Archive — history, not state. Kept for its reasoning and its evidence; its plan is closed.

Archive

Closed arcs, superseded plans, and the historical ruling record. Nothing here describes what is being worked on now — for that, go back to State. These pages are kept so that work is not retrod: they say why something was built the way it was, and what was already tried and rejected.

Read a page here for its reasoning and its evidence. Do not read it for its plan — that plan is done, abandoned, or replaced.

Closed and superseded

• The board — every active line, one view (2026-08-04)This file is the INDEX of active work, not a new plan: each line's substance
• Clerk — ARCHIVED 2026-08-03 (C1–C6 landed)The C line landed in full. C1–C3 established file-granular acquisition,
• Compiler optimizing experiment: restart handoffBranch: compileroptimizing
• The compiler lane — throughput and output qualitySplit out of plans/one-binary.md P4/P6 by John's ruling, 2026-08-03. The
• Derivations — the shop's build model, rebuilt on closuresWritten 2026-08-02 from John's ruling, verbatim in spirit: "I do not understand
• EdictsStanding rules for this work, from John. Every subagent brief points here. When code and
• Edicts — movedThe golden rules live in the docs now, which are the single source of truth:
• The finalization arc — THE live plan (ruled 2026-07-29)The one plan in flight. The target design lives in docs/ — normative pages for this arc:
• The next arc — two streams, dispatch-ready briefsRuled by John 2026-07-26/27. Stream 1 is the recipe arc (target knowledge + meson removal);

night-2026-08-04/

• Battery speed — where the test battery's seconds actually goBranch cp/battery-speed, worktree .claude/worktrees/night-battery, base dev@7c02555dd.
• Compiler-speed trendline — night of 2026-08-04 → 05All numbers are pulled from the night's section reports (each states its
• L6 cross-builds — linux, windows, macOS from tonight's tipare pressed through pit.mcode.program@1). Artifacts live in the worktree at
• Link-time optimization — night of 2026-08-04 → 05by the agent that measured them.
• Night log — chronologicalFormat: HH:MM state — event. Times are local (America/, John's machine).
• P8 — native bake-in and direct C-call finalizationBranch p/native-bake, from af0126485 (the press switch: ship/boot presses run
• The press switch — ship/boot presses go through the P6 semantic linkerBranch p/press-switch, from dev @ 0c6f6aacd.
• Seed speed — attribution round, 2026-08-04 nightBranch cp/seed-speed-probe, worktree .claude/worktrees/night-seedspeed, base
• Warm-store staleness — a linked package's snapshot was never the truthBranch cp/warm-staleness, worktree .claude/worktrees/night-staleness, base dev@0c6f6aacd.
• Compiler switch report — what each switch changes on a real artifactMeasured 2026-08-05 (night of 2026-08-04) on branch cp/switch-report,
• Night of 2026-08-04 → 05 — Table of Contentsisolated worktrees; every landing gated and pressed per protocol. Read

• One binary — the floor ruling and the mcode-linker arcWritten 2026-08-02 from a design session with John. This plan supersedes item 1 of
• Overnight run — judgement calls made without a ruling (2026-07-28 → 29)John left with: "Continue on and finish out 6 and 7... If you encounter blockers like bad

perf-2026-07/

• Direct call op: call dest, fn, argc, args... replaces frame/setarg/invokeDate: 2026-07-08. Baseline: e07750e6 (frame-stack harvest). Gate: make check
• Frame-stack harvest: call frames stop being GC objectsCherry-picked the seven-commit stackframes lineage (two-ended actor memory
• Profiling axes: separating what carries to native from what doesn'tPerformance here has three distinct dimensions, and improvements must be
• resolve v2: parameter backward inference + intrinsic returns + write-model soundnessDate: 2026-07-08. Baseline: 58f6ecdd (direct call op). Gate: make check
• calling sequence v2 — measured cost attribution (before build)Base: b1dbf1e6 (branch work/calling-v2), make check ALL GREEN. Machine
• Canonical CLI one-shot contractThe terminal is transport and presentation, not an actor-vocabulary parser.
• The compiler contract — constraints a resolve/emit rewrite must honorStatus: inventory extracted from the current implementation (tokenize.cm, parse.cm,
• Federation, Playdate, and Apple-platform checkpointStatus: implemented and locally verified on 2026-07-14.
• Active structured mid-end: exact value lanes across finalized mcodeStatus: isolated performance experiment on codex/perf-active-structured-midend.
• Dynamic call activation-cache experimentOriginal measurement branch: codex/call-activation-experiment at
• Pit language performance campaign — 2026-07This is the audit ledger for the contained performance experiment based at
• Compiler optimization campaign: final decisions and tradeoffsBranch: compileroptimizing
• Compiler optimization campaign: final performance matrixStatus: final integrated campaign seed measured and validated. External-runtime
• Mach exact type-span consumption and constant-load fusion auditExperiment branch: codex/perf-mach-exact-fusion
• Mach guard/branch fusion experimentBranch: codex/perf-mach-guard-fusion
• Mach nullable-numeric decode experimentBranch: codex/perf-mach-guard-fusion
• Mach text-guard fusion and rejected local-fusion experimentsBranch: codex/perf-mach-guard-fusion
• Native suspension and freeze-safe callsStatus: required architecture and experiment gate; no implementation accepted.
• CLI reply deadlineStatus: superseded by the canonical one-shot transport contract.
• Closure outer-frame fast-path experimentDate: 2026-07-14. Campaign baseline: compileroptimizing at the shared
• Compact linked value/origin/effect spineStatus: isolated Gate-1 representation and benchmark-consumer experiment. It
• Integer-range analysis budget experimentThe compiler used to skip intranges whenever a function exceeded 128 mcode
• Exact linked pure-leaf substitutionStatus: rejected by the production performance gate on 2026-07-14. The active
• Linked optimization, shapes, and PGO arcBranch: compileroptimizing
• Native guarded-array store experimentBranch: codex/perf-barrier
• Native complete checked-array-store foldStatus: accepted isolated experiment on 530a9fcf; not merged upstream.
• Compact native raw-recursion continuationsExact base: 0401fb314a71311a476cd140f3930a242e6268bf
• Native fannkuch loop-pressure experimentBaseline: 7fca24a4 (the exact parent performance-campaign tip when this work
• Native raw signed-int self recursion experimentMeasurement base: e0a4a67e
• Native record-literal templates: experiment reportStatus: implemented and validated on the isolated performance branch. This is
• Native suspension experimentStatus: research-only; do not merge.
• Native suspension hot-path follow-upStatus: completed isolated experiment, based on 836a3151.
• Unified native suspension experimentStatus: completed composed integration candidate.
• pit performance — state of the world & the next partWritten 2026-07-10 at the close of the first perf arc, on the fully merged
• Final nursery sizing sweep protocolStatus: sweep executed on campaign checkpoint da2a46b5; results and
• Generational nursery revisit (2026-07-13)Branch: codex/perf-nursery-revisit
• Native raw-number experimentAccepted in the contained performance campaign. This experiment retains
• Dense static record shapes — performance experimentStatus: retained experimental implementation on compileroptimizing, with
• Spectralnorm inline-callee cleanup and generated-code auditStatus: accepted on the experiment branch after composed artifacts, strict A/B,
• Spectralnorm transient raw inline chainStatus: retry rejected; compiler implementation deleted. The QBE-only proof
• Structured ordinary-loop numeric phisStatus: accepted isolated candidate on codex/perf-structured-loop-phis, based on
• Structured native numeric lanes for self-TCOStatus: contained performance experiment, validated and ready for composition review.
• Wide integral literal correctnessStatus: retained correctness repair; final integrated campaign validation is
• pit-native (AOT) vs mach — benchmark findingsMeasured 2026-07-09 on this branch (macOS arm64, Opus session). Harness:
• Shop actor and native-runtime split handoffStatus: merge checkpoint prepared 2026-07-15.

• The v1 plan/cart format — complete writer/reader map (recon, 2026-07-28)Produced by the keystone recon sweep for landing 4b. This is the COMPLETE baked-caller
• Review queue — ARCHIVED 2026-07-25for later review. He now answers as things come up, per plans/method.md, so rulings go
• Roadmap — from the green base to what the docs describedocs/ says how the system should work. This says what is between here and there, in
• Stones 6 and 7 — the execution map (recon 2026-07-28, read-only, verified against the tree)Produced because the roadmap's rows 6 and 7 are each one line and each is 5-6 landings.
• What to do nextOne list, ordered roughly by leverage. Each item says how to start, so picking one
• System overhaul transition diaryStatus: CLOSED 2026-08-01, audited entry by entry against the post-L3/L4 tree
• Worklist — endowment, gate, and boot modelGenerated 2026-07-24 from a design review of docs/ against John's rulings. This is a