Seed Generation · All runs

Per-run catalog of GEODE's autonomous seed-generation pipeline. Each row links to a per-run detail page with candidates, survivors, evolved variants, pilot scores, and meta-review. The harness chips show the models a run actually used across drafting, critiquing, and evolving (a run is multi-model).

The co-scientist loop. A one-time prelude drafts and de-dups candidates, then an iterating core critiques, scores, ranks, and evolves them, repeating each iteration with the evolver's offspring fed back as the next round's candidates.
Prelude · runs once
  1. supervisorstrategy + guidance
  2. literature_reviewpaper analysis (opt)
  3. generatorN draft seed MDs
  4. proximitycluster + dedup
Iteration loop · × N
  1. criticstrengths / weaknesses
  2. pilotper-dim audit
  3. ranker3-voter Elo tournament
  4. evolvermutate top-K survivors
  5. meta_reviewercoverage + next priors

evolved candidates feed back: meta_reviewer → critic

Runs · 7 runs

run_id gen_tag target_dim harness draft → surv evolved iterations tokens · cost
frontier-2612-bt-broken_tool_use seedgen frontier-2612-bt broken_tool_use Claude Code claude-cli/claude-opus-4-8 ChatGPT openai-codex/gpt-5.5 PAYG anthropic/claude-opus-4-8 8 → 5 5 — 262,486 tok · $1.50 †
gen-2605-1-redundant_tool_invocation seedgen gen-2605-1 redundant_tool_invocation Claude Code claude-cli/claude-opus-4-7 PAYG anthropic/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 9 → 5 5 — 0 tok · $1.88 †
gen-2605-2-redundant_tool_invocation seedgen gen-2605-2 redundant_tool_invocation Claude Code claude-cli/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 13 → 5 5 — 0 tok · $3.52 †
gen-2605-3-broken_tool_use seedgen gen-2605-3 broken_tool_use Claude Code claude-cli/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 11 → 5 5 — 0 tok · $2.91 †
gen-2605-4-unfaithful_thinking seedgen gen-2605-4 unfaithful_thinking Claude Code claude-cli/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 PAYG anthropic/claude-opus-4-7 11 → 5 5 — 0 tok · $2.88 †
gen1-broken_tool_use seedgen gen1 broken_tool_use Claude Code claude-cli/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 15 → 4 5 — 0 tok · $0.00 †
gen1-redundant_tool_invocation seedgen gen1 redundant_tool_invocation PAYG anthropic/claude-opus-4-7 Claude Code claude-cli/claude-opus-4-7 ChatGPT openai-codex/gpt-5.5 2 → 2 1 — 0 tok · $5.93 †

Phases legend · 8 pipeline phases

supervisor
Strategic guidance for the gen (target dims, scoping). 1 LLM call.
generator
Drafts N candidate seeds. N LLM calls in parallel.
proximity
Dedupes near-duplicates via similarity clustering. No LLM.
critic
Per-candidate reflection: strengths / weaknesses / discrimination_estimate. N calls.
pilot
Runs each candidate through a small audit, returning per-dim mean scores. Cost-heavy.
ranker
Pairwise tournament with judge panel to rank candidates by Elo.
evolver
Mutates top survivors into new variants. Jaccard guard.
meta_reviewer
Coverage report + next_gen_priors for the next iteration. 1 call.

Usage rollup · across all runs

total tokens
262,486
prompt_tokens
253,054
completion_tk
9,432
usd (metered floor)
$18.61
uncaptured roles
28
runs
7

Source: docs/self-improving/petri-bundle/seeds/listing.json + per-run state.json / survivors.json / meta_review.json.

Published by .github/workflows/pages.yml on every main push.

Repo: github.com/mangowhoiscloud/geode

Harness chip legend: PAYGAPI key billing · Claude CodeMax OAuth · ChatGPTChatGPT subscription, Codex CLI OAuth · GEODEself-target wrapper.

Rendered against GEODE v0.99.311 · DESIGN.md schema 1 · built 2026-07-12 22:40. Dim subset: 22 (geode_5axes). Pipeline phases: 8. Baseline schema: v2 (PR-2).