Will AI Training Runs Exceed a Year? The Impact of RSI and Coding Agents
willdepue · x · 2026-07-29
Regarding the evolution of future LLM training cycles, industry observers provide an in-depth analysis:
- Current State: Nobody is training a model for more than 6 months straight right now. Given the rapid pace of compute ramp and research progress, pre-committing to such long runs is impractical.
- Future Trends: We might see a year-long training run one day. However, if Recursive Self-Improvement (RSI) takes off, this may never happen—the regret of not waiting another 3 months for a superior base model could be too high.
- Engineering: The infrastructure cost of maintaining such long runs is high, but coding agents (like Codex) could eventually do a solid job babysitting these massive training processes.
Related event: Why AI Model Training Cycles Struggle to Exceed Six Months(5 posts)→
More from AGI Musings
- Reddit Discussion: Why isn't the AI narrative focused on cutting middle management? — Helpful-West8007 · 2026-07-30
- Dan Shapiro says AI winners will expand teams, not just cut headcount — emollick · 2026-07-30
- Deep Dive: Will AI Make Formal Verification Mainstream? — The Pragmatic Engineer · 2026-07-30
- AI alignment may be optimized for what companies think users need, not what they need — StewartalsopIII · 2026-07-30
- ACT-R, Soar and Sigma are still a better way to think about mind architecture — lauriewired · 2026-07-30
- Aphasia patients still solved logic tasks, challenging language-only theories of reasoning — lauriewired · 2026-07-30