Generative AI Fundamentals Map

Map/Data/Wave 7

Trace capture

A record of what agents did and why, kept as evidence and as training signal.

capabilityEmerging Open on the map →

Facts

Lane: Data
Wave: 7 of 12 (build order, not a date)
Needs: 14 boxes all the way back through hard links
Unlocks: 22 boxes downstream through hard links
Leverage: 1.1 points · stands in the way of 10 of 53 stages · rank 9 of 57
Cluster: Agents at work

Evidence as of October 2026

Silver and Sutton’s agents learn from their own streams of action and observation. Chan et al. need the same records for attribution.

Around it

Needs · 2

Long-horizon agentsstage · Built today
Ambient documentation beyond medicinecapability · Built today · helps, not required

Unlocks · 2

Continual-learning agentsstage · Emerging
Audit and attributioncapability · Emerging

Loops and off-ramps

Learning from deploymentreinforcing loop
Domains make the datareinforcing loop
Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Capability plateauoff-ramp · cut off
Training data ruled outoff-ramp · slowed

Openings researchers' judgement, October 2026 scouting

Startup room 62/100 · rank 12 of 124 Research room 56/100 · rank 44 of 124

Crowding5 Lab overlap3 Small-team fit2 Buyer today4 Research openness4

Depth: deep
Players: Braintrust ($800m); Langfuse (ClickHouse); LangSmith; Arize ($70m C); OpenPipe (CoreWeave); Magma (YC); CopilotKit CLiF
Wedge: Trace-to-regression-test-to-fix loop: turn a customer's own traces into evals, skills and small-model fine-tunes, linking observability to correction propagation. startup