My honest field notes on the super-exponential + why “I’ll catch up later” is now the riskiest bet

Source: Nate’s Newsletter Date: 2025-12-29 URL: https://natesnewsletter.substack.com/p/my-honest-field-notes-on-working

Summary

AI capability growth is super-exponential — not just accelerating but accelerating faster, with the doubling cycle for delegatable task complexity compressing from ~7 months to ~4 months. Using Claude Opus 4.5’s 50% success rate on five-hour tasks as a baseline, Nate projects ten-hour complexity by spring, twenty hours by summer, full work weeks by fall. The warning: “I’ll catch up later” is now the riskiest bet as the gap between early adopters and everyone else widens faster than intuition suggests.

Implications

AI economics thread. The 50%/80% performance gap is the operational challenge that raw benchmarks obscure: 50% success means AI is “possible” for a task, not “dependable.” Organizations that deploy at 50% reliability will see mixed results and attribute failure to AI; those that wait for 80% may find the window of competitive advantage has closed.

Capital thread. If delegatable task complexity reaches full work-week scope by fall 2026, the implications for headcount, hiring, and organizational structure are no longer hypothetical. Capital allocation decisions made in early 2026 will be made against a capability landscape that looks materially different by year-end.

Watch: Whether the super-exponential trajectory holds — recursive AI-trains-AI loops could steepen it further, or capability plateaus could flatten it. The trajectory itself is the key variable, not any single benchmark number.