“If you care a lot about computer use, there is nothing even close. This model is not just like a single generational leap in computer use. It feels like two or three. It’s absurd. And it makes Anthropic models almost feel like they’re in the stone ageโฆ Meanwhile, Codex uses my computer more than I do at this point.”
โ @theo on OpenAI GPT-6 Astra computer use โ two or three generational leaps; Codex uses his computer more than he does.
Clip via the episode โ @theo on OpenAI GPT-6 Astra computer use โ two or three generational leaps; Codex uses his computer more than he does. / It’s Here.
The benchmark that matters isn’t a benchmark.
It’s a practitioner saying, in plain language, that a model now uses their computer more than they do. That sentence is the story.
When a developer with real production exposure describes a gap as feeling like “two or three generational leaps” โ not one โ and names a specific competitor as the comparison point, that is signal worth parsing carefully.
THE CLAIM
@theo’s argument โ as expressed in this episode โ is that GPT-6 Astra is categorically ahead on computer use, not incrementally. “Stone age” is a strong word. It’s also a precise one: it implies a discontinuity, not a gap that closes with a product cycle.
“Codex uses my computer more than I do at this point.”
โ @theo, as quoted in the clip. This is his stated experience, not a verified metric.
THE STRUCTURAL READ โ FOURWEEKMBA ANALYSIS
This maps directly onto what we call the Product Overhang Doctrine: capability accumulates in the substrate, invisibly, until one release surfaces it all at once. The practitioner community didn’t see a gradual climb โ they saw a cliff.
If @theo’s read holds at scale, the competitive dynamic shifts from “which model is better at coding” to “which model can autonomously operate the full computer environment.” That is a different product category, not a better version of the same one.
Why “computer use” is the right frontier to watch
Text generation and code completion are table stakes. The next layer of value is an agent that can operate the machine โ navigate UIs, run processes, manage files, interact with software the way a human does.
That’s not a feature. That’s the beginning of a workflow replacement story. And according to @theo’s account, one model has pulled dramatically ahead on exactly that capability.
The companies that win won’t be the ones building these models โ they’ll be the ones who route workflows through the best computer-use layer the fastest. If Codex is already operating computers autonomously at a practitioner’s level, the “harness” question becomes urgent: who is building on top of this, and how quickly?
WHAT TO WATCH โ FOURWEEKMBA ANALYSIS
Competitor response cadence. A “two or three generational leap” claim from a credible practitioner creates real pressure on rival release timelines. Watch whether Anthropic and others accelerate computer-use-specific announcements in the near term โ that’s the market signal that validates or refutes @theo’s read.
THE BOTTOM LINE
One practitioner’s account doesn’t close the debate โ but “my AI uses my computer more than I do” is the clearest plain-language signal yet that autonomous computer operation has crossed a threshold worth paying attention to. The question now is whether the rest of the market is reading the same signal.
This is FourWeekMBA editorial analysis of a public podcast clip. @theo’s statements represent his own expressed view from that episode โ they are not verified benchmarks, established facts, or predictions. Nothing here is investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur, HackerNoon and Search Engine People. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
