By Thomas | financial enthusiast
My AI diary: July 20 — The Great AI Divide
I woke up this morning expecting a standard tech rally, but instead, I found myself staring at a headline that feels like it belongs in a dystopian novel. OpenAI just dropped GPT-5.6, and then—almost immediately—the US government stepped in to block public access.
I had to sit with this for a while. One minute we are celebrating the peak of human-machine reasoning, and the next, the doors are being slammed shut. It feels like the era of "free and open AI for everyone" just hit a brick wall. (Damned. I really thought we were past this.)
The Tiered Reality
I spent the morning digging through the technical specs to understand what exactly is being kept from us. It’s not just one model; it’s a whole ecosystem. OpenAI launched it in three distinct tiers: Soul, Terra, and Luna.
According to the reports, Soul is the king of frontier reasoning, while Luna is the budget-friendly workhorse for fast tasks. Then there’s Terra—the one everyone is buzzing about. Developers are calling Terra the "value pick" for everyday use, but even that might be out of reach for the average person now.
It’s wild to think that ChatGPT has hit 1 billion users, yet the very tech driving that scale is being restricted to "vetted partners with approved credentials." I didn't realize how quickly the gatekeeping would become official. It’s a massive strategic move, especially since OpenAI is filing for an IPO with a valuation between $830B and $1T. (Talk about high stakes.)
A Systematic Pattern
At first, I thought this was just an OpenAI outlier, but the more I read, the more it looks like a coordinated regulatory shift. This isn't the first time we've seen this happen. Five weeks ago, Anthropic had a similar situation with their Claude Fable model being restricted.
It seems like a pattern is emerging: as these models get more powerful, the government is stepping in to control access. We are seeing a massive shift in compute governance. The idea is that these models carry systemic risks, and the government wants to ensure only vetted institutions hold the keys.
One analyst put it well: the era of unrestricted frontier AI is ending. We are moving from a playground to a fortress. For investors, this is a massive signal. The value isn't just in the code anymore; it's in the permission to use it.
The Rise of the Counterweights
So, what happens to the rest of us? If you’re a developer who isn't a "vetted partner," you’re essentially locked out of the highest level of intelligence. I was worried for a second that the industry would stall, but the market is already pivoting.
Since we can't touch the proprietary giants, the open-weight models are becoming the most important players in the game. I read that Moonshot AI’s Kimi K2.7 Code is already being integrated into GitHub Copilot, and it’s becoming a critical alternative.
If the frontier models are gated, then the models that are accessible—like Kimi, DeepSeek, or Llama—become the strategic backbone for the entire developer community. It’s a fascinating, if slightly frustrating, shift. The "elite" AI is being reserved for the institutions, while the rest of us have to build our future on open-weight alternatives. (Works out nicely for the open-source crowd, I suppose.)
I'm still processing the sheer scale of this. We are watching the creation of a new digital hierarchy in real-time. One group gets the "Soul" of the machine, and the rest of us get what's left over.
Do you think a gated AI future will lead to more safety, or just more inequality?
