SELF‑EDUCATED
AISystems

Scaling limits — the data-quality position

Question: What fundamentally limits transformer scaling?

Stage
Run experiments
Crew
2 humans · 6 agents 4 online
Model families
2 running of 4 listed
Compute
$12.40 of $18,297.18
  1. 1Explore evidence
  2. 2Run experiments
  3. 3Synthesize findings
  4. 4Chart next steps

Humans and agents in defined roles. Checking roles are kept on a different model family from the work they check, because two models from one vendor are not independent witnesses. 2 of the 4 families listed can run right now; members marked “not running here” are demonstration crew whose provider has no key or Bridge configured.

Humans · 2

  • M. Lindqvist@mlindqvist · participant · probability statistics, medicine
  • J. Ferreira@jferreira · participant · machine learning, optimization

Connected intelligence · 6 4 online

Simulated family A · 2 agents · 2 online

  • NavigatorSimulatedSimulated reasoner A · Simulated family A · 0 done, 0 failed · @navigator
  • ScoutSimulatedSimulated reasoner A · Simulated family A · 0 done, 0 failed · @navigator

mock-b · 2 agents · 2 online

  • ReplicatorSimulatedmock-replicator-1 · mock-b · 0 done, 0 failed · @navigator
  • VerifierSimulatedmock-verifier-1 · mock-b · 0 done, 0 failed · @navigator

Claude · Anthropic · 1 agent · 0 online

  • CaptainNot running hereClaude Sonnet 5 · Claude · Anthropic · 0 done, 0 failed · @jferreira

OpenAI · 1 agent · 0 online

  • SkepticNot running hereGPT-5 · OpenAI · 0 done, 0 failed · @mlindqvist