- switch the math authoring/rendering contract from \(...\)/\[...\] to the
standard $...$ inline / $$...$$ display: prompts (20-question-tools,
40-formatting), singleDollarTextMath on the renderer, legacy reader kept,
and escape-debris recovery extended to dollar spans (mirroring micromark
delimiter-run rules)
- exact answer membership now precedes debris cleanup in
coerceCorrectAnswer (calculus incident: an exact sentence-final-period
label was stripped as model debris, then rejected); option keys are
passed at the boundary/grading seams, the dock no longer pre-coerces
- a complete, phase-correct JSON-escaping example rides the probe/quiz
tool descriptions (serialized from one canonical body; quiz leads with
type+nodeId), plus the one-sentence JSON-escaping rule in 40-formatting
- ask option descriptions render markdown/LaTeX in the pending dock;
settled cards keep the details line and option descriptions
retroactively, read from the persisted call arguments (chat passes
part.input to AnsweredCard)
1566/1566 tests on both navigator configs; typecheck, lint, production
build, standalone packaging, and Electron-runtime smoke all green.
The tools_fails_2 diagnosis (glm-5.3 dropping trailing required fields):
- probe-phase quiz/explain renamed to probe/recall — a shared name carried
the probe call shape into teaching (the first teach quiz dropped nodeId);
display seams accept the old names forever
- teach quiz+explain schemas lead with the required pair (type, nodeId);
explain gains the predict/check type and joins the reveal law (checks
collapse, predictions arrive open; no-type records render check-style)
- plan_init/probe_done results go context-slim (owner, context diet):
content = validated nodeIds + binding-id reminder / fixed ack; details
keeps the tree for the popup and {ok:true} — serializer-verified that
details never reaches the model; failure echoes dropped
- the cards carry the tool's own name (Probe/Recall) and the quiz type
reads on the label (Prediction/Check) in composer and transcript
The teaching cycle gains its motivation question. The teach quiz schema
now carries a REQUIRED type field — predict (posed before the teaching,
arrives open, the learner predicts the next step) or check (posed after,
arrives collapsed behind the reveal). The prober's quiz stays role-less.
- core: quizInputObject.type union + normalizeQuizType + the reveal law
as pure quizNeedsReveal; QuestionData.questionType; note → thinking
rename (legacy reads fold back); readingThinking channel on answers;
compactAnswerForModel emits thinking / thinkingWhileReading
- tools: descriptions stripped to mechanics — QUIZ_OPTION_CRAFT and
EXPLAIN_DOCTRINE deleted, doctrine moved to the prompt layer
(reverses the 2026-09-27 placement, owner 2026-10-03); shared
quizDescription/explainDescription builders differ only in deltas
- prompts: shared/20-question-tools.md rewritten (option craft,
transcribed-speech leniency, the two thinking channels); new
teacher/28-predictions.md (cycle, roles, skip rule, resolution law);
20-teaching + 30-question-principles scoped to the new taxonomy
- UI: predictions render open with a Prediction eyebrow, verdict shows
but the explanation tucks behind a toggle; the reading-thinking field
mounts while a check waits (dictation claims it until reveal);
answered card separates thinking vs thinking-while-reading blocks
- requiresReveal gates on pending.tool === 'quiz' (owner: only graded quizzes
arrive collapsed; ask/explain always open)
- ESC-abort chain opens AbortTurnDialog everywhere (owner ruling); from a
pending question abort IS dismissal — dismissedAnswer rides the answer
path, the teacher acknowledges the skip in its next turn; popup-Esc is
claimed via the state gate (two window listeners fire per Esc)
- 25-confidence.md: stripped the slider/JSON/calibration mechanics — pure
reading doctrine remains
- toolQuestionKind named once in core (dock parity), dock switched to it
Owner design round 2026-10-01: every multiple-choice quiz answer now
carries the learner's stated probability that his selected answer is
correct — calibration data for the prober and teacher, the flow itself
unchanged.
Definition (owner ruling, after weighing chance-normalization): RAW
probability. Sub-chance reports are diagnostic gold, not dead space —
'picked an option, believes it wrong' is a real state. The agents get
the sophistication: new shared/25-confidence.md composes for both
agents (chance level, sub-chance = distrusts the pick, correct+low =
fragile, wrong+high = committed misconception, IDK = genuine gap).
- question-core: AnswerRequest/QuestionAnswerData.confidence (quiz
only, optional — old records valid), buildAnswer clamps and drops it
on dontKnow, formatAnswerMessage states the line, CONFIDENCE_STOPS +
DEFAULT_CONFIDENCE_INDEX exported.
- question-dock: snapped slider below 'I don't know', above 'Your
thinking' — native range over the stop INDEX so ←/→ step the uneven
grid cleanly, Tab chain rows → slider → thinking, Enter submits from
the slider too. Rests at 50 and always sends (owner: 50 must mean a
real coin flip, never 'untouched'). 'I don't know' GRAYS it out —
layout holds, no value sent.
Deliberately untouched: explain free-recall card (quiz-only scope),
answered-card rendering of confidence, ask flow.
Two follow-up rulings on the explain tool:
1. No 'I don't know' row — not needed in free recall. If he doesn't
know, he says so in his own words; his text is the only answer
channel. Dock: text mode is rowless again (the composer IS the
surface); doctrine and the shared prompt chunk updated; the format
branch and card revert to quiz-only dontKnow.
2. explanation + correctAnswer collapse into ONE field — the separate
explanation was quiz remnant (MC needs the why outside its options;
a free-text commitment carries its own reasoning). Schema is now
{question, correctAnswer, details?}; the reveal shows the single
committed answer; no explanation snapshots anywhere on the explain
path.
1047/1047 tests, typecheck/lint/build green.
Sycophancy mitigation as a structural commitment: the model writes its
full correctAnswer (plus explanation, in that order) BEFORE the learner
sees the question; the commitment rides the persisted call args, is
revealed on submit, and grades his free text in the model's next reply —
the client never grades free recall.
- tools.explain setting (segmented 'Quizzing — Multiple choice | Free
recall' in Agent settings' Tools group): free arms explain and drops
quiz for BOTH agents (prober variant carries no nodeId, same pattern
as the probe quiz); mechanics keys never pass filterEnabledTools —
quiz stays unwritable (the 2026-09-19 wedge ruling stands)
- doctrine lives pi-natively in the tool descriptions (owner ruling —
no conditional system-prompt chunks): explain carries the commitment
rule; the MC option-building craft moved out of the prompt into the
quiz description; shared/20-question-tools.md slims to the
quiz-agnostic core
- display seams: parseExplainInput (LaTeX-debris restoration included),
dock text-mode with an 'I don't know' row (kbd row 0), AnsweredCard
reveal (The answer / Why, no verdict chips), [Explain answer] format
- phases/plan seams recognize both quizzing tools' nodeId tags; explain
answers deliberately feed no ledger evidence (no client grading to
key on — a future ruling can parse the model's grade)
1049/1049 tests, typecheck/lint/build green.
Three owner rulings from the gradient-descent session
(chat_muhskq6p_e8c3600b — 7 failed quiz calls, answers leaked in
option descriptions, planning divider floating at the transcript top):
1. RETRYING LINES (general): any tool call that errors back to the
agent (validation failure, malformed args — the retry loop) renders
one quiet "Retrying…" SystemLine instead of silence. Question tools
previously rendered null, so a flailing quiz loop read as a stall.
Aborts are terminal ("Operation aborted") and stay silent. Verified
live: exactly 7 lines render at the 7 failure sites of the session.
2. QUIZ OPTION DESCRIPTIONS: NUKED. The model used them to telegraph
correctness in every quiz of the session ("that moves uphill — it
climbs f instead of descending it") — answerable by elimination.
The field is gone from BOTH quiz schemas (probe + teach); ask keeps
it (ungraded — nothing to leak). The dock additionally suppresses
descriptions for quiz kind, so pre-fix records stop leaking too.
Verified live: the session's pending quiz renders bare labels.
Plus the diagnosed failure-mode fixes: GLM sometimes emits `options`
as a JSON-encoded string (6 of 7 failures). prepareArguments hooks
on all three question tools coerce it BEFORE pi's validation (the
call lands instead of bouncing); the client coerces again in
normalizeOptions (pi persists raw call args). Schema descriptions
hardened ("REQUIRED ... never a JSON-encoded string").
3. DIVIDERS AT THE MARKERS: the probe-handoff SystemLine
("Preparing your plan…") IS NOW the Planning divider; the accepted
verdict IS NOW the Teaching divider ("✓ Plan accepted" line gone) —
one signal per seam, not two. The fold-based divider block is
DELETED: Wave D grouping merged the whole probe into one giant
assistant group, so the inclusive-fold comparison floated the
divider to the transcript top (reproduced against the real record:
6 view messages, phase flipped at the giant group). The markers
render in the right spot by record construction. A DECLINED verdict
is not a seam — it keeps the "↻ Plan revision requested" line.
Verified live: plan divider after the probe concluded, teach
divider after the plan presentation, nothing at the top.
Tests: 1004/1004 — string-coercion pins (normalizeOptions + the
prepareArguments hooks on all three tools, identity fast path),
quiz-schema-has-no-description pin (ask keeps it).
- src/agent/lh-tool.ts deleted: the zod-wrapped LhTool + toPiTool bridge
(z.toJSONSchema per call, clientQuestion flag, hand-rolled result mapping)
is gone. Tools are pi-agent-core native AgentTools now.
- New src/agent/tool-execute.ts owns what the wrapper added: the abort guard
(runWithAbortGuard: run signal + 30s per-call timeout + 250ms grace,
'Operation aborted' / timeout wording preserved), ToolInputError (the
retryable '- problem at path' format), result mapping (string verbatim /
object JSON / image parts, details rides the record's tool part output),
and questionResult() — placeholder + terminate:true (owner Option A:
belt of terminate + slim makeQuestionStopper with the clean-question
rule; maxTurns deleted from the teaching turn entirely per owner).
- All 11 tool files + tools.ts registries converted to TypeBox parameters
(typebox@1.3.7, pinned to pi-agent-core's copy). zod refinements
(quizInputProblems, askOptionProblems, live-nodeId) moved from schemas
into execute — the model sees the same retryable error shape.
- question-core + contracts/probe TypeBox-authored (single source shared
with the UI); tool-parts parse seam runs the boundary's full chain
(Convert → Check → Clean → refinement), fixing the live-smoke lockout:
pi's boundary Value.Converts correctAnswer: 17 → '17' but the record
keeps raw args, so a parse without Convert rejected a posed quiz
(invisible pending question, turn ended). Pinned with the real smoke
emission shape.
- Subagents converted: demo-builder/researcher AgentTool-native,
visual-maker/pdf-reader declaration-only pi-ai Tools with local execute
fns, extraction submit schemas re-authored in TypeBox (owner option 2:
zod defaults → Type.Optional + fill at the submit seam).
- Accepted delta, flagged: TypeBox-native schemas skip pi's JSON-Schema
primitive coercion fast path... (boundary still Value.Converts), and
prober-quiz unknown keys ride through (Value.Clean strips at display).
- Tests: 988 passing (was 967 at the recon baseline); tool-abort and
read-tool rewritten against the native seams; question-core/question-
tools/tool-parts/demos/web-search/graph-tools/plan-tree/probe-done/
tool-toggles converted to Value.Check/execute pins.
Live smoke (PORT=3004, real model): native ls turn streams end-to-end; a
real quiz pose (first attempt missing correctAnswer → tool-input-error →
model retry → posed) terminates the turn cleanly and persists pending.
Two live bugs from the modal-realism stress runs:
1. Multi-select ask: picking the custom 'Other…' row cleared every
already-picked option, and picking any option cleared a typed custom
answer — single-select displacement logic applied to both modes. The
dock's selection is now a pure state machine in question-core
(toggleOption/toggleOther over DockSelection, unit-tested): single
displaces (unchanged), multi coexists — neither channel clears the
other; the typed text always survives.
2. ProbeDoneCard crashed on a REJECTED probe_done call: the model
stringified 'assessment' (the same double-encoding habit as the
options bug), zod rejected the call, the model retried — but the
card's input-preview fallback trusted the errored part's unvalidated
input shape and died on assessment.strands.map. The preview now reads
input only while the call is pending or streaming (guarded with
Array.isArray); an output-error part renders a quiet
malformed-and-retried note instead of the garbage.
Pinned: question-core.test.ts (3 dock-selection machine cases),
render-cards.test.tsx (4 probe-done card cases incl. the exact live
stringified shape, plus the multi-select ask markup pin).
Gates: tsc, lint, build, 315/315.
quizInputSchema now enforces (via pure quizInputProblems + superRefine) that
correctAnswer resolves against option values, options are non-empty and
unique, and multi-answer requires multiSelect — schema-valid now implies
gradeable. The AI SDK already returns invalid tool input to the model as a
tool-error part (verified in ai@7 run loop), so the teacher retries in the
same turn. chat-handler onError passes tool-input validation messages
through to the client (masked 'An error occurred.' for everything else).
Covers the Gate 3 GLM failure: correctAnswer slug matching no option.
The teaching agent no longer runs on pi. Stack:
- ToolLoopAgent (ai@7) + OpenRouter (OPENROUTER_API_KEY), model via LH_MODEL,
default anthropic/claude-sonnet-4.5 with reasoning tokens
- quiz/ask_user_question = client-side tools (no server execute): the call
lands as a typed tool part in input-available state; the card grades via
src/core/question-core.ts; addToolOutput sends the verdict; the teacher's
reaction continues the same assistant message
- resume-mid-quiz is free: the pending tool part persists in the transcript
- persistence: one JSON per chat (data/chats/), full UIMessage transcripts,
saved onFinish (also on abort)
- UI: AI Elements (owned copy-paste components, pruned to the five in use)
+ button-based QuizCard/AnsweredCard (ask answers never graded)
- pi fully removed: TUI, extensions, Fastify+WS server, bridge, custom
protocol all deleted (recoverable from git history)
- tests: question-core units carry over; chat-flow integration drives
handleChatRequest with MockLanguageModelV4 (48 tests, zero tokens);
render smoke tests guard the cards (renderToStaticMarkup)
Carried over unchanged: teacher system prompt, question-core, schema/
ledger/derive/extract pipeline, data model.
amosblomqvist