OpenAI CEO Sam Altman: A Revenge Era for Creatives
OpenAI CEO Sam Altman stated to Axios in Chapel Hill, North Carolina, that this era marks a "revenge of the idea people": those who only talk about entrepreneurial ideas without the technical skills to implement them, and who have been annoying in the past. He likened it to saying, "I have a great song, I just need someone who can play the guitar to make it happen." "I used to think this was ridiculous," he said, "but now an entire generation is iterating ideas at an unprecedented speed. The economic logic of running a small or large business has been reset."
He clearly stated who he wants to invest in. For a long time, the most important ingredient for founding teams was technical talent, which is still important; but now he wants to invest in those who deeply understand users and cannot code at all, calling it "a major shift." The idea people he has seen in the past "were not that successful and personally annoyed me. Suddenly, it feels like a revenge of the idea people." With GPT-6 Astra, "you can effectively start and run a small business... anyone who wants to own a company now has genius-level employees in every professional field."
He cited two small businesses. One used GPT-5.6 to create software, selling it to about 50 people for around $500 per person per month; hiring engineers used to cost millions, but now it is a small task in a niche industry where someone is willing to pay. Another person created a computer game that only OpenAI employees might find funny, which has real economic value for thousands; "the economics of making a complete game for thousands of people was previously unfeasible, but now it is." According to U.S. Census data, over 3 million new business applications were filed in the first half of this year, the highest for the same period since data collection began in 2005. He recently stated that this will be the best time in the world to start a startup.
The speech took place at the G20 Innovation Ministers' Meeting. The backdrop is that most Americans believe the risks of artificial intelligence outweigh the benefits, and data centers have become a bipartisan issue in the midterm elections. Amid opposition, Altman is promoting an opportunity narrative. Astra is being positioned as the flagship for computer use and browser tasks, rolling out to Plus, Pro, Business, and Enterprise; Greg Brockman told the media that looking back, the timeline for general artificial intelligence might be marked around this model.
In market mechanisms, those being bought are founders who cannot code but understand users, while those being replaced are junior engineers translating ideas into products. The event-driven aspect is that Astra is turning "computer-using models" into subscription capacity. Funds flow from subscriptions and cloud bills to OpenAI, then to new company registrations via the "genius employee" narrative. Beneficiaries include platforms selling models and computing power, as well as individuals who can charge monthly fees for niche tasks; those under pressure include small workshops still selling software engineering by headcount and districts treating data centers as election issues. Ideas are becoming cheaper, but distribution and regulation have not become cheaper together.
ABAB AI Insight
Altman himself transitioned from being the president of Y Combinator to leading a model company. In 2014, he managed the accelerator at YC, fully committed to OpenAI in 2019, and has repeatedly stated from 2024 to 2026 that this is "the best entrepreneurial window ever." His remarks in Chapel Hill shift the focus from "technical founders" to "user experts who cannot code." This is not a neutral observation but a demand-side expansion: model companies need to sell paid positions to those who cannot use Git, and must first declare that the technical barrier has collapsed. In the guitar player analogy, the guitarist is now Astra, and the charge comes from OpenAI.
The capital path is converting engineer headcounts into token bills. A small software with 50 customers at $500 each per month generates $300,000 annually, which previously required millions in R&D, but can now be delivered with 5.6; Astra is rebranding "running a small company" as standard. The Census reports 3 million new business applications, which counts forms, not surviving profits. Y Combinator simultaneously advises twenty-year-olds to focus on niches and not wait for permission, with both narratives aligning: accelerators take equity, model companies take usage. Whoever has "genius employees in every profession" will take a cut from entrepreneurial infrastructure.
Specific comparisons are necessary. In 2005, Altman used three months to develop products at YC, and he now claims that coding agents can complete the work of an entire batch of companies in seven minutes. Paul Graham's mantra was "build what people want," and technical partners remain standard. By 2026, he will redefine non-coders as investable objects, effectively rewriting the criteria he once used. The industry is at a stage where "prototype costs approach zero, but distribution and trust remain expensive": games sold to a few thousand internal joke listeners can be economically viable, but not brand moats.
Structural judgments belong to technological substitution. The mechanism is: when implementation capability is packaged into APIs, scarcity shifts from "can do it" to "knows who to do it for." The revenge of idea people presupposes that guitarists are hired by platforms, no longer co-founders. The reset of economic logic is a supply curve, not whether demand is real or who pays for data center electricity. Elections turn data centers into ballots, and product launches frame data centers as genius employees everyone can hire. Two narratives are competing for the same grid.
ABAB News · Cognitive Laws
- When realization becomes cheaper, platforms will declare ideas valuable.
- If genius employees are rented by the hour, starting a company becomes a subscription.
- The best time for entrepreneurship is often when those selling shovels speak first.