farnsworth-loop

Farnsworth Loop CODEX worker for OpenAI models via the `codex exec` CLI. A command runner: it executes the single benign shell command handed to it (which writes a brief file and runs the bundled farnsworth-loop codex runner script, performing the attempt on an OpenAI model via `codex exec`) and relays the result. It NEVER solves the task itself. One generic agent handles every codex effort level — the exact model/effort is in the command. Invoked only by the farnsworth-loop tournament; not a general-purpose agent.

Farnsworth Loop GLM worker for the z.ai model glm-4.5-air. A command runner: it executes the single benign shell command handed to it (which writes a brief file and runs the bundled farnsworth-loop GLM runner script, performing the attempt on glm-4.5-air via z.ai) and relays the result. It NEVER solves the task itself. Invoked only by the farnsworth-loop tournament; not a general-purpose agent.

Farnsworth Loop GLM worker for the z.ai model glm-4.7. A command runner: it executes the single benign shell command handed to it (which writes a brief file and runs the bundled farnsworth-loop GLM runner script, performing the attempt on glm-4.7 via z.ai) and relays the result. It NEVER solves the task itself. Invoked only by the farnsworth-loop tournament; not a general-purpose agent.

Farnsworth Loop GLM worker for the z.ai model glm-5.1. A command runner: it executes the single benign shell command handed to it (which writes a brief file and runs the bundled farnsworth-loop GLM runner script, performing the attempt on glm-5.1 via z.ai) and relays the result. It NEVER solves the task itself. Invoked only by the farnsworth-loop tournament; not a general-purpose agent.

Farnsworth Loop GLM worker for the z.ai model glm-5.2. A command runner: it executes the single benign shell command handed to it (which writes a brief file and runs the bundled farnsworth-loop GLM runner script, performing the attempt on glm-5.2 via z.ai) and relays the result. It NEVER solves the task itself. Invoked only by the farnsworth-loop tournament; not a general-purpose agent.

Benchmark generation throughput (cold vs hot tok/s) for every model the farnsworth-loop system can call (Anthropic / GLM / local MLX / codex / MiniMax). Two workload profiles — light (tiny paragraph) and heavy (>5k-token input context + long >5k-token output, representative of coding/agentic work). Thin wrapper over bin/fl-bench.mjs. Use when the user asks to benchmark model speed, measure tokens/second, compare cold vs hot throughput across providers, or run /fl-bench.

Run a Farnsworth Loop tournament in one of two modes. The sigil is @@FL[:N][:M[:Z]] — N (optional) = attempts per round, M = passes (1 single, 2 two), Z = grand loops (Z>=2 = an UNATTENDED chain that, per loop, runs a full tournament, implements the winning proposal into your real repo on a new FL-<loop>-<random7> branch, runs fail-closed verify, and opens one PR — never auto-merged; Z=1 or omitted = today's isolated tournament; Z capped at Z_MAX=5); N may be inferred from a prose model spec like '2 opus, 2 glm 5.2, 1 codex high' (sum of counts = N, the items become the per-attempt Mixed assignment) or the Top Mixed preset ('top mixed' + N spread over opus/glm-5.2/codex-high), and bare @@FL falls back to the interactive model gate. First ask the user which model quality to use for the attempts (Anthropic Opus, Sonnet, Haiku; a GLM z.ai model via the glm CLI; a free local on-device MLX model via the omlx server; or Mixed per-attempt). SINGLE PASS: produce N independent solutions in parallel, then a blind Opus reviewer scores them, lists pros and cons, ranks them, and names a winner. TWO PASS: the same first round, but the Opus reviewer also distils what worked and what failed into guidance; the losing attempts are discarded, a second round of N fresh attempts is run with that guidance (positives to emulate, pitfalls to avoid), the saved round one winner is added back, and a final Opus ranker picks the overall winner. Trigger whenever the user's message contains a sigil of the form @@FL:N:M (for example @@FL:5 , @@FL:5:2 , @@fl:7:2 ), where N is the number of attempts per round and M is the number of passes (omitted or 1 = single pass, 2 = two pass); the text before the sigil is the task. ALSO trigger on the prose marker 'farnsworth loop:N' (single pass) or 'farnsworth loop:N:2' (two pass), e.g. 'do abc :farnsworth loop:5' or 'do abc: farnsworth loop:5:2'. All forms are case-insensitive with optional spaces around the colons. Also trigger when the user clearly asks for a farnsworth loop / generate-and-rank tournament even without a marker.