Skip to main content

Dual-axis model fit — power and specialty

· 2 min read
Anchor maintainers

Anchor’s fit check used to answer one question well: is this model strong enough? Mythos-core rule 11 and model-fitness.md sent weak-column and orchestration-class work up with SUGGEST-ESCALATE. That stopped silent overreach — but it did not catch the other failure mode: a model that is strong enough yet the wrong kind of product for the job.

What shipped

Fit is now dual-axis:

  1. Power — weak column, orchestration, under-tier → SUGGEST-ESCALATE: <target> — <reason>
  2. Specialty — power OK, wrong computational/product profile → SUGGEST-REROUTE: <target or profile> — <reason>

Good fit on both axes means silence — no model pitch on every prompt.

Closed specialty profiles (v1): coding-agent, terminal-agent, critic, planner, general-chat, multimodal, swarm-local.

Example (lateral, not “pick Opus”):

SUGGEST-REROUTE: coding-agent — leave multi-file software for a software-dev optimized model

scripts/orchestrate.py honors first-line SUGGEST-REROUTE the same way as escalate (no retry burn; --insist still forces proceed). Plans may list profile tags next to tiers in Preferred models; mechanical pickers still gate on tiers and names only — tags guide self-assessment. Optional plan_fit.py --profile adds a soft specialty_hint in JSON.

Where to read it

  • Model fitness — protocol, profile table, per-model profiles
  • Doctrine — right-size diagram includes specialty re-route
  • Platform briefs: Claude Code, Grok Build, Chat, local models