/loop — Iterative Improvement

What It Does

/loop runs the Algorithm as a loop — multiple full Algorithm cycles on the same target, each iteration building on the last. By default a human reviews and redirects between iterations. Unlike /optimize (an autonomous mutation loop), /loop runs full Algorithm passes with that human review in the seam.

The Problem

Some work doesn't finish in one pass. A skill, a prompt, a diagram, a piece of writing gets meaningfully better each time you run a full cycle on it — but only if each cycle remembers what the last one learned and what it already tried. Run the cycles by hand and you lose that thread: you re-explore dead ends, forget which approaches got rejected, and have no record of whether the score actually moved. /loop carries ISC criteria and a dead-ends ledger across iterations so each pass starts from where the last one ended.

How It Works

Each iteration is a full Algorithm cycle (OBSERVE → LEARN). The LEARN phase of one cycle feeds the OBSERVE phase of the next, the ISA tracks iteration count and cumulative improvements, and a human approves or redirects between iterations unless autoresearch mode is enabled.

Invocation

/loop --target "path/to/target" --iterations 5
/loop --target "~/.claude/skills/Art/Workflows/TechnicalDiagrams.md" --goal "make diagrams more consistent"
/loop --resume       # Resume a previous loop
/loop --status       # Show iteration history

What Happens

Each iteration is a full Algorithm cycle (articulate → climb → verify → learn) with:

• ISC criteria that evolve between iterations
• Each cycle's learnings inform the next cycle's scaffold
• ISA tracks iteration count and cumulative improvements
• Human approves/redirects between iterations

Arguments

Algorithm Integration

The iteration field tracks cycle count. (mode: is retired — never write it.) Each cycle re-enters the Algorithm with accumulated context from prior iterations.

Autoresearch Mode (opt-in)

--autoresearch switches /loop from supervised multi-pass improvement to autonomous iteration, borrowing three patterns from pi-autoresearch (davebcn87, MIT):

1. No human review between cycles — each iteration's LEARN feeds directly into the next OBSERVE. Cycle continues until --iterations reached, target met, or explicit interrupt.
2. Dead-ends ledger — ISA maintains a ## Dead Ends section. Every failed iteration appends one line with the rejected approach and reason. Resumes read this to avoid retrying rejected paths.
3. MAD confidence on iteration score — if the target has a measurable score, compute |delta|/MAD(iteration_scores) per cycle. Flag red (<1.0×) iterations as noise-floor and log marginal; do not update baseline. See LIFEOS/ALGORITHM/optimize-loop.md → Confidence Gating.

Invocation:

/loop --target "path" --goal "X" --iterations 20 --autoresearch

Default /loop behavior is unchanged — autoresearch is opt-in only. Intended for overnight runs on targets where human-in-the-loop review between cycles is too slow.

Examples

/loop --target "~/.claude/skills/Research" --goal "improve output quality" --iterations 5
/loop --target "prompts/summarize.md" --goal "more concise, less filler"

Gotchas

• Loop runs multiple full Algorithm cycles. Each cycle is a complete OBSERVE→LEARN pass. This is expensive in time and tokens.
• Set a clear exit condition. Without one, loops can run indefinitely.
• Human review happens between cycles. Don't skip the review step — it's the feedback mechanism.