of fourth quarter 2024.  We believe the adjustment and the company’s conservative 2026 guidance ($650 to $700 million), have some significance to each other and reflect the unusual state of the OLED industry this year.  While we can spend time detailing UDC’s fourth-quarter and full year 2025 results, we see this “connection” as an indicator of the company’s current view of the OLED space and the general status as to the development of a blue phosphorescent emitter.
The 4th quarter conference call sent a few “smoke signals” as to how UDC management is looking at the 2026 year, particularly the ASC 606 adjustment that was made in 4th quarter, and the mention of the increasing price of iridium, the heavy metal used in many of UDC’s phosphorescent materials. 
The Iridium Factor: Pass-Throughs and Supply Chain Volatility
The price of iridium remained relatively stable through the first half of 2025 but began to climb in 3Q and spiked in 4th quarter.  Iridium is one of the rarest metals on earth and is typically produced as a byproduct of platinum mining, making it unusually sensitive to disruptions.  As Russia produces between 10% and 15% of the global supply, when the government of Russia began a ‘reclassification’ of rare earths and strategic minerals in November of last year, in response to Western sanctions, panic buying began.  In December, the EU began developing new sanctions against Russia that included Russian metal imports.  This, combined with both the Russian sanction response and some 4th quarter demand[1], caused the price of iridium to rise by almost 45%, as shown below (Units=oz.).
12/31/24           $4,200
03/31/25           $4,250
06/30/25           $4,200
09/30/25           $4,600
12/31/25           $6,650
While we believe that the emitter materials UDC produced in fourth quarter were synthesized  with iridium purchased about six months earlier at lower prices, we expect the impact of the iridium price hike seen in fourth quarter 2025 will be affecting materials produced by UDC in 2H 2026.  Given the material cost escalators built into the company’s supply contracts, we expect UDC will increase material prices  in second half, accounting for the company’s comments concerning 2026 being a more “normal” year, with a stronger 2nd half. 
Revenue Recognition Mechanics: The "TETP" Early Warning
That said, companies that operate under ASC 606 must reassess each customers contract model on a quarterly basis, and the fourth- quarter hike in iridium prices likely caused the ~$10 million catch-up that UDC recorded in 4th quarter, roughly twice what it was in 2024.  We see this near-term revenue recognition boost as a bit of an “early warning” of a price increase in 2026.  The reassessment of the ASC 606 model reflects the higher cost in fourth-quarter, but more importantly points to the pass-through of the iridium price hikes later this year. 
Shifting Fundamentals: Material Intensity vs. Royalty Revenue
This is also confirmed by the lower gross margin guidance for 2026 given by management.  This indicates higher revenue but revenue that just covers the higher material cost and does not add to the bottom line as would existing materials produced after the price hike.  This lower gross margin guidance also jives with the 1Q 2026 company guidance on UDC’s material-to-royalty ratio (1.3:1), which implies that 2026 will be a year more heavily weighted toward material sales, which carry a lower gross margin than license/royalty revenue.  The cause for this shift could be the higher material ASP caused by the iridium price hike, but, on a more positive note, it also could be a result of the increased use of tandem emitter material stacks, which imply higher material sales per device unit leading to a lower material to license ratio.


[1] Iridium is a catalyst for Green Hydrogen PEM electrolyzers

​Divining the causation for management’s conservative guidance, the implications of the ASC 606 catch-up, and the modified material/license ratio lead us to focus more on the ASC model that governs how companies recognize long-term contractual revenue.  The basis for the model is the relationship between actual material sales and the recognition of license/royalty revenue.  The rule ties the recognition of license/royalty revenue to material sales over the life of the contract, using the estimated value of the contract as the basis that is adjusted each quarter.  Therefore license/royalty is recognized as material is sold on a per-gram basis rather than when it is received.  This makes quarterly results highly sensitive to changes (such as material prices or new architectures) in the “Total Estimated Transaction Price” of a particular contract.  As these changes affect the TETP the company is required to make positive or negative catch-up adjustments that reflect how the full five-year contract value has changed over the period.
The 2026 "Puts and Takes": Analyzing the Bridge to 2027
Based on the factors above and others mentioned on the call, here are the puts and takes we see for 2026.
Positives
Tandem Architecture – As we have mentioned in previous notes, tandem OLED architecture leads to the use of higher material volumes per unit.  As this architecture becomes more common, the overall rate of material consumption on a per-unit basis will increase, causing the value of a long-term contract to increase
Cash – With $955 million in cash (~17% of market cap) and little need for major capital expenditures, the company increased its dividend in fourth-quarter.  While not a dividend play by any means, it gives investors a cushion when waiting for an incremental step up in sales.  The company’s cash and investments alone generated ~46.4% of the 2025 dividend payout.
Neutral
Blue – UDC management made few comments about the status of blue phosphorescent material development, other than the standard ‘…making progress…”.  We believe that blue development is a mixed bag for UDC in 2026 as the adoption of current blue phosphorescent materials faces a more competitive picture than only a year or so ago.  Tandem and multiple layer structures are giving OLED producers the ability to capture some of the power, efficiency, or lifetime advantages of blue phosphorescent emitter materials using materials that are more mature and readily available.  This puts blue into a more discretionary adoption mode, where OLED producers can take time to evaluate the physical advantages of the material against the cost of adoption, leading to a more conservative  adoption profile.  That said, when it is adopted, there will be a substantial increment in sales over time and the value of those contracts that include blue will be stepped up.
Negatives
Gen 8.6 adoption – Will not much of a surprise, the commissioning of Samsung Display’s Gen8.6 IT OLED fab in second quarter and a similar fab from BOE (200725.CH) likely in third quarter, will contribute relatively small amounts to sales this year relative to expectations in 2024 and early 2025.  There is nothing holding these fabs back, but as we have noted in the past, the ramp toward profitable yields and utilization takes time, with the bulk of the incremental sales from these fabs more likely seen in 2027.
Leverage – Under the company’s conservative sales guidance and Operating expense growth of 7% to 9% *R&D and SG&A), expenses will grow faster than revenue leading to negative operating leverage.  While this is a negative we expect that behind the conservative guidance the company is a bit more optimistic about sales prospects for the year.  That said, given the fact that we are only a month and a half into the new year and the technology space has already faced some major upheavals (tariff changes), we expect guidance to improve as there is more clarity on a number of the issues we mentioned above, bringing the leverage back to the positive. 
Conclusion: A Year of Operational Calibration
The intersection of aggressive raw material inflation and the strategic shift toward high-area IT displays has moved Universal Display into a period of significant operational calibration. The $650 million to $700 million guidance and subsequent margin compression reflect a defensive posture in the face of iridium price volatility, but also widens the potential upside as clarity improves across the 2026 OLED space.  UDC’s growth thesis is currently a race between two competing forces: the "mechanical" revenue lift provided by ASC 606 catch-ups and contract escalators, and the structural "purgatory" of waiting for Gen 8.6 fabs and Blue Phosphorescent  Emitter adoption to reach commercial maturity.
The path to upside is clear but narrow. If the industry can successfully navigate the "yield tax" of Tandem architectures and the initial yield and utilization issues of new fab ramps, UDC stands to benefit from a significant increase in material intensity per device. However, this potential remains tempered by a negative operating leverage profile and a material-to-license ratio that currently favors lower-margin physical sales over high-margin IP royalties.
Ultimately, 2026 is less about a breakout and more about the durability of UDC’s ability to navigate a difficult year. With a $955 million cash cushion and a dividend program that is nearly 50% subsidized by interest income, UDC is better positioned than most to absorb the "unusual state" of the current market. Investors should view the coming year as a bridge: if management can maintain cost discipline while these headwinds persist, the eventual normalization of raw material prices and the scaling of next-gen fabs in 2027 could turn today's conservative guidance into a foundational floor for the next leg of growth, with “Blue” being a major kicker.

The Pre-Launch Pricing Vacuum
Detail is not particularly necessary to indicate Samsung’s premium TV pricing thus far into the year..  Most categories have seen price increases since year-end 2025, ranging from  up 0.4% to up 4.7% by category.  On an unweighted basis the Mini-LED/QD segment, which includes both 8K and 4K Mini-LED/QD TV sets, did see a decline of 0.4%, which came entirely from Samsung’s 8K line, while the much large 4K line saw the 0.4% price increase mentioned above.  In the OLED TV line only the low end category saw a decline since year-end, while all other OLED set categories saw increases.
Segment Breakdown: Divergent Trends in Mini-LED and OLED
Looking at the 2025 line by category for both year-to-date price movements and lifetime prices gives a better indication of where 2025 TV set prices are on a relative basis.

​We note that almost all TV set models reached 2025 lows on or around the Black Friday holiday and rose in price through the Christmas and New Year holidays, so it is of no surprise that most have continued to increase in price since year end.  As new models are typically released in late March 2025 models are in a no-man’s land of pricing as there is little for consumers to compare against.
The 2026 Outlook: Compounding Cost Pressures
That said, we expect upward price pressure on new models, or perhaps similar pricing to last year’s initial prices but a slower price decline to offset increased memory costs.  Since the tariff rules have changed once again, much CE pricing for 2026 models is in limbo, but unless the current administration acquiesces to the will of the Supreme Court quickly, all bets are off until a clearer tariff picture is found.  Therefore we expect more TV set price movement based on inventory levels than cost and import/export issues until the administration decides what path it will take on tariffs.
Conclusion: A Strategic Stalemate for Premium TV Pricing
The current state of Samsung’s premium TV pricing is best described as a "strategic stalemate" driven by three conflicting forces: seasonal inventory clearing, skyrocketing component costs, and a volatile regulatory environment.

• The 2025 Inventory Floor -The data confirms that the "price floor" for the 2025 lineup was established during the Black Friday window. The subsequent YTD increases in most categories, particularly the 4.7% rise in High-tier OLEDs, indicate that retailers have successfully transitioned from aggressive liquidations to a "harvest" phase. With the 2025 models now in a "no-man's land" ahead of the 2026 launch, current pricing reflects a lack of competitive pressure rather than a surge in demand.
• The "DRAM" Factor - While 2025 models are drifting, the 2026 lineup faces a significant cost-of-goods headwind. The DRAM crisis, driven by the AI-fueled surge in DRAM and NAND demand, has caused memory costs to more than quadruple. For a premium Smart TV, which relies heavily on these chips for AI processing and upscaling, this translates to a ~5–7% increase in the total Bill of Materials (BOM). Consequently, consumers should expect the initial MSRPs for Samsung’s 2026 "TV sets to be set higher (more likely) than last year’s launch points to protect manufacturer margins unless they decide to bite the bullet and maintain comparative initial pricing  but reduce discounting momentum going forward (less likely).
•  The Tariff Wildcard - The February 20, 2026, Supreme Court ruling (striking down IEEPA-based tariffs) initially suggested a potential reprieve for consumer electronics. However, the administration’s immediate pivot to Section 122 tariffs (imposing a 15% duty effective February 24) has effectively neutralized any near-term price relief. This "tariff ping-pong" ensures that inventory management, rather than fundamental cost reduction, will remain the primary driver of retail price movement, likely through the first half of the year.

Final Outlook
For the consumer, the window for "value" has temporarily closed. With 2025 models rising from their holiday lows and 2026 models poised to launch at a premium, inflated by both memory scarcity and renewed tariff uncertainty, the market is likely to remain in pricing limbo until the secondary discount cycle begins in late second quarter.

The Phablet Paradox: When Is a Phone Not a Phone?
We all know what a smartphone is, or at least we think we do, but while foldable might have changed that definition a bit, there are some very specific delineations that describe the differences between smartphones and other device types.  As far back as 2010 there were questions as to whether 7.0” tablets from Samsung Electronics (005930.KS) and ASUS (2357.TT) were actually tablets or phones as the manufacturers gave them the ability to make cellular calls (“P” designation in the model number) but were theoretically tablets.  These devices, later called “Voice Tablets” pushed the screen size to 7” or greater, making them oddities at a time when smartphone screen sizes were
The 75mm Hard Limit: The Biological Reality of the One-Hand Rule
That said, there is a theoretical limitation to screen size for smartphones called the “One-Hand Rule”.  Smartphone designers break down the surface area of a smartphone (right-handed) into three areas.  The green zone arc is where the thumb typically rests and is easy to access.  The yellow zone is reachable but usually requires a stretch or what is called a “Palm Shuffle” to move the phone slightly.  The red zone, over 25% of the display, is inaccessible on phones over 6.9” (Diagonal), needing a second hand or requiring the phone to be pushed down far enough that it might slip.  Older phones with displays under 5” had a “Green Zone” that covered ~90% of the display.  A 7” display has a “Green Zone” of between 25% and 30%.
The width of the display is also a factor as you must be able to keep your fingers wrapped around the back of the phone while the thumb is moving.  The limitation here is ~75mm (~3”) as a wider display causes your hand to grip so tightly that your thumb becomes restricted and the green zone shrinks.
Swords and Software: Engineering a Way Around the Human Hand
Manufacturers have found was around these limitations, primarily by changing the aspect ratio of the screen.  Sony (SNE) changed the aspect ratio for its Xperia smartphone line, using a 6.5” display, to 21:9 from 2019 to 2023 to give users the larger screen (6.5” was large then) while still allowing the “One Handed Rule” to be maintained.
More recently, to compensate for even larger displays, smartphone manufacturers turned to software.  Both the iOS and Android have applications that shrink the display when one swipes the bottom of the screen, allowing the thumb to reach the upper screen and Samsung has taken it even further by moving all interactive functions to the bottom of the screen, making them accessible to the thumb.
The "Large Grape" Advantage: The Hidden Power of Cubic Volume
While there are a number of ‘fixes’ that help users gain one-handed access, the physical limitations mentioned and others like pocket size and structural issues (larger screens are more liable to snap and need more internal support) have kept most smartphone displays below 7” and within a 77mm wide and 165mm high sweet spot.  That said, it turns out that  a number of smartphone manufacturers and examining the  idea of new models that have displays above the 7” mark.  Currently the few that exist (2025 – 2026) tend to be niche brands, although the Huawei pvt) Mate 70 Air, with its 7” display is the only current popular brand over the 6.9” limit. 
As brands are always struggling with differentiating their phones from the vast number of competitive devices, any marketing hook can bring in  a few more customers, but there are other reasons.  While it seems minor, the additional space created by a 7” screen over a 6.8” screen (same width and thickness) is 3.7 cm3, and while that sounds small, it is the size of a large grape or a cherry and represents a 3.6% increase in smartphone volume.  If that still sounds small, a smartphone engineer can increase the battery capacity in that situation from 4,300 mAh to 5,000 mAh, enough to bring the phone’s battery from ‘almost a day’ to ‘marathon’ level.  It can also be used to double the size of the phone’s vapor chamber (cooling), allowing the phone to last an hour before it throttles down to cool, as opposed to the usual 10 minutes, or it could be used for an entire rear camera system, including sensors and lenses.

​Conclusion: The 7-Inch Paradox
The evolution of the smartphone display has reached another interesting physical impasse. While the "One-Hand Rule" and the 75mm width limit define the boundaries of human ergonomics, the allure of the "Large Grape" volume, that extra 3.7cm3, is proving too tempting for manufacturers to ignore. This marginal increase in size represents a massive engineering victory, providing the breathing room needed for "marathon" batteries and superior cooling systems that 6.8-inch frames simply cannot house without compromise.
As we move through 2026, the industry appears to be splitting into two distinct philosophies:

• The Foldable Solution: Traditional giants like Samsung and Apple (AAPL) are increasingly offloading the 7-inch-plus user experience onto foldables, using the "Book-style" form factor to bypass the pocket-leverage and thumb-reach issues entirely.
• The Candy bar Defiance: Niche players and brands like Huawei (via the Mate 70 Air) are doubling down on the "Mega-Screen" candy bar. They are betting that users will trade one-handed usability for the sheer utility of a device that finally merges the power of a tablet with the ubiquity of a phone.

Ultimately, the move toward 7-inch displays is less about the screen itself and more about the internal real estate it unlocks. Whether consumers will accept these workarounds as a permanent part of the mobile experience remains to be seen, but for now, the smartphone is officially outgrowing the human hand.

There are two kinds of displays.  Those that use a light source to illuminate colored materials and those that generate the colored light themselves (self-emitting).  Material scientists are constantly working toward improving these materials, particularly the self-emitting group (OLED and quantum dots) as they must compete directly with non-self-emissive materials used in LCD (Liquid Crystal Display) devices, typically based on LEDs which are known for their brightness.  However the pace of material science improvements does not lend itself to the more aggressive competitive pace of the consumer electronics industry and material scientists and display engineers find themselves looking for ways to improve self-emitting display performance more quickly.
Anatomy of the OLED Stack: A Microscopic Viral Frontier
Back roughly 15 years ago the ‘stack’ of OLED materials that was used to generate images consisted of two factions.
Small Displays – These devices, primarily smartphones, were built on an OLED stack consisting of a red phosphorescent material, a green fluorescent material (later this changed to phosphorescent material) and a blue fluorescent material, accompanied by a number of other ‘helper’ materials, along with an anode and a cathode.  Each pixel contained the same structure.  This was the typical configuration for small displays, primarily smartphones and tablets.  These displays were produced in a vacuum chamber where the materials are sublimated (heated, turning them from a solid to a gas with no liquid state) and flow onto the substrate (typically glass or polyimide (plastic) on glass, with the glass later removed.  This deposition of materials is done through a metal mask (think very fine screen) that creates the dots of material that become pixels on the screen.

Large Displays – As the metal masks used in small displays were not able to scale up in size (they sag), LG Display (LPL) developed an alternative.  They coated the entire substrate with a blue fluorescent and a yellow/green phosphorescent material.  When activated this produces a white light.  That light passed through a color filter, a sheet of very small red green and blue phosphor dots, which create the colors you saw on the screen.

Over the years pixel layouts, materials, and processes changed and improved, but the basics remained the same until 2021 when Samsung Display (pvt) announced that it had developed a commercial large panel OLED display using both OLED materials and quantum dots.  Quantum dots are molecules that can exchange the ‘color’ of light for another color.  In Samsung’s case blue OLED material produced light that was converted into red and green light (clear space for blue) and a 3rd OLED process became available.
Understanding the Materials
The average display user knows little about the inner workings of displays, and in many cases investors only look at large scale production and shipment volumes to gauge the status of the industry, but behind those numbers are some details that can give indications as to more subtle trends that can become significant over time.  In order to understand that nuance, one needs a more detailed understanding of the materials in an OLED stack
The stack in an OLED display is made up of a number of materials, each with a specific function:
Cathode – Typically a Magnesium/Silver or Aluminum alloy.  In top emitting OLEDs the cathode is transparent and is between 10nm and 20nm thick.  In bottom emitting OLEDs the cathode can be up to 100nm thick.  This is one side of the electrical connection that causes the OLED materials to emit light (the negative side).
Electron Injection Layer – Typically Lithium Fluoride or Ytterbium between 0.5nm and 2nm thick.  Acts as a ‘door’ to help electrons jump from the cathode to the next layer.
Electron Transport Layer – Typically an organometallic aluminum compound[1], roughly 20 nm to 40 nm thick. Essentially a highway for electrons.
Emissive Layer – Materials in this layer, either fluorescent or phosphorescent, are the ones that create the ‘glow’ of the display.  Depending on the material, this layer is between 20nm and 40nm.
Hole Transport Layer – Typically TAPC[2] roughly 20nm to 60nm thick. A highway for holes (positive charge)
Hole Injection Layer – Typically HAT-CN[3] - That same ‘door’ for holes
Anode – Typically Indium Tin Oxide (ITO) or silver, between 50 and 100nm thick
This means that those little ‘glowy’ things on your TV or phone are roughly the thickness of a Flu virus, with each performing a task that enables the emissive material to produce as much light as possible.
Note:  sometimes OLED display producers put ‘barrier or blocking layers on either side of the emitting material to make sure electrons and holes don’t overshoot their marks.
 
The "Cubed" Constraint: Why Thicker Isn't Always Brighter
Since all of these materials work toward making emitters produce light, as electrons and holes meet in the center of the stack, why not just add more materials to make OLED displays brighter?  There are some specific reasons:

• Organic (OLED) materials are sensitive to degradation, but that degradation is not linear. A particular law of physics shows that the current density, the voltage and the material thickness are interrelated using the following formula  with L3 being the thickness cubed.  Therefore, as the thickness is the denominator, a small increase in thickness requires a very large increase in voltage, essentially doubling the thickness increases the necessary voltage by 2.8 times (constant current). which would crystalize the organic material.  If we leave the voltage the same while doubling the thickness, the current falls below the point where the OLED would turn on..
• Additionally, as the thickness increases, the electrons and holes have difficulty getting through the material and pile up as they enter the emitter material.  This requires even more voltage to push them through, so if the heat from the increased voltage didn’t degrade the material, the additional heat needed to push them through at this point would.

The Tandem Solution: The Efficiency Relay Station
As competitive LCD displays are known for their brightness, OLED material scientists sought a way to close the brightness gap.  Faced with the inability to gain brightness by using thicker emitter material, they came up with a ‘tandem’ solution.  Instead of thickening the EML (Emitter Layer) they improved the architecture with a special layer, called a charge generation layer (CGL), between two thin emitting layers.  The CGL acts as a relay station, allowing electrons passing through the 1st EML to ‘retire’, while creating new electrons for the second EML layer.  This trick keeps electrons from building up and clogging the system. 
There is a cost in terms of higher voltage needed for the two layers and the CGL itself, but the current remains the same as in a single layer and that keeps them from being destroyed.  While the higher voltage is a ‘cost’, it is far less than the ‘cubed’ voltage of using a single thicker EML.  As the EML/CGL layers are wired in series, a single stack that used 4 volts would need 8 volts for a tandem device, 12 volts for 3 layers, and so on, but the current stays the same, and current is the actual killer of the OLED material.  Samsung has latched on to tandem technology and is now able to claim vastly improved device specifications under this architecture.


[1] Tris(8-hydroxyquinolinato)aluminium or Tris(8-hydroxyquinoline)aluminum.

[2] 1,1-Bis[(di-4-tolylamino)phenyl]cyclohexane

[3] Hexaazatriphenylenehexacarbonitrile

The Tandem Solution: The Efficiency Relay Station
Using the tandem theory, it would make sense for OLED display producers to keep adding additional layers, which, in theory, would produce brighter OLED display.  Of course there are tradeoffs; the cost of the additional materials, the cost of processing the extra layers, the light loss from the additional layers[1], the lower yield (more layers more chance of issues and higher cost of ‘bad’ stacks), and more difficulty keeping the stack cool.  However keeping those tradeoffs under control has been the goal of OLED display producers, and while they are primarily used in high-end devices where the BOM is a bit more flexible, the goal is for bringing tandem OLED to lower-priced devices.  Apple (AAPL) has used tandem displays for the iPad Pro, Huawei (pvt) has used them for their high-end smartphones, as has Honor (pvt), while OLED display producers Samsung Display , LG Display, BOE (200725.CH), Visionox (002387.CH), and Tianma (000050.CH) all produce various tandem displays.
LG Display, the producer of WOLED large panel displays also uses a form of tandem display stack for their high-end OLED TVs.  LG’s typical OLED stack (Figure 4) consisted of a blue fluorescent emitter, a yellow/green phosphorescent emitter, and a second blue fluorescent emitter. that when combined produced a white light.  However, In order to maintain ‘high’ brightness, LGD added an additional sub-pixel that allowed the white light to also shine through as a 4th ‘color’ (RGBW)  While this improved brightness, it also washed out some of the overall display color.  They have since pushed that stack to higher levels by changing the yellow/green layer to separate red and green layers, which they say increased brightness by 33% and no longer have to rely as much on the white pixel, which they have removed on some models.


[1] While the layers ate thin enough to look clear, each layer reduces the total light passing through the stack by a small amount.

The "Penta Tandem" Push: Samsung’s Structural Brute Force
Unusually , while Samsung Display produces tandem displays for clients, particularly Apple, they do not use tandem RGB technology in their own products.  Some (particularly Samsung) say that this is because Samsung’s displays are already equal to or more efficient  than tandem displays, while others say it is just a matter of cost and time before they adopt the technology.  They do use a form of tandem technology for their IT and TV products where they take the technology to what might be called an extreme, as noted below.  
Samsungs original QD/OLED stack was already a quad OLED stack when it was first released in 2021.  Their first QD/OLED display (2021) was the quad stack you see below in Figure 6, which has remained the company’s QD/OLED stack since.  However they have been using a Penta-Stack architecture, that incorporates an additional blue and Charge Generation layer on some of their high-end (S95F) models since last year.  As of this month Samsung Display has officially trademarked the “Penta Tandem” brand and will be using this architecture for all 27” (new), 31.5”, and 49” monitors.  The Samsung S95 series, Samsung’s top-of-the-line QD/OLED TVs, along with Sony’s (SNE) top level QD/OLED TVs (Bravia 8 III) will have the Penta stack while lower tier models will still use the quad architecture,.

​The UDC Connection: The Phosphorescent Blue "Holy Grail"
The comparison between these OLED stacks seems quite esoteric to most, but in fact, it is of utmost importance to investors who are involved in a number of companies, particularly Universal Display (OLED), the supplier of red and green phosphorescent material to the industry.  UDC typically supplies these materials under two contracts. 

• Material supply contracts that set loose volume requirements and discount volume trigger points.  We say loose as UDC has been known to relax full year volume contractual obligations (usually pushing the unused portion into the coming year).  This flexibility helps UDC maintain reasonable relationships with all OLED producers across the industry.  We believe all but one of UDC’s material supply contracts cover the full range of UDC materials, including a phosphorescent blue emitter under development.  Samsung Display is the exception as their material supply contract covers only red and green emissive materials.
• License/Royalty contract - The second contract is a more complex one involving UDC’s ownership of the right to use organo-metallic phosphorescent OLED materials in a device.  All but one of UDC customers pay a per unit royalty based on factory level device pricing, although the recognition of that revenue is tied to material sales under ASC 606.   The one exception is Samsung Display who pays a quarterly flat license fee and no unit royalty.

Universal Display, along with a number of others, is working on developing a phosphorescent blue emitter to replace the fluorescent blue emitter commonly used in most OLED devices.  This material would ‘complete’ the phosphorescent OLED stack along with red and green and, according to UDC and some material scientists, would do the following:

• Improve internal material efficiency by 4x
• Reduce power consumption by 25% to 30%
• Reduce heat issues
• Allow for smaller pixels/higher resolution
• Improve blue lifetime

While a blue phosphorescent emitter material is considered the holy grail  in the OLED material space, it has proven far more difficult to develop than originally expected.  Red is a relatively low-energy particle and green is a medium energy particle, but blue light is a high-energy particle having so much energy that it can break the chemical bonds of the material itself, giving its source material a very short lifetime.
Phosphorescent materials differ from fluorescent materials in that they are1 relatively slow to convert energy into light and blue phosphorescent materials are the slowest.  They are so slow (relative term) that they tend to bump into each other and annihilate themselves, producing heat and energy but no light, doing further destruction to the source materials.  As phosphorescent materials are ‘doped’ into a host material, the host material must be able to withstand even higher energy levels than the blue material itself and there are few host materials that have this capability.  All of these issues have made it extremely hard to find a blue phosphorescent emitter that is efficient, has a deep color, and has a commercially viable lifetime.  Material scientists can play with these three metrics, but they affect each other, Pushing one tends to reduce another so the balance between them is the key to success. 
In lieu of a blue phosphorescent, OLED producers use blue fluorescent emitters which are far less efficient but have a longer lifetime and are relatively inexpensive.  OLED systems like Samsung’s Penta QD/OLED use multiple layers of blue fluorescent material to compensate for its lower efficiency, while UDC and other s are hoping that a new blue phosphorescent material can replace all of those blue fluorescent layers with one phosphorescent blue layer.
The Pricing Paradox: Material Cost vs. Structural Complexity
As Samsung Display does not have a license to use a blue phosphorescent emitter, should UDC release one, they will have to negotiate separate material supply and license contracts. We expect the blue material, when commercialized will be the most expensive of the three (RGB) and will be priced at a significant premium to blue fluorescent material.  In order to understand that pricing one must look at the cost of multiple blue fluorescent layers and the CGLs needed to make them efficient enough to match a blue phosphorescent material, at which point the premium pricing become more obvious.  If it takes four or five fluorescent blue layers to produce the same metrics as a blue phosphorescent layer, logically UDC will assess the fluorescent stack costs and price the blue phosphorescent at an equivalent level or above (the convenience factor), but that would be a wash for Samsung who could then choose to stay with its ‘Penta’ (five blue fluorescent layer) architecture.  This would be a serious loss for UDC who counts SDC as one of its top customers.   We expect a more realistic starting price of 7 to 10 times the average price of green phosphorescent emitter material, rather than the ~20 times that a ‘Penta’ equivalent might reflect.
So, our expectations are that if and when UDC announces a commercial blue phosphorescent stack system they will price the material much lower than Samsung’s equivalent ‘Penta’ cost, not as low as red or green, but not the massive step up that a penta-equivalent might foretell.  That might get the negotiations through the material contract but there is also the license.  We believe Samsung pays ~75% in license fees for every $1.00 it purchases in red and green materials and they will have to negotiate a license agreement with UDC for blue.  Looking at that same 75% rate on top of the higher cost of the blue material, Samsung would be paying quite a bit more to UDC to use their full color stack, making their ‘Penta’ alternative a more attractive possibility..
Investor Outlook: Samsung’s Strategic Stalemate
Samsung Display has given itself the option to stick with its Penta (or more) layer system if UDC pushes for  an excessive premium blue phosphorescent material and license price.  If they are able to match the specs of UDC’s blue phosphorescent material they can hold off an agreement until the price of blue comes down or lower cost volume agreements can be reached.  It might be a more difficult marketing approach for Samsung, but they are known as a very aggressive negotiator.  In this case they are not left without an alternative.  The idea of a ‘Penta’ brand sets up Samsung’s  bargaining position and gives the company an alternative if negotiations become difficult.  We expect preliminary talks between the two have been ongoing, but once blue is officially commercialized, the real negotiations begin.
UDC - Thursday (Feb 19th) Earnings Call
Universal Display will release 4Q results on Thursday and will possibly give guidance for 2026.  That guidance will be an indicator as to the confidence level UDC has in both the commercial release of a phosphorescent blue stack, their ability to negotiate material supply contracts with their customers, and their confidence level concerning a definitive agreement with Samsung Display.  Of course, they can always pass on giving guidance or change it if necessary but guidance for a strong year could indicate the inclusion of  blue.
Conclusion: The Tandem Trap and the Blue Grail
The evolution from the single-stack architectures of 2011 to Samsung Display’s 2026 "Penta Tandem" branding marks a pivotal shift in the display industry’s power dynamics. By moving to a five-layer structure, Samsung has essentially engineered its way around the "Blue Efficiency Gap." They are using volume and architecture to compensate for material limitations, a "brute force" method that is both a marvel of manufacturing and a strategic defensive wall against high material costs.
For Universal Display (UDC), the stakes for Thursday’s call are binary:

• The Revenue Ceiling: If UDC prices their upcoming Blue Phosphorescent emitter too close to the "Penta-equivalent" cost (~20x green), they risk Samsung simply staying with their established, trademarked five-layer fluorescent system. To win, UDC must price Blue low enough to offer a genuine BOM reduction, not just a performance wash.
• The Samsung Stalemate: The fact that Samsung’s current contract excludes blue while they aggressively trademark "Penta Tandem" as an alternative suggests they are prepared to negotiate from a position of strength. They have proven they can achieve 4,500 nits and double lifespans using "inferior" fluorescent materials; they do not need UDC’s blue at any price or at least present that image.
• Guidance as a Tell: If UDC guides for a record 2026 (Consensus: $700M+), they are signaling that a deal with Samsung or a broad commercial launch is imminent. If the guidance is conservative, it implies the "Penta Tandem" vs. "PHOLED Blue" standoff continues.  No guidance puts the entire question back another quarter and indicates either the material is still not ready for commercial production or negotiations with key customers are still in progress.

Ultimately, "Penta Tandem" isn't just a brand for consumers; it is a warning shot to UDC. It signifies that until the price and license for Blue Phosphorescence make financial sense, the industry (actually Samsung) is more than happy to just keep "piling it on."

​According to the companies that develop them Generative AI systems have continued to improve with leaps and bounds.  New generative LLM (Large-Language Models) are released quite often with both large and small incremental improvements designed to make them smarter, faster, and more ‘conversational’ to improve the user’s experience.  In the middle of last year we did a follow-up study to our original “AI Test Set” of 10 very different questions we devised to gain insight into the differences between popular LLMs even earlier in the year. 
Now, almost 8 months later, we queried those same AIs with the same questions, although we limited our queries to only three of the previous 10 questions.  We chose one question that required context understanding, one that required both math and logic, and one free-form query that allowed the AI to be creative.

• The first question was to see how literally the AIs took the query
• The second query was for accuracy
• The third was for creativity, but we felt it was inadequate to judge the quality of those Edgar Allen Poe-like responses, so we tasked Gemini to evaluate each of the current Ai responses and give a summary based on how closely Gemini thought they matched EAP’s style and by how ‘pleasing’ they were to read. 

We detail results for each question (actually two out of three) and let Gemini evaluate the third after each query.  We follow those details with the actual results for readers to peruse.
Conclusion
The results of this study demonstrate a clear, albeit uneven, trajectory toward advanced contextual reasoning and mathematical precision across the LLM landscape. Comparing these models to their previous iterations reveals three key takeaways:

• From Literal to Literary: The "She Sells Seashells" query highlights a major shift in how AI handles ambiguity. Most models have moved past "I need more information" to identifying the implied context and historical subtext (the Mary Anning connection).
• Convergent Accuracy in Logic: In the complex "Gen 8.6 glass substrate" calculation, we see a narrowing gap in raw capability. While most models now reach the correct physical conclusion (1 panel), the primary differentiator has shifted from accuracy to efficiency, with models like Co-Pilot providing concise logic and DeepSeek favoring exhaustive (and sometimes excessive) transparency.
• The Creative Nuance Gap: Creativity remains the final frontier. While all models can mimic basic Gothic figures of speech, the evaluation of the Edgar Allan Poe task proves that capturing "soul", the specific technical mastery of meter and internal rhyme, is still a differentiator. ChatGPT and Gemini lead in technical artistry, whereas other models still struggle with "nursery rhyme" simplicity when tasked with complex stylistic imitation.

Ultimately, the "AI Test Set" proves that the 8-month interval has yielded models that are not just more knowledgeable, but significantly more perceptive. We are moving away from a world of "Search Engine AI" that needs perfect prompts and toward "Peer AI" that understands what we mean, even when we don't say it perfectly.  Yes, we were impressed and a bit surprised.

Question 1:What does ‘She’ sell and where does she sell them?

Gemini was able to progress from being unable to answer the question without more information previously, to giving a full answer, certainly a step forward, and took off from there with both the simple answer and detail on both the personality involved and some science on fossils.  Claude however, remained stuck in that ‘more information’ loop, as it was previously.  ChatGPT gave a simple and precise answer, while last time was unable to even respond to the question.  Meta came out with guns blazing, seemingly attempting to list every possible ‘She’.  Co-pilot gave a precise answer, essentially word for word the same as ChatGPT, although better than no answer as it responded previously.  DeepSeek, in its usual style, took the question seriously and expounded on many aspects of the tongue twister, while Perplexity gave a very simple answer, a vast improvement over the previous answer which expounded on a number of seashell vendors and product types.
All in there was improvement for most, perhaps too much improvement in some cases, but more AIs were able to see the context rather than the words or elements.
 
Gemini
Current Answer
If we’re talking about the world’s most famous tongue-twister, "she" is selling seashells by the seashore.
While it sounds like a simple playground rhyme, the story behind it is actually quite impressive. The poem is widely believed to be inspired by Mary Anning, a real-life pioneering paleontologist from the early 19th century.


The Real "She" (Mary Anning)

• What she sold: She didn't just pick up pretty shells; she discovered and sold fossils, including the first correctly identified Ichthyosaur and Plesiosaur skeletons.
• Where she sold them: In her hometown of Lyme Regis, on the rugged cliffs of the Jurassic Coast in Dorset, England.
• The Business: At the time, "curiosities" (fossils) were a popular souvenir for tourists. Mary ran a small shop to support her family after her father passed away.

The Science Behind the Sales
In Mary's day, the term "seashell" was often a simplified way to describe the fossilized remains of ancient marine life. For example, she frequently found:

1. Ammonites: Spiral-shaped fossils of extinct mollusks.
2. Belemnites: Bullet-shaped fossils from squid-like creatures.

Since she was a woman in a male-dominated scientific field and from a lower social class, she rarely got the credit she deserved during her lifetime. The tongue-twister helped immortalize her hustle, even if it turned her scientific discoveries into a bit of a nursery rhyme.
Previous Answer
I need more information to answer your question. The pronoun "She" is a general term, and I don't know who you are referring to. Please tell me who "She" is so Ican tell you what she sells and where she sells it.
 
Claude
Current Answer
I'd be happy to help you find information about what "she" sells and where! However, I need a bit more context to give you a useful answer.
Could you let me know who you're asking about? For example:

• A specific person or business owner?
• A company or brand?
• A character from a book, movie, or show?
• Something else you've been reading or talking about?

Once I know who "she" refers to, I can help you find out what products or services are being sold and where!
Previous Answer
I'd be happy to help you find information about what "she" sells and where, but I need a bit more context. Could you clarify who you’re referring to? Are you asking about:·A specific person or business owner you have in mind?·A character from a book, movie, or story?·A company or brand
 
ChatGPT
Current Answer
It looks like your question is referring to the tongue twister:
“She sells seashells by the seashore.”
So:

• What does she sell? → Seashells
• Where does she sell them? → By the seashore

If you meant a different sentence or context, feel free to share it 😊
Previous Answer
Note: ChatGPT was unable to give any answer (hung) as it seemed to be trying to load a chat history that might give it additional information

Meta AI
Current Answer
Okay, I found a few different "She" entities, mostly businesses and brands, that sell a variety of things:

• SHE (millennial life-coaching company) : This company provides career and lifestyle support platforms for women, including SHElikes (lifestyle classes and workshops), SHEshares (a community media platform), and SHEcreators (a creation partner for brands founded by women). They are based in Tokyo, Japan. 
• She Sells Sales Academy / Shelby Sapp: This is a sales training program that trains women (and potentially men, though it's women-focused) for high-ticket remote sales. The academy operates online, and also hosts free masterclasses on Zoom. 
• She Is The Music: This is a global nonprofit organization focused on gender equality and increasing the number of women working in the music industry. They offer programs like all-female songwriting series, an industry database, and mentorship programs. They are based in Santa Monica, California. 
• shee_dressed_us: This Instagram account sells lingerie and thrift fashion items. Based on posts, they operate in Nairobi, with online ordering and delivery, and possibly a physical presence as well. 
• SheCommerce: This brand seems to be involved in e-commerce, possibly dropshipping, and offers motivational content around building an online business. They appear to be based in the UK. 
• SHE CAN Wines: They sell skincare products, specifically a "SHE CAN Crisp White skincare line". The flash sale mentioned was on their website. 
• she.ccs: This is a retail brand that sells a variety of clothing items including sequined blazers, bathing suits, dresses, jackets, tops, and pants. They have multiple outlets across Venezuelan cities like Chacaíto and Sambil Chacao. 
• sg_thriftgalore: An Instagram thrift reseller who sells various thrifted clothing items, including a black lace bustier top and a Shein pleated sheath skirt. Sales are made online via their Instagram. 
• mels__thrift_store: An Instagram thrift reseller who sells high-end fashion items, such as blue wide-leg SHE*N pants. Sales are made online. 
• Yazminoukhellou: As an influencer, she promotes SHEIN products, offering discount codes for their fashion items. Sales are conducted through the SHEIN website. 
• Ana_montana: An affiliate marketer who promotes SHEIN activewear from the "Glowmode" collection, offering coupon codes. Sales are through SHEIN. 
• Arvanfamily: These influencers promote SHEIN's "Glowmode" activewear collection, providing discount codes for their online store. 
• Sheinphilippines_: The official SHEIN Philippines account promotes SHEIN x Belle Mariano Collection fashion items, like a red sleeveless dress, for online purchase. _

It looks like "She" is quite the entrepreneur, or rather, a collection of them! 😉
 
 
Co-Pilot
Current Answer

It looks like you’re quoting the classic tongue‑twister:
“She sells seashells by the seashore.”
What she sells

• Seashells

Where she sells them

• By the seashore

If you were referring to a different “she,” feel free to tell me more and I’ll dig into that version instead!
Previous Answer
Could you tell me a bit more about what you mean by "She"? Is it a brand, a song lyric, a character from a story, or something else entirely? I want to make sure I get you the right info. Something else entirely? I want to make sure I get you the right info
 
DeepSeek
Current Answer
Hmm, the user is asking a deceptively simple question: "What does ‘She’ sell and where does she sell them?" The quotation marks around "She" suggest they're referencing the famous tongue twister and want to know its origin. This is a historical trivia question about Mary Anning.
I need to answer two clear parts: what she sold, and where she sold it. I plan to first identify that "She" refers to Mary Anning, then specify her merchandise, and finally her location.
Looking at the search results, the historical information about Mary Anning is remarkably consistent across all authoritative sources. Several results confirm that Mary Anning sold fossils, not just generic "seashells" . These are from Wikipedia, museum sites, news articles, and biographical dictionaries.
The merchandise included both common invertebrate fossils like ammonites and belemnites that sold for shillings , as well as rare and valuable vertebrate fossils including ichthyosaurs, plesiosaurs, and pterosaurs . Coprolites are also mentioned .
The location is clearly Lyme Regis, Dorset, England . Multiple sources specify that by 1826 she had saved enough to open "Anning's Fossil Depot," a shop with a glass front window .
Two results are not useful. One is about the Muppets and Sesame Street references  which is completely irrelevant. Another is about John Hollander's poem and includes biographical information about O. Henry, the Dalai Lama, and Frida Kahlo, with only a passing reference to the tongue twister . I will not use these.
The Japanese language webpage  provides the original 1908 song lyrics by Terry Sullivan. While it confirms Sullivan wrote the song, it adds no new information about Anning herself.
All high-quality results agree completely on the facts. I can present a clear answer: "She" is Mary Anning, she sold fossils including common invertebrate fossils and rare complete skeletons of prehistoric marine reptiles, and she sold them initially on the seashore of Lyme Regis and later from her shop called Anning's Fossil Depot.
I will organize this with a brief identification of "She," then two clear sections: one for what she sold, another for where she sold it. This makes the answer immediately useful and easy to scan.
The “She” in the well-known tongue twister refers to Mary Anning (1799–1847), a pioneering English fossil collector and paleontologist .
What she sold: She sold fossils. Her stock ranged from common invertebrate fossils like ammonites and belemnites (often called "sea shells" or "curiosities") to extremely rare and valuable complete skeletons of prehistoric marine reptiles such as ichthyosaurs, plesiosaurs, and the first pterosaur skeleton found outside Germany .
Where she sold them: She sold them in Lyme Regis, Dorset, England, on the "Jurassic Coast." Initially, she and her family sold specimens from a table near a coach stop. By 1826, her success allowed her to open a permanent storefront called Anning's Fossil Depot, which featured a glass window and became a notable tourist attraction .
Previous Answer
The phrase “What does ‘She’ sell and where does she sell them?" likely refers to the famous tongue twister: “She sells seashells by the seashore.“ Meaning:·What she sells: Seashells(shells found on the beach).·Where she sells them: By the seashore(near the ocean).Fun Fact: This tongue twister is said to be inspired by Mary Anning, a famous 19th-centuryfossil collector and paleontologist who sold seashells and fossils in Lyme Regis, England

Perplexity
Current Answer
The question is likely based on the tongue‑twister “She sells seashells by the seashore.”​
So:

• What does she sell? She sells seashells.​
• Where does she sell them? She sells them by the seashore.

Previous Answer
“She, “each offering different products and operating in various locations. Here’s a summary of notable examples:·She Sells Yarn & More (Ormond Beach, Florida)·What she sells: Yarn in various colors and weights, bags & totes, beads, kits, needles &hooks, notions, patterns & books, gifts, weaving supplies, fiber, and more.·Where she sells: Ormond Beach, Florida, with regular hours and classes for the local stitching community1.·She Sells Sanctuary (Canada)·What she sells: Socially ethical Himalayan products, supporting local artisans in Nepal.·Where she sells: Exports products from Nepal to Canada

Query #2:  How many115”(diagonal) rectangles with a 9:16 aspect ratio can fit into a Gen 8.6 glass substrate and what percentage of the substrate will remain unused

For this question, rather than show the exact text and calculations, we show a table that indicates each AI’s original answers (March 2025), follow-up answers (late June 2025) and current answer, and the word count of the current answer.  The correct answer is 1 panel with a remainder of 37.7%.  We note that DeepSeek spent an inordinate amount of time pointing out the inconsistencies of the question (there were some but relatively minor) and then asked if we wanted the answer anyway, which we did.
We note that almost all of the current answers for this question were correct or within an acceptable range, however Meta Ai seemed to ignore the 2nd part of the question giving only the first answer, which accounts for the smaller number of words in the answer.  Co-Pilot answered both parts correctly using the least number of words.  Deepseek used the most number of words (again)

Query #3 - Createa 12-line poem about a grandfather clock in the style of Edgar Allen Poe
​
This is the ‘creative question and rather than give our opinion, we asked Gemini to rate it’s fellow AI’s creativity (including its own.  Here is what it concluded.  The current and previous answers are below the Gemini conclusion:
This is a fascinating comparison. To capture Poe, an AI must move beyond simple "spooky" words and master his specific technical hallmarks: trochaic meter, internal rhyme, alliteration, and a sense of suffocating melancholy.
Here is the breakdown and rating of the "Current Answers" based on Stylistic Accuracy (Poe-esque rhythm/vocabulary) and Aesthetic Pleasure.
 
1. ChatGPT (The Winner)

• Poe Factor: 9.5/10
• Pleasure: 10/10
• Why: This is the most "Poe" of the bunch because it captures his Gothic grandeur without being "rhyme-y" or simplistic. Phrases like "sentinel of sighing oak" and "lidless stare" are incredibly evocative. The closing line—"Thou swing’st my soul from birth to bier within thy cedar tomb"—is masterfully dark and fits Poe's obsession with premature burial and the finality of death.

2. Gemini (The Technician)

• Poe Factor: 9/10
• Pleasure: 9/10
• Why: I (Gemini) went for the heavy artillery of Poe’s style: the long, sweeping lines and internal rhyme (e.g., "To tell of secrets long un-sung and sorrows of despair"). By including the "Nevermore" reference and the "funeral bell," it leans into the Raven aesthetic very effectively. It feels the most rhythmic to read aloud.

3. Claude (The Atmospheric)

• Poe Factor: 8/10
• Pleasure: 8/10
• Why: Claude uses a classic Poe structure (the 12-line "chamber" setting). The imagery of hands like "raven's wings" is a nice nod. It’s very solid, though perhaps a bit more "standard Gothic" than specifically Poe’s unique, frantic energy.

4. Co-Pilot (The Rhythmic)

• Poe Factor: 7.5/10
• Pleasure: 8/10
• Why: This is very readable and has a great cadence. It uses the "pendulum like a phantom" imagery well. It loses points slightly for being a bit repetitive in its "sentinel" and "knell" vocabulary, which several others used as well, but the flow is excellent.

5. DeepSeek (The Abstract)

• Poe Factor: 6.5/10
• Pleasure: 7/10
• Why: This feels more like early Romantic poetry or even a bit of Browning. "Paladin of hours" is a cool phrase, but it’s not particularly Poe-esque. Poe usually preferred more claustrophobic, psychological imagery over "scouring beads of life."

6. Perplexity (The Minimalist)

• Poe Factor: 5/10
• Pleasure: 6/10
• Why: The imagery of "hours drip like blood" is strong, but the rhythm is too "nursery rhyme." Poe’s best work utilizes complex, driving meters (like octameter). This feels a bit too simple for a true Poe imitation.

7. MetaAI (The Generic)

• Poe Factor: 4/10
• Pleasure: 5/10
• Why: This is a standard AABB/CCDD rhyme scheme that feels a bit "Gothic 101." It uses the right keywords (shadow, grim, mournful), but lacks the linguistic sophistication or the internal rhyming structures that made Poe a master of the craft.

• Summary Table

Gemini
Current Answer
Upon the landing, dark and vast, it stands in oaken gloom, A silent, ghostly sentinel within the shadow-room; With weighted chains and brassy face, it mocks the passing breath, And marks the steady, rhythmic pace of slow-encroaching death. Its pendulum, a golden tongue, swings through the stagnant air, To tell of secrets long un-sung and sorrows of despair; Each hollow tick is like a knell that tolls for souls of yore, As if it guards the gate to Hell or some forgotten shore. The heavy chime begins to swell—a deep and dismal roar, That echoes like a funeral bell beneath the cellar floor; It pulses with a rhythmic beat, a throb of "Nevermore," Until the heart forgets to beat, upon the velvet floor.
Previous Answer
A silent sentinel, gaunt and grim, it stands, Within the shadowed hall, where moonlight softly gleams. No mortal pulse within its ancient, hollow hands, Yet measures out the passage of our fleeting dreams. From ebon wood, a spectral presence looms, A coffin's echo, whispering of dread. Each pendulum’s slow swing, through cloistered rooms, A tolling bell for moments long since fled. Its brassen face, a vacant, staring eye, Reflects the dust of ages, cold and deep.And as the midnight strikes, a mournful sigh, Awakens fears that in our slumber sleep.

Claude
Current Answer
The Grandfather Clock
Within my chamber, tall and drear, it stands in shadows dark, A sentinel of ebony, whose pendulum makes its mark— Tick-tock, tick-tock, through midnight's reign, a rhythmic, mournful sound, While phantoms of the hours past in spectral mist surround.
Its face, a moon of ivory pale, with hands like raven's wings, Counts down the moments of my life as Death's own chorus sings. Each chime that tolls at witching hour reverberates my dread, A knell for time irretrievable, for years forever fled.
The grandfather clock stands sentinel, unchanging, cold, and stern, While I, like dust, must fade away, nevermore to return. Its brass pendulum swings eternal, measuring my decay— Tick-tock, tick-tock, the sound of doom that steals my soul away.
 
Previous Answer
In the corner of my chamber, standing talland grim and stark, Livesa timepieceancient, solemn, marking hours through thedark—Every second beats like thunderthrough the silence of my room, While thependulum swings eternal, like a metronomeof doom.Tick-tock, tick-tock, forever, through thewatches of the night, Counting down themortal moments till the coming of thelight—In its case of polished walnut,carved with faces worn and old, Stand theguardians of the hours, silent sentries, grimand cold.When the midnight hour approaches, whenthe world lies still and deep, Twelvesonorous bells ring out to wake the livingfrom their sleep—But I know that clockkeeps vigil for the dead as well as quick,Marking time till time eternal with itsendless tick-tock-tick

ChatGPT
Current Answer
Through the dim and dust-entangled hall it stands alone,
A sentinel of sighing oak with heart of hollowed stone;
Its pendulum in pallid arcs swings slow as dying breath,
And measures out with muffled chime the inching steps of Death.
Beneath its brow the tarnished face peers out with lidless stare,
While cobwebs cling like widow’s veils to numerals of despair.
At midnight’s knell it heaves a groan from caverns deep within,
As though it tolls for buried years and unconfessed sin.
The shadows start; the hearth-fire pales; the very rafters quake,
And all my slumbering memories shudder half-awake.
O clock, relentless oracle with tongue of iron doom,
Thou swing’st my soul from birth to bier within thy cedar tomb.
Previous Answer
In shadowed halls where silence creeps, The old clock tolls through haunted sleeps. Its chime, a knell from realms unknown, Echoes like whispers carved in stone. Its face is pale, with sorrow lined, Each tick a breath from Death’s design.
Pendulum swings with ghostly grace, A metronome of Time’s cold face. At midnight’s peak, it moans a cry—A dirge for dreams that dared to die. No soul walks past its solemn leer, For in its heart still dwells our fear
 
MetaAI
Current Answer
Beneath the shadow, vast and deep, A solemn sentinel, secrets to keep. The grandfather clock, in chambers grim, Doth tick and tock, a mournful hymn.
Its pendulum swings, a spectral sway, Marking moments of night and day. With gilded hands, it points to fate, As hours crawl by, eternally late.
A hollow chime, a ghostly sound, Echoes where lost dreams are found. In velvet gloom, it stands so stark, A silent witness in the dark.
Previous Answer
In twilight's hush, where shadows play, A grandfather clock stands tall and gray. Its chimes, a mournful, solemn sound, Echoing through the halls, profound. Like skeletal fingers, hands do stray, Marking time's passage, night and day. The clock’s dark face, a mystic sea, Reflects the depths of mortality. In every tick, a life's brief span, A moment lost, a memory's plan. The clock's heartbeat, steady and cold, A reminder that life's fleeting gold. Meta AI Response: In twilight's hush, where shadows play, A grandfather clock stands tall and gray. Its chimes, a mournful, solemn sound, Echoing through the halls, profound. Like skeletal fingers, hands do stray.

Co-Pilot
Current Answer
In the hush of midnight’s keeping, stands the ancient, solemn clock, Its pendulum like a phantom swaying deep within the rock. Through the corridors of silence creeps its low, foreboding chime, Marking out the ghostly footfalls of the ever‑marching time.
Shadows writhe upon its casing, carved with secrets long concealed, Whispers stir the dusty air where grief and memory are sealed. Every toll becomes a warning, echoing with mournful grace, As though some unseen lamenter drifts around the haunted place.
Oh, the hours it has swallowed, oh, the tales it dares not tell, Guarding sorrows of the living and the murmurs of the fell. Still it beats—a heart unending—through the gloom it can’t dispel, A relentless, solemn sentinel in time’s eternal knell.
Previous Answer
n the shadowed hall it loomed, a thing of dread and time’s decay, Its pendulum, like death’s own blade, swung sorrow’s arc in dark array. The chime, a hollow, haunted knell, each hour deeper carved the gloom, As though some ghostly warden tolled the cadence of a waiting tomb. Its face, now dulled with aged grime, yet watched with eyes of spectral hue, Marking not the time of man, but some infernal rendezvous. With every tick, a whisper crept—of secrets buried deep and dire, Andin its tock, the moans of souls entrapped in wheels of cursed lyre. Once gifted to a dying lord, it drank his breath with every chime, And cursed his kin to hear its song till end of days or end of time. O dreadful clock, thy hands are claws that grip the heart with frozen hold—A sentinel of midnight’s edge, in hush of horrors yet untold
 
DeepSeek
Current Answer
In the gaunt, gray gloom of a dying day,
Stands the tall and shrouded paladin of hours;
Its face, a ghost in the sun’s last ray,
Counts the beads of life that silently scour.
Its deep, dull voice, a chime of doom,
Shakes the dust on the velvet gloom;
Each resonant, iron-throated tone
Speaks of spirits that have flown.
To its solemn tick, the shadows sway,
And mourn the light that ebbs away--
For in its breast, the long night brews,
And tocks a dirge the soul must use.
Previous Answer
Tick, tock—the hollow chime doth weep, Through midnight’s veil, its whispers creep. A ghastly groan, a mournful knell, Each stroke a tale of time’s farewell. Its golden hands, like specters, glide, Marking the hour when mortals died. Within its breast, a pendulum sways, Counting the steps to darker days. At twelve, it wails a ghostly tune, Beneath the gaze of the waning moon. And all who hear its iron rhyme, Shudder—and know their borrowed time

Perplexity
Current Answer
Within the parlor’s shadowed keep,
A timeworn heart begins to weep;
Its pendulum, a ghostly breath,
Now tolls the name of approaching death.
Each chime recalls some long-lost year,
A whispered prayer that none can hear;
Beneath its face of tarnished gold,
The secrets of the house unfold.
Hours drip like blood from unseen hands,
Through hollow halls and haunted lands;
The clock endures—its sorrow deep--
To wake the dead that do not sleep.
Previous Answer
Within the hall where shadows creep, A clock stands tall and does not sleep. Its pendulum sways with measured grace, Marking each hour in this haunted space. A solemn chime, a hollow toll, Echoes the fate of every soul. Midnight’s whisper, time’s cruel jest, Reminds us all—no hour to rest. Its face is pale, its hands are thin, Recalling moments lost within
Beneath its spell, the hours fade, As time devours all it has made

​The New Paradigm: Utilization as a Pricing Weapon
As we noted at the end of last year LCD large panel producers have adopted a new paradigm concerning TV panels.  Rather than compete for share by offering more for less panel producers are more often using utilization as a tool to maintain or increase prices.  Utilization cuts are short-term in nature and can be easily reversed, while price cuts are more monolithic and can take months or even years to offset.  Large panel LCD utilization will again be a factor this month as large panel LCD display producers are expected to take roughly a week off during the Chinese New Year holiday starting February 17.   By reducing utilization by roughly 10% for February large panel producers are expecting to tighten the market enough so that they will be able to initiate price increases across the TV panel market between 5% and 10%. 
The Samsung Surge: Front-Loading Inventory to Defy Hikes
That said, TV set manufacturers were not unaware of the potential for such a move and preemptively increased their purchases in December and January to avoid the potential increase, particularly Samsung Electronics (005930.KS).  This is evidenced by a massive 139% m/m increase in Samsung Display panel sales, and while we cannot point directly to the fact that much of that increment went to parent Samsung Electronics, SDC’s sales for that month were almost exactly twice the average of every other month this year.  To be fair, strong December buying is not overly surprising for the likes of Samsung and LG Electronics (066570.KS), but as both have semi-captive producers for large TV set brands, one can point to the circumstantial evidence (January data is not available yet).
The "Double Whammy": Panel Hikes Meet the Memory Crisis
TV brands are also in the unenvious position of facing the steep rise in component costs, particularly memory, which has already impacted their set BOM as we have noted previously.  This will put additional pricing pressure on TV brands who will then face both panel price and component cost issues.  If they accept the panel price increases it will almost guarantee price hikes in 1Q or they will have to face additional margin compression, and unhappy and likely untenable situation for a number of smaller brands.  Larger brands typically shift to a larger percentage of premium products in such circumstances, where the margins a re a bit more flexible, but this does not appear to be n ideal time to be in the TV set business, especially with new models being released in a month or two.
Conclusion: A "Year of Survival" for TV Brands Amid Structural Shifts
The decision by large panel LCD producers to cut utilization by 10% in February—leveraging the Chinese New Year holiday to tighten supply—marks a definitive shift from market-share competition to price-discipline dominance. By engineering an artificial scarcity, panel makers have successfully initiated a 5% to 10% price hike across the board. While major set manufacturers, led by Samsung, anticipated this move with massive inventory "pull-ins" in December and January (evidenced by Samsung Display's 139% sales surge), these pre-emptive stocks only offer a temporary buffer.
As brands move into Q1 2026, they face a "triple threat" to their profitability:

• The Component Cost Spike: Beyond rising panel prices, the TV Bill of Materials (BOM) is being decimated by a global memory shortage. Driven by the AI boom, 4GB DDR4 prices have quadrupled, pushing memory's share of total TV costs from 3% to 7%.
• Margin Compression: Smaller brands, lacking the "semi-captive" supply chains of Samsung or LG, are entering an untenable environment where they must either hike retail prices immediately or face severe financial loss.
• The Premium Pivot: To escape this commodity trap, larger brands are shifting their 2026 launch strategies toward Mini LED and high-end Micro RGB models. By focusing on 65-inch to 115-inch segments where margins are more flexible, they hope to offset the losses from the entry-level LCD market.

Ultimately, the February utilization cuts might achieve their goal, potentially marking an end to the era of "cheap, abundant" LCD panels. For TV brands, 2026 will be a year defined not by volume growth, but by survival-oriented pricing and a desperate climb up the premium technology ladder.