Dan Jeffries: RSI Will Hit a Hidden S-Curve, Leaving AI Useful but Not Magical
Dan_Jeffries1 · x · 2026-09-29
AI commentator Dan Jeffries predicts the end of the current subsidized-compute cycle: companies are giving away massive inference and free agents, but when the "magic version of RSI" fails to materialize, the musical chairs stop.
He argues RSI won't go FOOM because there's no free lunch, physics can't be violated, most knowledge isn't verifiable, and there's a subject/object problem: "how do I know I improved when I am the judge?" AI will improve usefully in many areas but not everywhere to infinity — it becomes a hidden S-curve like everything else.
Conclusion: we'll get super useful, revolutionary but flawed, spiky, "normal-ish" machines — groundbreaking like the steam engine, printing press, or Internet, wonderful and awful at once, but not the end of all things or the start of utopia.
More from AGI Musings
- Defending Aella: Her Sex Work Doesn't Undermine the AI Safety Movement — anpaure · 2026-09-29
- 10 TED talks that teach you more about AI than a college degree — thisdudelikesAI · 2026-09-29
- David Patterson mocks AI safety rhetoric: encyclopedias contain bomb plans, so they're 'misaligned' — davidpattersonx · 2026-09-29
- Instinct CEO's agent pitch: restaurants prioritizing birthdays via agents gets mocked — himanshustwts · 2026-09-29
- Stripe and Notion converge on the same bet: agents will be both sides of knowledge work — ivanhzhao · 2026-09-29
- Ben Todd argues mass AI unemployment may arrive later: task decomposition and deployment lag — ben_j_todd · 2026-09-29