Altman on the founder profile: people who cannot code can be funded too
In an Axios interview given in Chapel Hill and published on 7 September 2026, OpenAI CEO Sam Altman said technical talent had been the central ingredient in founding teams for a long time, that it is still very important, but that he now wants to fund people who understand their users deeply and cannot code at all. With OpenAI's new model GPT-6 Astra, he said, a small business can be started and run effectively, and the economics of building a new business have been reset. The article also reports that more than 3 million new businesses were formed in the United States between January and June 2026.
The Axios interview widens the definition of a fundable founder
Axios published its interview with OpenAI CEO Sam Altman on 7 September 2026. The conversation took place at the Carolina Inn in Chapel Hill, North Carolina, around the G20 Innovation Ministerial. Altman, who is 41, used it to describe a shift in the founder profile: technical talent was the central ingredient in a founding team for a long time, it is still very important, but he now wants to fund people who understand their users deeply and cannot code at all.
Axios framed the piece as 'Altman sees revenge of idea guy'. The article does not describe this as a reversal; what it describes is a change in which skill wins backing at the founding table. In Altman's telling, what moved is not the appetite of founders but the economics of starting a business.
GPT-6 Astra, reset economics and 3 million new businesses
Altman's reference point is OpenAI's new model, GPT-6 Astra, which the article presents as the company's powerful new model. With it, he says, 'you can start and run a small business effectively'. He adds that the economics have been reset for what it means to build a new small or large business, that 'the space of viability has increased dramatically', and that a whole generation is now iterating on ideas faster than it ever could. One distinction is worth keeping: an anecdote about a user producing software refers to GPT-5.6, while the claim about starting and running a business is attached to GPT-6 Astra.
The question has a history inside the piece. Stripe CEO Patrick Collison asked Altman on stage at Stripe Sessions 2026 last spring how the traits of successful founders had changed in this period. The article also reports that more than 3 million new businesses were formed in the United States between January and June 2026, a level no other January to June period matches in Census Bureau records going back to 2005. That figure comes from the article itself, not from Altman. Axios also discloses a licensing agreement between itself and OpenAI.
What changed is the skill the founding table rewards
Three readings follow. One, if the quality capital looks for in a founding team is moving, that definition reaches job posts, partnership talks and team building decisions quickly, because investor language usually shifts before the rest of the economy does. Two, the role assigned to the model goes past producing software: starting and running a business is a wider set of obligations than writing code. Three, nobody is saying technical talent lost its value; the claim is that a second capability placed next to it, understanding users deeply, now counts as fundable.
One caution belongs here. The number of new United States businesses and the capability claim about the model sit side by side in the same article, yet the piece draws no causal link between them. Keeping the two apart is a condition of reading the news correctly, since business formation also moves with tax, employment and cycle effects. The same applies to the model claim itself: the interview offers no measurement showing that a business built this way stays alive. That assessment is ours.
No Türkiye data in the source, so the reading is ours
The source carries no data, regulation or example specific to Türkiye. The interview was given by the head of a United States company, and the business formation figure is United States data. What follows is our reading of what the same facts could mean for small and mid sized businesses in Türkiye. That reading is ours.
Two practical points. For an owner with no technical team but close knowledge of their own customers, the distance from an idea to a working process gets shorter, and because the cost of trying drops, what used to be tested once a year can be tested monthly. Starting and running, on the other hand, removes none of the obligations that come with invoicing, accounting, personal data responsibility and customer communication. A sensible route is to pick one small, measurable process, hand that to the model, log the output, keep the consequential decision with a person, and look at the numbers after two or three weeks before widening the scope. Repetitive parts of customer questions, quote preparation or post order updates fit that description, because the cost of an error is low there and measurement is easy. That assessment is ours.
The UNALSOFT take
In our view this is not a model announcement, it is a story about the definition of a founder. Someone who knows their customers well now has a shorter path from an idea to a running system, but which data that system touches and which approval gates it passes through is a setup question. In our Agentic AI work the opening question is not which model to use, it is which decision stays with a person.
Sources
Have the idea but no technical team?
We can map which process can be handed to a model and which decision should stay with you. A short call is enough to start.