Pontification
Most companies have the same basic goal of generating profits. Depending on the type of company, the profits might be funneled to the company’s owners, investors, or even employees, but the motivation is the same, and when compensation is tied directly to performance, even more so. Some Ai companies are different in that they are either non-profit corporations, Public Benefit Corporations, or privately held (we exclude the internal divisions of public companies like Google’s (GOOG) Deep Mind or Meta’s (FB) internal model development division), and while these Ai companies are perceived as somewhere between high-end research labs and academia, they tend to be held in higher ethical regard than your average for profit corporation.
In fact there is an unending flow of statements, moralizations, and even blowviations from AI company CEOs, thought leaders, and some very wealthy industry representatives who state how important it is for everyone to work together to bring the benefits of Ai to the world.
In fact there is an unending flow of statements, moralizations, and even blowviations from AI company CEOs, thought leaders, and some very wealthy industry representatives who state how important it is for everyone to work together to bring the benefits of Ai to the world.
Here are one or two:
"AI will be an integral part of solving the world’s biggest problems, but it must be developed in a way that reflects human values.” - Satya Nadella, CEO of Microsoft
The playing field is poised to become a lot more competitive, and businesses that don’t deploy AI and data to help them innovate in everything they do will be at a disadvantage.” - Emad Mostaque, Founder and CEO of Stability AI.
But are things really that way, even among those who share this moral high ground? Not really… While the architects, interface designers, and coders might work in a rarefied atmosphere of colaboration at the highest scientific level, the organizations themselves do not. Here’s the most recent example:
In 2022 the one-year old compiler and virtualization company Windsurf (pvt), saw the light and converted itself into an AI coding company, creating a tool called Agent IDE. As background, coding is a complex process that is attractive to persons who are able to follow multiple flow paths at the same time and hundreds of variables in order to create an application. A single incorrect letter or number can completely halt an application comprising thousands of lines of code and days or weeks of work, and digging out the error and correcting it can become the most tedious part of the process. AI coding platforms, like Agent IDE, are sets of tools that help not only to automate debugging, but can write bits of code based on the developers requests, in some cases drastically reducing production time.
Each of these Ai tools is based on a single or set of models that the programmer accesses directly through the platform, the most popular being Gitbub Co-Pilot (open source), Anthropic’s (pvt) Claude 3.x (3.5 Sonnet, 3.7 Sonnet and Claude 4.0), and OpenAI’s (pvt) GPT 4.0. Stable access to these models and high speed processing is essential for the platform’s success. On May 6, OpenAI announced the purchase of Windsurf for $3b and 5 days later Anthropic decided to cut off all direct access to its Claude models from Windsurf clients.
AI model companies issue software ‘keys’ to their customers, in this case Windsurf, and all of Windsurf’s users access those models through the Windsurf key, essentially being responsible for its clients connection to AI models. While Windsurf allows client access according to each client’s Windsurf plan, all access to Anthropic models goes through Windsurf’s key. By turning off that key Anthropic forced Windsurf clients to buy expensive 3rd party access keys directly to Anthropic’s servers and manage them individually (what they called “BYOK” – Bring your own key). Other models were unaffected, but it left Windsurf clients unhappy about the extra cost, inconveniene, and potential capacity issues.
Why would Anthropic do such a thing when “we all need to band together to help bring AI to the masses?” They did it because they believe that the purchase by OpenAI will make Windsurf a more GPT-centric platform and therefore less Anthropicy. The money was not an issue as they still receive the same fees from Windsurf’s BYOK clients, but are making it much more inconvenient for Windsurf clients, which they expect will reflect negatively on both Windsurf and its new owner. To assuage its customers, Windsurf lowered the price of using the Google Gemini 2.5 Pro model by 25%. We note also that when Claude 4.0 (among the most powerful coding model), Windsurf competitors all announced access to this new model, all except Windsurf, who was not granted direct access (to this day),
The Kumbaya moments that have sprouted up around Ai over the last year or two are just that, moments. As valuations for AI companies rise into the hundreds of billions, the attraction of those vast piles of money far outweigh that ‘lets all join hands’ feeling that a new technology can sometimes bring. Just as in other technology segments, as the days wear on, more founders will be replaced with more ‘practical’ folk who the board feels might have a ‘better feel for the current AI environment’, and the competitive nature of business will overshadow the ‘around the campfire’ feel that Ai had for a few short moments. We live in a fast-paced society, one that keeps shortening the sharing and caring period of paterntal/child development and gives technology progressively less time to understand the implications of what its doing with each new discovery. Its no wonder the current generation hides in their rooms, and no surprise that those Ai companies with even the best of intentions lose those good intentions quickly.
"AI will be an integral part of solving the world’s biggest problems, but it must be developed in a way that reflects human values.” - Satya Nadella, CEO of Microsoft
The playing field is poised to become a lot more competitive, and businesses that don’t deploy AI and data to help them innovate in everything they do will be at a disadvantage.” - Emad Mostaque, Founder and CEO of Stability AI.
But are things really that way, even among those who share this moral high ground? Not really… While the architects, interface designers, and coders might work in a rarefied atmosphere of colaboration at the highest scientific level, the organizations themselves do not. Here’s the most recent example:
In 2022 the one-year old compiler and virtualization company Windsurf (pvt), saw the light and converted itself into an AI coding company, creating a tool called Agent IDE. As background, coding is a complex process that is attractive to persons who are able to follow multiple flow paths at the same time and hundreds of variables in order to create an application. A single incorrect letter or number can completely halt an application comprising thousands of lines of code and days or weeks of work, and digging out the error and correcting it can become the most tedious part of the process. AI coding platforms, like Agent IDE, are sets of tools that help not only to automate debugging, but can write bits of code based on the developers requests, in some cases drastically reducing production time.
Each of these Ai tools is based on a single or set of models that the programmer accesses directly through the platform, the most popular being Gitbub Co-Pilot (open source), Anthropic’s (pvt) Claude 3.x (3.5 Sonnet, 3.7 Sonnet and Claude 4.0), and OpenAI’s (pvt) GPT 4.0. Stable access to these models and high speed processing is essential for the platform’s success. On May 6, OpenAI announced the purchase of Windsurf for $3b and 5 days later Anthropic decided to cut off all direct access to its Claude models from Windsurf clients.
AI model companies issue software ‘keys’ to their customers, in this case Windsurf, and all of Windsurf’s users access those models through the Windsurf key, essentially being responsible for its clients connection to AI models. While Windsurf allows client access according to each client’s Windsurf plan, all access to Anthropic models goes through Windsurf’s key. By turning off that key Anthropic forced Windsurf clients to buy expensive 3rd party access keys directly to Anthropic’s servers and manage them individually (what they called “BYOK” – Bring your own key). Other models were unaffected, but it left Windsurf clients unhappy about the extra cost, inconveniene, and potential capacity issues.
Why would Anthropic do such a thing when “we all need to band together to help bring AI to the masses?” They did it because they believe that the purchase by OpenAI will make Windsurf a more GPT-centric platform and therefore less Anthropicy. The money was not an issue as they still receive the same fees from Windsurf’s BYOK clients, but are making it much more inconvenient for Windsurf clients, which they expect will reflect negatively on both Windsurf and its new owner. To assuage its customers, Windsurf lowered the price of using the Google Gemini 2.5 Pro model by 25%. We note also that when Claude 4.0 (among the most powerful coding model), Windsurf competitors all announced access to this new model, all except Windsurf, who was not granted direct access (to this day),
The Kumbaya moments that have sprouted up around Ai over the last year or two are just that, moments. As valuations for AI companies rise into the hundreds of billions, the attraction of those vast piles of money far outweigh that ‘lets all join hands’ feeling that a new technology can sometimes bring. Just as in other technology segments, as the days wear on, more founders will be replaced with more ‘practical’ folk who the board feels might have a ‘better feel for the current AI environment’, and the competitive nature of business will overshadow the ‘around the campfire’ feel that Ai had for a few short moments. We live in a fast-paced society, one that keeps shortening the sharing and caring period of paterntal/child development and gives technology progressively less time to understand the implications of what its doing with each new discovery. Its no wonder the current generation hides in their rooms, and no surprise that those Ai companies with even the best of intentions lose those good intentions quickly.
RSS Feed