A week is a long time in the world of AI. On 1 September, when Anthropic launched Fable 5.1, some people on my Instagram feed posted where they were when the advanced LLM dropped, as if it were a moon landing. Two days later, when OpenAI launched GPT-6 Astra, the same people called it a watershed moment, having seemingly forgotten about their own excitement from 48 hours ago.
By 6 September, Nvidia chief Jensen Huang had declared on X, in response to Astra, that AGI (artificial general intelligence) had arrived. The same day, OpenAI's own chief scientist Jakub Pachocki urged “extreme caution and voluntary slowdowns” in AI research and deployment, arguing that humanity is not prepared to handle the rapid rise in machine intelligence.
Real world impacts
All this begs the question: Is the first week of September really the most definitive in AI’s short history? The short and obvious answer is that it’s too soon to tell. But the longer answer depends on who you ask. Shagun Ohri, co-founder and CEO of Thundercat, a GTM AI infrastructure company that operates between San Francisco and Bengaluru, has a filter for new releases like these. “We ask what capability became materially better, and does it improve our job? The faster the model layer moves, the more important it is to become very clear about what you are building, who you are building it for.”
The mistake, she adds, is treating these developments as some kind of product roadmap “Technology alone is not going to continue to be your moat. If your advantage depends on someone else’s model remaining bad at something, that advantage is going to disappear very quickly,” she adds.
Aakrit Vaish, co-founder of Activate, an Indian AI venture capital fund, says that when technology becomes open, considering real-world use becomes more important than a better new model just for the sake of newness. “When it becomes very easy to build and replicate ideas, the depth of one’s understanding matters more than anything else.”
As an example, Vaish says, consider a founder building AI for fraud detection in banks. “We look at whether this person has worked in banks and is deeply aware of how fraud detection works in banks? AI is actually the table-stakes part.”
Disrupting work horizons
Or take the example of someone whose planning horizon can be affected by the release of a new AI model. For Srijan A, a 29-year-old software developer in Bengaluru, those horizons shrunk this week. “Earlier, if I thought something needed five engineers for six months, I could plan around that with reasonable confidence,” he says. “Two model releases later the same thing might be doable by one or two people.”
You can spend months building something internally, and suddenly the models you’re depending on just get good enough that the whole problem collapses, he notes. “Something you spent years building has suddenly become much less defensible. There is definitely a sense of disbelief and even sadness, because the job itself is changing so quickly,” rues Srijan.
So a week’s worth of new AI releases that can look almost routine for a company and an investor, can precipitate a professional rupture for an individual. Those calling AI model releases “historic” have their own specific reasons; the loudest of them sells the chips these models run on. Srijan’s adjustment to the news is more like the quiet reality. He’s focusing on building shorter loops and experimenting more, but with less attachment to any technical decision. “It’s exciting,” he says, “but also strange to watch skills and ways of working that defined your career get reshaped in real time.”