generate_via_llm consumed each validator Rejection internally (`last`), fed it into the
next attempt's prompt, and dropped it. So Step 5 was real but unobservable: a caller could
see THAT a proposal validated, never that it validated on attempt 2 after the deterministic
validator falsified attempt 1. It was the one step of the eight with no output to show.
The seam is a typed return value -- GenerationResult(outcome, refinements) -- rather than an
out-parameter or a callback: a returned value cannot be silently lost by a caller that forgets
to pass a collector, and mypy forces every call site to acknowledge it.
refinements carries ONLY rejections that were actually fed back. When the attempt budget runs
out the final rejection IS outcome; counting it here would be double-counting, and the bounded
control test goes red on the collect-everything implementation that gets this wrong.
The loop's bound is untouched: max_attempts and meter.tick_round stand, and `last` still drives
the prompt alone, so prompt growth is unchanged. run.py accumulates across _evaluate calls, so
_evaluate_mandate is untouched; RunResult.refinements defaults (the coverage precedent) and is
concatenated across approaches rather than keyed per approach -- stated as an honesty limit.
The simulation now shows it: the scripted proposer overclaims 250000, which the validator
falsifies against P90 = 90000, and the corrected 30000 validates. Only the overclaim is
scripted -- the rejection is computed. scripted_factory takes a per-role reply selector so this
needs no second scripted client body.
README records the two accuracy changes only (Step 5 is now inspectable; the simulation trace
shows the correction). The level-2 publishing claim stays deferred until after the demo (O4).
Load-bearing MEASURED against the full suite with a control, four mutations all red:
detach the returned history (4 tests) - collect-everything (control only) - detach the run
wiring (2 tests) - revert the simulation's proposer to a constant (the demo-protection test).
Control: 759 passed / 4 skipped; ruff, format and mypy clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017CcWFcREUi6YPjEpN3ACDP
`docs/bestille-en-kjoring.md` is the commissioning half of the expert-facing pair
(`ekspert-svar.md` is the judging half): the mandate file field by field, how to
run it, and — separated deliberately — what a commission does NOT do. It directs
what is evaluated, never what is approved.
Registered in _LIVE_DOCS, so it cannot silently fall behind the code.
The example output in it is COPIED FROM A REAL RUN, not composed, and running
that run is what found the defect fixed here: three approaches against the same
cost line each validated at 30000 NOK, and the settlement printed
"Validated total: 90000 NOK". Commissioned approaches are ALTERNATIVES — they
usually attack the same line — so summing them reports money the project cannot
realise. A domain expert reading that total would reasonably believe the run
found 90k.
The settlement now reports how many approaches held and which one the run
carries: a selection, not an arithmetic claim. That also removes the last money
addition from this module, which is the right place for it not to be.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ULCqjLF61rehj5cZmdUoR3
Walked Door A from a fresh clone: materializing the file-family golden
manifest writes index.md plus one concept file per extraction, and pointing
the run at the result is refused —
run refused: IR projection not found in bundle: 'validator-input.json'
A clean fail-fast, but nothing adopter-facing said it was coming, while the
README actively invites it ("swap --bundle-dir for your own bundle"). The
run path needs the bundle's IR projection, which ingest does not and cannot
produce: ingest materializes source documents, the projection states the
candidate measure. Both docs now say so, with the shape reference named.
Also corrects a live-doc claim that was wrong in both halves: the MCP
timeout is `anyio.fail_after` nested inside both task groups, not
`asyncio.wait_for`, and `tests/test_ingest_golden_mcp.py` covers it
(verified — 2 passing timeout tests). And no bundled example ships a
`cost-baseline.json`, so the text no longer implies one does.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GyAbxJoyypnLLUDcMvnKh8
An independent reviewer found it and the claim was verified by measurement
before being accepted, not taken on trust.
test_portfolio_scripted_pass_makes_no_real_client raised from the patched
_default_factory to prove the scripted portfolio pass never builds a production
client. It cannot: the factory is called inside run_project's coroutine, and
run_portfolio gathers with return_exceptions=True, so the AssertionError was
collected into a RunFailure and never escaped. Measured directly -- with BOTH
the client_factory wiring and the rc rule detached, the blade stayed GREEN. It
was going red on rc alone, which means it was testing defect B while claiming
to test defect A. The original seven-mutation sweep did not catch this because
each mutation was applied singly, and dropping the wiring alone still flips rc.
Replaced with a call sentinel: a list appended inside the factory and asserted
in the test body, which the wave handler cannot swallow. Re-measured -- red on
the wiring detach alone, and red on both detaches together.
The docstring now also states what the sweep could not: blades 1 and 8 are
environment-conditional. The local profile points at loopback, so on a machine
running a local model server the wiring detach would make real calls and could
complete the pass. Their red was real on the machine it was measured on and is
not portable; blade 2's is.
Separately, the budget-stop print is marked as defensive and currently
unreachable from main(), because main() never constructs a PortfolioMeter and
every write to budget_stop is gated on one -- the strict=True precedent
directly above says untested future-proofing must be labelled as such. The
README claim that a portfolio pass reports a cap stop is corrected to say the
cap has no CLI flag yet. Noted for whoever wires that door: BudgetRefused is a
RuntimeError and the existing except clause would not catch it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0118noV9rCfrdREH26XqZB5z
The operator is not a domain expert, so the domain content is mine to own --
and the one thing the loop asks a human for is exactly the thing no example
existed for. docs/ekspert-svar.md is written for whoever has to deliver the
verdict: the two forms a judgement can take (a --rationale string during the
run, a JSON file in the inbox for later runs), where each field comes from, and
four complete paste-ready answers.
Every command and every verdict in it was RUN from a fresh clone before it was
written. The `hitl pending` line quoted is verbatim output. The rejection
answers close the gap STATE has carried since the demo shipped: the README
shows the VALIDATOR refusing a number, but nothing showed an EXPERT refusing a
proposal whose numbers are fine -- the only judgement in the whole loop that a
machine cannot make. Two rejection shapes are given, because "not feasible
here" and "right measure, wrong cost base" teach the system different things.
Everything is marked AI-authored and not verified professional judgement.
Also corrects the --outbox-dir help text, which claimed sharing a folder with
--verdict-dir "re-ingests raw agent output past the Step-8 promotion gate".
Measured, by pointing both at one folder and running twice: it does not. The
outbox artefacts are named {run_id}-*.json and carry none of the verdict keys,
so the tolerant inbox loader skips them and the run is unaffected. The hazard is
real but latent -- a future verdict-shaped artefact in the outbox -- so the
warning stays and says what is actually true. This also answers STATE's open
question about enforcing the distinction in the CLI: no. There is no reachable
contamination to refuse, and a guard for an unreachable case is the kind of
error handling this repo declines to write.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0118noV9rCfrdREH26XqZB5z
Steps 6 and 7, each run verbatim from a fresh clone before being written.
Step 6 is the portfolio pass, which the walkthrough could not reach until the
scripted door was wired into portfolio mode. It is also the clearest single
demonstration the framework has: all four reference projects carry a cost line
01.1 at four different amounts, so one unchanged proposal yields one
ValidatedProposal and three Rejection -- the gate is anchored to each project's
own baseline, not to the proposal's internal arithmetic. The shared verdict id
is explained rather than hidden: a verdict is keyed on the candidate, not the
project, and that key is how a later run finds the earlier judgement.
Step 7 documents the value report and, more importantly, the gap a downloader
hits first: nothing in the shipped code writes a savings ledger. Measured, not
assumed -- no .save call on a ledger exists outside the library API. That is by
design and is now said out loud: the ledger records savings actually realized in
the world, which is not a conclusion the system may draw from its own proposals.
A validated proposal is a claim; a ledger entry is a result. The refusal a
downloader sees before creating one is quoted, and the library snippet that
creates one is shown.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0118noV9rCfrdREH26XqZB5z
Measured before writing: the commands existed but were scattered across the
mode partition, and there was no path a newcomer could walk end to end. The
capability gap that made a complete offline walk impossible is closed in
3abc61b; this is the door onto it in the README.
Five steps, each RUN FROM A FRESH CLONE before being written down (git clone +
uv sync + uv run pytest -> 652 passed): read the knowledge base, watch the
learning loop close, run the loop with your own scripted answers, watch the
validator say NO, and price a real run before spending anything. Step 4 is the
one that was missing entirely -- the demo only ever showed a yes, and a
refusal carries far more weight than another approval.
Also documents that `Rejection (..., decision=approved)` is not a
contradiction: the first is the validator's outcome, the second echoes the
human's recorded verdict. Documented rather than changed -- altering a public
output format is the operator's call, not a side effect of writing docs.
`costsim` gets its first mention in the README at all; it was finished, tested
and completely invisible from the surface.
Also points the commons link at open/ (published 2026-08-04). NB: the
repo-standard v0.3.0 register still lists 19 repos and does not know that repo,
so pointing at the correct live URL now trips a false LINK-DEAD. Reported to
org-ops; the URL answers HTTP 200.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GWsexbQjPo9rsV3aUE54ZS
Ran `repo-standard` (v0.1.1, class `standalone`) and fixed everything it
flagged as ERROR, plus the WARN links that were genuinely dead.
README first screen:
- opening line is now byte-identical to the forge description, so
description == catalog == README is machine-checkable (badges moved below).
- `## Install` (required for class `standalone`): clone + `uv sync`, stated as
clone-only because the shared spec, persona skill and example bundles under
`shared/` are read from the working tree at run time. `uv run pytest` named as
the verification, with the fact that no CI runner exists said out loud rather
than implied by a badge.
- `## Non-goals` (required): the five limits already binding in CLAUDE.md —
not a compliance product, not a portfolio-level reallocator, not autonomous
decision-making, not turnkey, not a model benchmark.
Dead relative links (measured, not guessed):
- `docs/plan/2026-07-10-sesjonsplan-fase2-6.md` pointed at
`../2026-07-14-revisjonspakke-DF-DI.md` six times; the file sits in
`docs/plan/`, not `docs/`. (The sibling `../review-2026-07.md` links are
correct and untouched.)
- the Fase-1 spike brief linked repo-root-relative from
`.claude/projects/…/`; re-anchored with `../../../`.
The one remaining README ERROR was a gate false positive: `checkInternalLinks`
resolves targets against `git ls-files`, which lists files only, so a link to a
directory can never resolve. `[shared/](shared/)` now points at
`shared/README.md` — a better target anyway, since that file carries the
pull-only subtree rule. Not fixed here: the classifier lives in another repo.
Remaining WARNs are all inside `shared/`, deliberately untouched: it is a
pull-only commons subtree, and the nav-golden files are byte-level fixtures
that gate `test_nav_golden_*` — four of them are OKF bundle-internal links,
and the `/etc/passwd` ones are the negative escape fixture doing its job.
Suite green: 630 passed, 4 skipped (markdown-only diff; no test touched).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ri3aVJPfynCZtHRhesCzUH
Every stage of validate_proposal reasoned only about numbers the proposal itself
supplied, so an internally-consistent hallucination cleared the whole gate (F3).
A new stage 0 reconciles each affected_item against the project's CostBaseline
before the CBC solve: an unknown cost code is rejected, and a real code carrying
a quantity/unit_cost outside the configured tolerance (5% default, relative to
the baseline value) is rejected. Validation, never repair.
The baseline argument is OPTIONAL (None = pre-S4.0 behaviour), but both run
paths set it: the road path projects project.cost_items, the bundle path loads
cost-baseline.json when the bundle ships one. Bundles written before the
amendment stay un-anchored, so the commons-owned goldens run byte-identically;
a baseline that exists but is malformed still raises on both loaders.
F8: the method-specific cap now comes from the METHOD_CAPS registry (measure
type -> fraction, injectable) instead of an energy_efficiency string comparison.
The baseline format and tolerance semantics were decided locally — the commons
amendment (D-A pt. 2) never arrived, exactly as in S3.2. D7 mirroring stays open.
Three portfolio fixtures quoted cost codes belonging to OTHER projects; the new
gate caught them. They now quote each project's own lines, and the two copied
REPLIES tables import the single source instead of drifting from it.
Load-bearing measured (tests/test_s40_cost_baseline_loadbearing.py), six
mutations all red: detach the reconciliation stage; detach the magnitude
tolerance; detach the road wiring; detach the bundle wiring; ignore the injected
cap registry; make the optional loader tolerant of malformed content. Control:
with the road wiring detached the repaired portfolio fixtures still pass, so
they are not masking the seam. 597 -> 612 tests.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JdwK7bQ4BZkWH4t8MRDKb4
seed_store_from_bundle keyed EVERY `type: verdict` file on bundle_candidate_features — the single
candidate the bundle's validator-input.json describes. A bundle carrying verdicts about several
candidates collapsed them onto one key, so a verdict about candidate B scored a perfect structural
match against candidate A's query and could be folded into A's hypothesis prompt. The ExpeL
substrate was single-candidate by construction.
A verdict file may now carry its own structural key in frontmatter (affected_codes / measure_type /
claimed_saving_nok); absent, keying falls back to the bundle candidate, so every pre-S3.2 seed keeps
working unchanged. promote_verdict writes the three fields, so a promoted verdict — frequently about
a different candidate than the target bundle's projection — does not impersonate that candidate.
Semantics decided HERE, not pulled: commons' seeding rule (method-spec §3 Steg 1 + bundle example)
has not arrived; we said we would build locally first. D7 mirroring stays open.
- ALL THREE fields or none. A partial declaration raises VerdictFrontmatterError rather than merging
with the bundle candidate, which would mint a key belonging to NEITHER candidate. Validation,
never repair (mirrors write_concept_file); the tolerant-skip rule belongs to the RAW inbox layer.
- claimed_saving_nok parses via json.loads — the SAME literal rule the IR projection went through —
and is written back with str() of the raw value. _mint_id hashes that value, so 30000 and 30000.0
are different keys; a normalising writer would split one candidate's signal across two ids.
- The structural key is signal-free, so it does not weaken the Step-8 no-leak property (Test C green).
Load-bearing MEASURED, five mutations all red: detach per-verdict keying · detach the fields
promote_verdict writes · make a partial/unparseable key tolerant · normalise the magnitude on write ·
remove the fallback (control — breaks the step1 suite at collection, proving the fallback bears load).
589 -> 597 tests. Full gate green (pytest, ruff, mypy).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QkjvTTxrg9LTrmghebfiij
PortfolioBudget + PortfolioMeter carry ONE token ledger over a whole portfolio
pass -- and, seeded from a persisted spend file, across passes -- while the
per-run Budget/TokenMeter pair is untouched. Three enforcement points, each
doing a different job:
- startup: a remainder that cannot fund one run raises BudgetRefused before
anything loads (a pass that can afford zero projects is a caller mistake,
not a result);
- wave assembly: an unfundable project is NEVER STARTED and the pass stops
structurally (budget_stop + stopped_early, completed runs preserved).
Because every member of a wave is funded against the SAME pre-wave
remainder, admission RESERVES each member's requirement -- otherwise a wave
of k over-commits the cap by up to k runs;
- pre-call: BudgetMiddleware refuses a call the remainder cannot pay for
instead of making it. The post-charge check stays: real usage is only
knowable after the response, so the guard stops the NEXT call, never the
one in flight.
budget_stop is its own field rather than a widened stop_reason -- a goal-stop
is success, this is resource exhaustion, and fusing them would make "we
stopped" unreadable. PortfolioMeter splits record/check so tokens the provider
already billed reach the ledger even when the same charge breaks the run's own
cap. read_spend raises on corrupt content (our own accounting state, unlike
the tolerant RAW inbox layer); write_spend takes a REQUIRED stamp with no
wall-clock default, mirroring promote_verdict.
Load-bearing MEASURED, not asserted -- 6 mutations, all red: detach the wave
check; detach the pre-call guard; detach the wave reservation; check the run
cap before crediting the global ledger; detach the startup refusal; make
read_spend tolerant. Files restored from shasum-verified copies after each.
Two findings worth keeping: the pre-call guard MASKS a detached wave check if
the test asserts on overspend (spend stays under the cap either way), so the
load-bearing assertion had to become failures == () plus never-started; and
the token arithmetic is probed (32 tokens/run at tokens=8), not guessed.
537 -> 553 tests, ruff + mypy green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015EaxFnaDAbMQkmTeX4u7sd
New English docs/knowledge-base-recipe.md grounded strictly in the D-H decision record
(revisjonspakke-DF-DI.md §3): setup is always a small team (technical + domain expert), the
deliverable is a recipe NOT a wizard (B9 onboarding interview + guided verdict command rejected),
domain expert delivers files in their own formats never schema/JSON, phased process (inventory ->
skeleton -> seed verdicts -> iterate), reading via Obsidian/VS Code. The honest 1-2 week
expectation is stated early and SOURCED verbatim to the record. Factory-dependent parts
(free-format verdict translation, clone-to-demo) are explicitly marked future/blocked-on-toolkit
so the doc never claims above the evidence level. Linked from README's Docs section with the
1-2 week expectation in context. SC5 (ASCII-only greps): file exists, '1-2 weeks' x2,
'knowledge-base-recipe' in README. src/ untouched; 431 passed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KNNiJRk1sSwxgVLS5AobT1
README: the stale one-line CLI mention replaced by the two-mode flag matrix (single-project vs
--portfolio), runnable 'uv run python -m portfolio_optimiser.run ...' examples, the --outbox-dir
!= --verdict-dir self-contamination warning (documented, not enforced), and the --decision/
--rationale inert-in-portfolio note. Honesty scoping: the prior-verdict fold (the learning step)
is stated to happen ONLY on the --bundle-dir path; a --docs-dir-only run is single-shot (no fold).
CHANGELOG [Unreleased]/Added: S5.3 CLI-parity entry (six flags + portfolio mode + load_dimension +
recipe doc) plus catch-up for the shipped-but-undocumented S4.1 (preflight), S4.2 (--live-dry-run),
S5.1 (hitl CLI), S5.2 (notify); stale test count 237 -> 431. extending.md verified accurate (B11
notifier note stands verbatim — main() auto-wires no notifier; no CLI section to sync), left
unchanged.
SC4 honesty grep clean (each hit in bundle-path context, none on a fold-less path):
grep -rniE 'learning loop|learns from|self-improv' README.md docs/extending.md CHANGELOG.md
README:11 (system-level) :56 (wiki substrate) :67 (8-step bundle loop);
CHANGELOG:13 (gated ExpeL fold) :17 (offline simulation). Full suite 431 passed (no code touched).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KNNiJRk1sSwxgVLS5AobT1
Rewritten around what the project is built on — Karpathy's LLM-wiki idea
as formalized by Google Cloud's Open Knowledge Format (spec-linked) — how
the 8-step loop works and is set up (shared commons core, per-project
bundles, specified ingest layer), and what the architecture enables.
Adds the AI-first framing: the wiki is written for the model, human
affordances (verdict inbox, promotion gate) are layers on top — the
inverse of a human-first wiki. Honest status kept: offline proof only,
ingest specified but not implemented.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AaQCFnfsh3tfq1VfzdJpoi
Findings 4-7 from the 2026-07-02 status analysis, per the session plan (S1):
- CHANGELOG rewritten truthfully (was: 'Plan phase - no framework code yet')
- README stack line names the split GA packages, not the agent-framework meta-package
- CLAUDE.md: MCP downgraded to extension point (in-process FunctionTool is the default seam)
- Verdict conflict semantics documented as chosen (store first-write-wins per id,
disk/wiki last-write-wins per file; full B10 taxonomy deliberately deferred)
- docs/extending.md: explicit 90%-principle cut-list (B10, B11, U12, U14, concurrent fan-out)
- .gitignore covers .trekexecute-progress-* (docs/.DS_Store was already untracked/ignored -
the plan's git rm --cached assumption was stale; no-op)
No code behavior changed (docstring only in verdicts.py). Suite 152/4 green, mypy clean,
ruff format --check clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AaQCFnfsh3tfq1VfzdJpoi
The expert reviewer was only a hardcoded verdict_input dict inside the offline
simulation. Build it as the real, shared artifact target picture §8 calls for:
shared/skills/expert-reviewer/ — a SKILL.md persona prompt (energy-advisor / M&V
role + the realization-gap methodology the validator cannot compute) plus a
canonical references/example-verdict.json. shared/ stays pure data; the MAF side
reads it via portfolio_optimiser.persona.load_persona_example (call-time,
fail-fast) and the Claude-SDK sibling reads the same JSON with its own loader.
This de-stubs the simulation: its persona judgement (decision + rationale + traced
marker) is now sourced from the artifact at call time, not an inline literal — so
the shared persona is genuinely consumed and cannot rot silently. decision is
binary (approved/rejected, the FeedbackContract the run path accepts);
approved_with_adjustment is rejected there and lives only in the bundle seed
frontmatter + the promotion gate, so the realization correction is carried in the
rationale prose.
Load-bearing trio (tests/test_persona_skill_loadbearing.py), each proven RED on its
own detach: structure + framework-neutrality, the example is valid pipeline input
(incl. FeedbackContract, on a throwaway copy), and the simulation's marker follows
the artifact file. Suite 149->152.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
The primary method proof, offline — a deliberate, cost-driven substitution
for målbilde §11.8's real-model run (the operator runs MAF against no real
model; API for both repos is too costly privately).
`portfolio_optimiser.simulation` drives `run_project` with a scripted
synthetic chat client across two runs separated by a promotion, and shows
the learning loop close end to end:
- ScriptedChatClient subclasses the LAYERED OpenAIChatCompletionClient (not
bare BaseChatClient — else the always-attached BudgetMiddleware no-ops),
constructs offline (loopback url + dummy key), role-keys proposer/checker
replies, and records every prompt into a shared sink.
- simulate_learning_loop: Run A (fresh wiki) -> validated, persona-approved
verdict carrying a realization marker absent from the bundle -> promote_verdict
into the OKF wiki -> seed_store_from_bundle re-reads it -> Run B's hypothesis
prompt carries the marker. An empty-wiki control on Run A proves causality.
- `python -m portfolio_optimiser.simulation` prints an honest trace.
Honesty (§1): this proves the plumbing, the deterministic spine, and that the
learning dataflow closes — NOT that a live LLM would produce the proposal or
verdict (scripted stand-ins). The genuine model-behaviour comparison lives on
the Claude-SDK side (a minimal API run); the scripted client is MAF-side
scaffolding, not part of the framework-neutral shared/ core.
Load-bearing: tests/test_simulation_loadbearing.py goes red when promotion is
detached (the marker never crosses into Run B). Suite 148->149.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Close the last agentic-loop seam (målbilde §3/§6/§7/§11 step 6): an
APPROVED verdict is promoted from the raw output layer into the context
layer (the OKF bundle) as a navigable `type: verdict` concept file, so
human/persona-approved knowledge reaches the next run's hypothesis.
- okf.py (pure stdlib, MAF-free): render_frontmatter / write_concept_file
/ link_in_index — the D7-portable OKF write counterpart of navigate.
- verdicts.py: promote_verdict + PromotionRefused gate (fail-closed; only
approved decisions enter the wiki, never raw agent output), provenance
stamp (who/experiment/when; timestamp a required kwarg), neutral index
label (signal reaches a prompt only via the gated ExpeL fold, never
bundle_context), _safe_filename_token (id sanitised for path/link).
- R4 = optional+gated: a public opt-in primitive, NOT wired into
run_project (mirrors write_verdict — the system reads, the gate promotes).
- Load-bearing trio (test_step8_promotion_loadbearing.py): gate refuses a
non-approved verdict, approved verdict is navigable, promoted signal
stays out of the read-context — all proven RED-on-detach. Suite 144->148.
Design hardened by an adversarial plan-critic (12 findings; the BLOCKER —
index-link leak into bundle_context via index_summary — closed by the
neutral label + a no-leak test). Honesty limits documented: promoted file
is minimal (signal as prose only), and the learning-key id means
same-candidate approvals share a filename (last-write-wins).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
The short loop captured the expert verdict inline into an in-memory store, so a
verdict arriving days/weeks later in a separate run could not influence any future
hypothesis (målbilde §5 row 7). Steg 7 adds the long timescale: run_project gains an
opt-in verdict_dir async inbox that load_verdicts_from_dir -> store.add MERGES into the
store BEFORE the Step-1 ExpeL fold, so a verdict dropped after an earlier run reaches a
separate, later run's hypothesis — fully resumable across runs separated in time.
- verdicts.py: verdict_to_dict / verdict_from_dict (id read verbatim, never re-minted),
write_verdict (public authoring primitive, NOT wired into run_project — system reads
the folder, expert/persona writes it, §3 role split), tolerant load_verdicts_from_dir
(missing/foreign/half-written files skipped, not raised — RAW layer per §10 R2),
VerdictStore.from_dir.
- run.py: verdict_dir kwarg; ingest-merge block after load_contracts (merge not replace
keeps run_portfolio's cross-project threading; store.add idempotent on content-hash id;
no change to the fold). CLI --bundle-dir/--verdict-dir thread the long loop to the
console entry. No auto-persist of the run's own captured verdict (outbox/Steg 8).
- Load-bearing PAIR (test_step7_async_loop_loadbearing.py): a verdict dropped after run A
must reach run B's prompt (run B uses a FRESH store -> the transfer is the file loop,
not in-memory carryover); empty-inbox control proves causality. Marker = a realization
value absent from the bundle (not the seed's 0.82). Proven RED on ingest detach.
Suite 138 -> 140 passed, 4 skipped; mypy + ruff check clean. Målbilde treated as frozen
(no §3/§5/§7 edit). Step 8 (gated wiki promotion) remains.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Closes gap #5 (maalbilde §5/§7): generate_via_llm's outer max_attempts loop
built the prompt ONCE and re-sent it identically — a BLIND retry. The validator's
per-attempt Rejection.reason was captured in `last` but never reached the next
prompt, so the proposer re-answered the same question with no knowledge of WHY
the prior candidate failed. Step 5 routes that reason into the next attempt.
- generate.py: _build_messages() gains prior_rejection; when set it appends a
revision block carrying ONLY the falsification reason verbatim (never the
rejected proposal JSON). None -> the byte-identical base prompt, so attempt 1
is unchanged. generate_via_llm() rebuilds messages inside the outer loop with
prior_rejection=`last` (None on attempt 1); _fetch_parsed() takes messages as
an explicit parameter. `last` is overwritten each round -> only the most-recent
falsification ("forrige"), never an accumulated history. Bound unchanged:
max_attempts + meter.tick_round (no new loop; §6 — "improve until good enough"
without a ceiling stays impossible).
- Scope honesty: the only per-attempt falsifier here is the validator. The
checker is a run-level, one-shot signal (run.py, before generation); seeding
generation with the checker critique is separately scoped and NOT done here.
The boundary is written into the generate_via_llm docstring + README + CLAUDE.
Load-bearing (maalbilde §7): tests/test_step5_refine_loadbearing.py is a PAIR —
the positive test keys the proposer's flip on the validator REASON PAYLOAD (the
rejected claim value, derived from validate_proposal(bad).reason so test and SUT
share one source of truth), and asserts the reason reached attempt 2's prompt
VERBATIM (the green-but-dead guard). It goes RED on detach (build messages once):
the flip token never arrives, so the outcome never flips AND the verbatim
assertion fails — proven double-red. The bounded control proves a never-fixed
proposer exhausts exactly max_attempts and returns a Rejection.
Adversarial Plan agent hardened the design pre-implementation (flip on payload
not wrapper/call-count; derive flip-key from the validator reason; drive through
generate_via_llm directly; docstring honesty). Suite 136->138 passed, 4 skipped;
mypy + ruff check clean. New test ruff-formatted; pre-existing ruff-format drift
(budget/verdicts/test_contracts) left untouched for a surgical diff.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Closes gap #3 (maalbilde §5): the GroupChat checker critiqued into the void —
output_from=[proposer] surfaced only the proposer, so an explicit checker
rejection was ignored and the deterministic validator was the sole gate. Two
falsifiers now act on the same candidate: the validator gates the NUMBERS
(blocking, unchanged), the checker gates the REASONING (maalbilde §2/§6).
- workflow.py: output_from=agents surfaces both participants; the checker
instruction ends with a VERDICT: APPROVE / VERDICT: REJECT - <reason> line.
- run.py: _authored_texts() reads author_name through out.messages (MAF 1.9.0
puts it there, not on the AgentResponse); _debate_text() now selects the
PROPOSER-authored output (fixes a latent texts[-1] regression that would feed
the checker's verdict to generation at even round counts); _checker_verdict()
parses the gate decision. An explicit REJECT overrides an otherwise-validated
outcome to a checker-sourced Rejection. Opt-in-reject (fail-open on a missing
marker). RunResult gains checker_verdict; provenance.validator_decision is
stamped from the validator outcome BEFORE the override, so it never conflates
the two falsifiers (provenance honesty).
Load-bearing (maalbilde §7): tests/test_checker_gate_loadbearing.py is a PAIR —
an explicit checker REJECT on a VALIDATOR-VALID proposal yields a Rejection whose
reason carries the checker's reason while validator_decision stays "validated";
the causality control (checker APPROVE, same proposer) validates normally. Proven
RED on BOTH detach points (revert output_from, or drop the override).
Suite 134->136 passed, 4 skipped; mypy + ruff check clean. Pre-existing
ruff-format drift (backends/budget/verdicts/test_contracts) left untouched for a
surgical diff.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Closes the honest Fase 2a limitation: docs_dir==bundle_dir let keyword
chunk-stuffing leak the verdict's realization rate ("0.82") into the debate /
generation prompt regardless of the ExpeL fold (it surfaced from both
verdict-led-fro.md AND golden.json). The realization signal now reaches the
hypothesis prompt ONLY via the gated ExpeL fold.
- okf.py: bundle_context() + Bundle.context_files render the navigated bundle
(index + frontmatter + cross-links) as the agent read-context, EXCLUDING
type: verdict (maalbilde §2/§4). Pure stdlib, still MAF-free.
- datasource.py: bundle_citations() derives first-class citations from the
navigated non-verdict files.
- run_project: on the bundle path context + citations + debate tools come from
navigation (tools=[]; navigation replaces query-time RAG); the road path keeps
chunk-stuffing unchanged.
Load-bearing (maalbilde §7): the marker is upgraded from the minted verdict id
to the realization signal itself. The empty-store control now asserts "0.82"
reaches NO prompt — RED against the pre-2b chunk-stuffing path, green after
navigation (TDD red->green). New okf-level test_bundle_context_excludes_verdict_layer
guards the seam directly.
Suite 133->134 passed, 4 skipped; mypy + ruff check clean. Reverted unrelated
ruff-format drift (backends/budget/verdicts/test_contracts) to keep the diff
surgical.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Closes maalbilde §5 gap #1 (the one missing "feedback-into-prompt" dataflow)
for the OKF-bundle path. Before, ExpeL was computed AFTER generation into a
discarded SessionContext, so a prior verdict could not influence any hypothesis
(context_providers=0).
- New okf.py: framework-neutral OKF bundle navigation (index + frontmatter +
cross-links), pure stdlib, no agent_framework/mcp (D7-portable), enforced by
test_okf_is_maf_free.
- verdicts.py: seed_store_from_bundle + bundle_candidate_features build the
ExpeL substrate + the pre-hypothesis query key from a bundle.
- run_project(bundle_dir=...): folds the candidate's prior verdicts into the
generation context BEFORE generate_via_llm; the road path is unchanged.
Load-bearing (maalbilde §7): test_step1_expel_loadbearing proves a prior verdict
reaches the hypothesis prompt and goes RED when the fold is detached (shown via
TDD red->green). The marker is the minted verdict id (content hash) because
docs_dir==bundle_dir lets keyword chunk-stuffing leak the realization rate;
clean layer separation is Fase 2b.
Suite 121->133 passed; mypy + ruff check clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
Persistent dev-fixture for energieffektivisering (energiledelse/M&V), valgt for
sin lærings-overflate: gapet mellom modellert besparelse (validatoren regner) og
faktisk realisert besparelse i drift (eksperten kjenner) — det ExpeL skal lære.
Ett kontorbygg, ett LED-retrofit-tiltak. OKF-bundle (index/project/hypothesis/
methodology/reference/verdict) bærer kontekst-laget; verdict-led-fro.md koder
realiseringsgraden (RR ≈ 0,82, forankret i National Grid SBS 2010) som ExpeL-frø.
Energi mappet inn i den EKSISTERENDE kost-IR-en uendret (affected = byggets totale
energikostnad, claimed = modellert besparelse ~10 % < 30 %-cap), så validatoren
kjører som-den-er — src/ urørt. golden.json fryser de seeded percentilene; testen
beviser at fixturen er konsumerbar (validerer, ikke Rejection), ikke bare til stede.
Domenetall verifisert mot primærkilder (EVO/IPMVP, DOE/NREL UMP, CPUC, fire
evalueringsstudier); norsk energipris mot SSB Q1 2026. README + shared/README
oppdatert (eksempel finnes, ikke lenger "planned"). Suite 121/4, ruff+mypy rene.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHR8iKxJRxDiDfNw8HZmWE
