Weight-level continual learning
Lessons move from the context window into the model itself.
capabilityEmerging Open on the map →
Facts
- Lane
- Algorithms and learning
- Wave
- 6 of 12 (build order, not a date)
- Needs
- 0 boxes all the way back through hard links
- Unlocks
- 3 boxes downstream through hard links
- Leverage
- 0.5 points · stands in the way of 2 of 53 stages · rank 16 of 57
- Cluster
- Agents at work
Evidence as of October 2026
"There is no pathway from context to weights" yet. Titans and Nested Learning are early attempts at models that update while they run.
Around it
Needs · 0
Nothing beyond the detail list. It can start at any time.
Loops and off-ramps
Openings researchers' judgement, October 2026 scouting
Startup room 52/100 · rank 26 of 124 Research room 57/100 · rank 37 of 124
- Depth
- deep
- Players
- Adaption Labs ($50m); Trajectory ($15m seed, then $40m at $300m); SSI (~$8bn); Thinking Machines (Tinker); Google Titans
- Wedge
- Systems layer for learning from production feedback on open weights: adapter updates gated against regressions, forgetting and poisoning, with rollback and a learning-over-time eval. startup