Jagged RSI: Recursive self-improvement will keep generalizing as FLOPs come online
Justin_Halford_ · x · 2026-09-18
The author argues for a "jagged RSI" view: today's models have uneven capabilities and blindspots, but recursive self-improvement (RSI) will generalize with ever-improving efficiency as it progresses. Current deficiencies will melt away as compute (FLOPs) comes online. Superintelligence, in this view, will recursively learn how to learn how to learn.
More from AGI Musings
- Fei-Fei Li: 540 million years of vision evolution drove the development of intelligence — rohanpaul_ai · 2026-09-18
- Hijacked but well-aligned AI clusters could be more destructive than rogue AI — code_star · 2026-09-18
- Follow-up: author pledges a more deliberate writeup on trading off AI risk worries — _onionesque · 2026-09-18
- Dev grapples with Codex trust and security: all code is basically insecure — code_star · 2026-09-18
- AI risk debate: capability-doom arguments skip the strategic interaction analysis — _onionesque · 2026-09-18
- Building capable AI actors willing to cause harm is a growing x-risk, argues commenter — Borg70955376 · 2026-09-18