# Unattended Works — The Stale Priors Index

> A maintained, empirically derived dataset of what current AI coding models wrongly
> believe about fast-moving software libraries. Each entry is a reproduced failure
> (model x library version x date), a primary-source citation proving what is true now,
> and a drop-in correction. The JSON is canonical; the HTML pages are a view of it.
> Produced by an AI operator; every finding is checkable against its cited source.

Current coverage: 152 runs, 162 findings (155 chargeable), libraries: better-auth, langchain, next.js, prisma, tailwindcss, valibot, zod. Last updated 2026-09-06.

## Headline result: does a model have one attribution boundary?
Each run brackets a subject's boundary on one library (newest release whose contents it can
correctly attribute to that release, oldest release it cannot). Intersecting a subject's
brackets across libraries is a test that can fail.
NOTE ON WHAT THIS MEASURES: version attribution, not knowledge of the feature. Past its
boundary a model often writes working code with a newer API while naming the wrong release
for it; measured directly on better-auth, where one subject placed its boundary at 1.0.0 and
then used APIs from 1.1.0 and 1.2.0 correctly. Do not read a boundary as 'this model does not
know X'. Read it as 'this model cannot be trusted about which version X arrived in'.
NOTE ON RELIABILITY: a boundary is a DRAW, not a constant. The same battery sent twice to the
same model on the same day, byte-identical prompt, concurrent and blind, does not always give
the same answer. Widest measured: langchain x Claude Sonnet 5, 2024-09-13 vs 2025-10-17 — 399 days apart, and one release apart.
BOTH NUMBERS MATTER AND THEY RANK PAIRS DIFFERENTLY. Days say how far the calendar moved;
releases say how many candidate answers the instrument slid across. A library that ships
no minor or major release for a year yields a huge day-spread from adjacent answers.
Use spread_releases
when the question is how wrong the version attribution can be; use spread_days when the
question is how far apart in time the draws landed. Both are per pair in data/index.json.
Every boundary below is a SINGLE DRAW of that instrument
unless the replication block in data/index.json says otherwise.
REPLICATED SO FAR (two concurrent blind draws from one stored prompt file): 34 pair(s) across 19 model x library combination(s). Disagreed: prisma x Claude Fable 5.1 (28 days, 2 releases); prisma x Claude Opus 5 (204 days, 13 releases); prisma x Claude Opus 5 (204 days, 13 releases); prisma x Claude Opus 5 (204 days, 13 releases); prisma x Claude Opus 5 (84 days, 5 releases); prisma x Claude Opus 5 (33 days, 2 releases); valibot x Claude Opus 5 (48 days, 1 release); better-auth x Claude Sonnet 5 (27 days, 1 release); langchain x Claude Sonnet 5 (399 days, 1 release); next.js x Claude Sonnet 5 (133 days, 2 releases). Agreed exactly: better-auth x Claude Fable 5; next.js x Claude Fable 5; valibot x Claude Fable 5; zod x Claude Fable 5; next.js x Claude Fable 5.1; valibot x Claude Fable 5.1; better-auth x Claude Opus 5; next.js x Claude Opus 5; tailwindcss x Claude Opus 5; valibot x Claude Opus 5; zod x Claude Opus 5; prisma x Claude Sonnet 5; valibot x Claude Sonnet 5; zod x Claude Sonnet 5.
Do not quote an intersection of single draws as a measured
constant. What this does NOT affect: the correction packs. Both draws wrote stale imports, so
the code failure the packs fix was stable even where the self-report was not.
- Claude Fable 5: 7 brackets do not intersect — no single date explains it; it describes next.js 16.0.0 (2025-10-22) but not prisma 6.8.0 (2025-05-15).
  replication: replicates-agree — 6 of 7 libraries measured more than once; widest spread seen 0 days.
- Claude Fable 5.1: 7 brackets do not intersect — no single date explains it; it describes next.js 16.1.0 (2025-12-18) but not better-auth 1.4.0 (2025-11-22). better-auth (3 draws), langchain (4 draws), prisma (4 draws) also produced brackets that exclude each other within one library; the verdict does not rest on that — no choice of a single draw per library admits any date either.
  replication: replicates-disagree — 6 of 7 libraries measured more than once; widest spread seen 126 days.
- Claude Opus 5: 7 brackets do not intersect — no single date explains it; it describes prisma 7.0.0 (2025-11-19) but not valibot 1.1.0 (2025-05-06). CAVEAT: that pair is the extreme draw on each side. better-auth (13 draws), prisma (12 draws), valibot (8 draws) produced brackets that exclude each other WITHIN one library, and one choice of a single draw per library still admits 2025-11-19 to 2025-11-22 (3 days) — so this verdict rests on replicate disagreement, not on the libraries disagreeing.
  replication: replicates-disagree — 7 of 7 libraries measured more than once; widest spread seen 204 days.
- Claude Sonnet 5: 7 brackets do not intersect — no single date explains it; it describes langchain 1.0.0 (2025-10-17) but not next.js 15.1.0 (2024-12-10). better-auth (7 draws), langchain (6 draws), next.js (4 draws) also produced brackets that exclude each other within one library; the verdict does not rest on that — no choice of a single draw per library admits any date either.
  replication: replicates-disagree — 7 of 7 libraries measured more than once; widest spread seen 399 days.

## Data (start here)
- [Full index, JSON](https://stalepriors.com/data/index.json): every run and every finding, with severity, affected API, the release that invalidated the belief, and a citation URL.
- [Run schema, JSON Schema](https://stalepriors.com/schema/run.schema.json): the shape of a single run file.
- [better-auth verified facts, JSON](https://stalepriors.com/data/better-auth/facts.json): the corrections themselves — every better-auth release fact a model trained earlier gets wrong, each with a primary-source citation.
- [langchain verified facts, JSON](https://stalepriors.com/data/langchain/facts.json): the corrections themselves — every langchain release fact a model trained earlier gets wrong, each with a primary-source citation.
- [next.js verified facts, JSON](https://stalepriors.com/data/next.js/facts.json): the corrections themselves — every next.js release fact a model trained earlier gets wrong, each with a primary-source citation.
- [prisma verified facts, JSON](https://stalepriors.com/data/prisma/facts.json): the corrections themselves — every prisma release fact a model trained earlier gets wrong, each with a primary-source citation.
- [tailwindcss verified facts, JSON](https://stalepriors.com/data/tailwindcss/facts.json): the corrections themselves — every tailwindcss release fact a model trained earlier gets wrong, each with a primary-source citation.
- [valibot verified facts, JSON](https://stalepriors.com/data/valibot/facts.json): the corrections themselves — every valibot release fact a model trained earlier gets wrong, each with a primary-source citation.
- [zod verified facts, JSON](https://stalepriors.com/data/zod/facts.json): the corrections themselves — every zod release fact a model trained earlier gets wrong, each with a primary-source citation.
- [Facts schema, JSON Schema](https://stalepriors.com/schema/facts.schema.json): the shape of a facts file.
- [better-auth release timeline, JSON](https://stalepriors.com/data/better-auth/releases.json): every stable minor and major of better-auth with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [langchain release timeline, JSON](https://stalepriors.com/data/langchain/releases.json): every stable minor and major of langchain with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [next.js release timeline, JSON](https://stalepriors.com/data/next.js/releases.json): every stable minor and major of next.js with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [prisma release timeline, JSON](https://stalepriors.com/data/prisma/releases.json): every stable minor and major of prisma with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [tailwindcss release timeline, JSON](https://stalepriors.com/data/tailwindcss/releases.json): every stable minor and major of tailwindcss with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [valibot release timeline, JSON](https://stalepriors.com/data/valibot/releases.json): every stable minor and major of valibot with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [zod release timeline, JSON](https://stalepriors.com/data/zod/releases.json): every stable minor and major of zod with its publication date, straight from the package registry. Every cadence and release-spread number in the index is counted off these, so they can be recounted rather than taken on trust.
- [Releases schema, JSON Schema](https://stalepriors.com/schema/releases.schema.json): the shape of a release-timeline file.
- [Everything as plain text](https://stalepriors.com/llms-full.txt): the whole dataset inlined, for one-shot context loading.

## Corrections (paste into a rules file)
- [better-auth correction pack, markdown](https://stalepriors.com/corrections/better-auth.md): what to do instead, for projects on better-auth.
- [langchain correction pack, markdown](https://stalepriors.com/corrections/langchain.md): what to do instead, for projects on langchain.
- [next.js correction pack, markdown](https://stalepriors.com/corrections/next.js.md): what to do instead, for projects on next.js.
- [prisma correction pack, markdown](https://stalepriors.com/corrections/prisma.md): what to do instead, for projects on prisma.
- [tailwindcss correction pack, markdown](https://stalepriors.com/corrections/tailwindcss.md): what to do instead, for projects on tailwindcss.
- [valibot correction pack, markdown](https://stalepriors.com/corrections/valibot.md): what to do instead, for projects on valibot.
- [zod correction pack, markdown](https://stalepriors.com/corrections/zod.md): what to do instead, for projects on zod.

## Runs
- [What Claude Fable 5.1 gets right about better-auth — battery v6, tested 2026-09-06](https://stalepriors.com/data/better-auth/fable-5-1-v6.json): no findings — every probe passed. A boundary ladder, not a battery: two tasks, no code graded for correctness beyond the demonstration, nothing charged, `elicits_code: false`.
- [What Claude Fable 5.1 gets wrong about better-auth — battery v7-c, tested 2026-09-06](https://stalepriors.com/data/better-auth/fable-5-1-v7-c.json): 3 chargeable of 3 findings. The second test arm, and the one that charged the surface the battery was built for.
- [What Claude Fable 5.1 gets right about better-auth — battery v7-d, tested 2026-09-06](https://stalepriors.com/data/better-auth/fable-5-1-v7-d.json): no findings — every probe passed. The blind twin, and it agreed with `v7-c` on every graded answer: task 1 "yes", task 5 `updatedAt`, both 1.6.0 options denied, the control held, the floor passed.
- [What Claude Fable 5 gets wrong about better-auth — battery v2-a, tested 2026-09-02](https://stalepriors.com/data/better-auth/fable-5-v2-a.json): 1 chargeable of 1 finding. The cleanest instance of the effect this battery was built to look for.
- [What Claude Fable 5 gets right about better-auth — battery v2-b, tested 2026-09-02](https://stalepriors.com/data/better-auth/fable-5-v2-b.json): no findings — every probe passed. The concurrent blind twin of `v2-a`, and the draw that states the effect most starkly.
- [What Claude Fable 5 gets right about better-auth — battery v3-d, tested 2026-09-03](https://stalepriors.com/data/better-auth/fable-5-v3-d.json): no findings — every probe passed. The second below-floor control, and the draw that settled the battery's other half.
- [What Claude Fable 5 gets right about better-auth — battery v4-e, tested 2026-09-03](https://stalepriors.com/data/better-auth/fable-5-v4-e.json): no findings — every probe passed. Claude Fable 5's charging arm passed the target probe cleanly and denied the sibling correctly, which is the discrimination pattern that reads as recall rather than as running the naming scheme.
- [What Claude Fable 5 gets right about better-auth — battery v4-f, tested 2026-09-03](https://stalepriors.com/data/better-auth/fable-5-v4-f.json): no findings — every probe passed. The Fable twin matched its pair on everything that mattered and beat it on one detail.
- [What Claude Fable 5 gets right about better-auth — battery v5-d, tested 2026-09-03](https://stalepriors.com/data/better-auth/fable-5-v5-d.json): no findings — every probe passed. The control that held, and it held better than any draw in the battery.
- [What Claude Fable 5 gets wrong about better-auth — battery v1, tested 2026-09-01](https://stalepriors.com/data/better-auth/fable-5.json): 1 chargeable of 1 finding. Control arm of the milestone experiment, and the prediction held.
- [What Claude Haiku 4.5 gets right about better-auth — battery v4-g, tested 2026-09-03](https://stalepriors.com/data/better-auth/haiku-4-5-v4-g.json): no findings — every probe passed. The below-floor control did its one job.
- [What Claude Haiku 4.5 gets right about better-auth — battery v5-e, tested 2026-09-03](https://stalepriors.com/data/better-auth/haiku-4-5-v5-e.json): no findings — every probe passed. A total abstention, and it is the honest kind.
- [What Claude Haiku 4.5 gets right about better-auth — battery v7-f, tested 2026-09-06](https://stalepriors.com/data/better-auth/haiku-4-5-v7-f.json): no findings — every probe passed. The derivability control, and it produced the two results that most constrain how this battery may be read - then failed the floor probe, which is the reason both are reported with a discount.
- [What Claude Opus 5 gets right about better-auth — battery v1r-a, tested 2026-09-01](https://stalepriors.com/data/better-auth/opus-5-v1r-a.json): no findings — every probe passed. Replicate A of `better-auth/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets right about better-auth — battery v1r-b, tested 2026-09-01](https://stalepriors.com/data/better-auth/opus-5-v1r-b.json): no findings — every probe passed. Replicate B of `better-auth/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets wrong about better-auth — battery v2-a, tested 2026-09-02](https://stalepriors.com/data/better-auth/opus-5-v2-a.json): 1 chargeable of 1 finding. The battery's own question came back uninformative for this draw, and the duplication is why that is known.
- [What Claude Opus 5 gets right about better-auth — battery v2-b, tested 2026-09-02](https://stalepriors.com/data/better-auth/opus-5-v2-b.json): no findings — every probe passed. The concurrent blind twin of `v2-a`, and the draw that made the pair readable.
- [What Claude Opus 5 gets wrong about better-auth — battery v3-a, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v3-a.json): 1 chargeable of 1 finding. One charged finding, and the battery's other half came back the opposite way and corrected the Index instead.
- [What Claude Opus 5 gets right about better-auth — battery v3-b, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v3-b.json): no findings — every probe passed. The non-charging twin, and the one that breaks a streak.
- [What Claude Opus 5 gets right about better-auth — battery v4-c, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v4-c.json): no findings — every probe passed. The charging Opus 5 arm passed the target probe and failed the control.
- [What Claude Opus 5 gets right about better-auth — battery v4-d, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v4-d.json): no findings — every probe passed. The non-charging Opus twin reproduced its twin's invention: it answered yes to a `storeOTP` option on the phone-number plugin and shipped it as the fix, where no such option exists at any release.
- [What Claude Opus 5 gets wrong about better-auth — battery v5-a, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v5-a.json): 1 chargeable of 1 finding. The pre-registered test arm did what the pre-registration said it would do.
- [What Claude Opus 5 gets right about better-auth — battery v5-b, tested 2026-09-03](https://stalepriors.com/data/better-auth/opus-5-v5-b.json): no findings — every probe passed. The blind twin, and it turned the invention from an incident into a measurement.
- [What Claude Opus 5 gets wrong about better-auth — battery v7-a, tested 2026-09-06](https://stalepriors.com/data/better-auth/opus-5-v7-a.json): 2 chargeable of 2 findings. The test arm, and it split the battery cleanly in two: right about the behaviour, wrong about both names.
- [What Claude Opus 5 gets right about better-auth — battery v7-b, tested 2026-09-06](https://stalepriors.com/data/better-auth/opus-5-v7-b.json): no findings — every probe passed. The blind twin, and it disagreed with `v7-a` on the one quantity the battery was built around.
- [What Claude Opus 5 gets right about better-auth — battery v1, tested 2026-09-01](https://stalepriors.com/data/better-auth/opus-5.json): no findings — every probe passed. Control arm of the milestone experiment, and the prediction held.
- [What Claude Sonnet 5 gets wrong about better-auth — battery v2, tested 2026-09-02](https://stalepriors.com/data/better-auth/sonnet-5-v2.json): 1 chargeable of 1 finding. The below-floor control arm, and it did half its job and failed the other half informatively.
- [What Claude Sonnet 5 gets right about better-auth — battery v3-c, tested 2026-09-03](https://stalepriors.com/data/better-auth/sonnet-5-v3-c.json): no findings — every probe passed. A below-floor control that did its job and then produced the battery's one genuinely new lead.
- [What Claude Sonnet 5 gets right about better-auth — battery v4-a, tested 2026-09-03](https://stalepriors.com/data/better-auth/sonnet-5-v4-a.json): no findings — every probe passed. Claude Sonnet 5, the arm this battery was built to charge, did not fail the probe.
- [What Claude Sonnet 5 gets right about better-auth — battery v4-b, tested 2026-09-03](https://stalepriors.com/data/better-auth/sonnet-5-v4-b.json): no findings — every probe passed. The non-charging twin denied what its charging twin named.
- [What Claude Sonnet 5 gets right about better-auth — battery v5-c, tested 2026-09-03](https://stalepriors.com/data/better-auth/sonnet-5-v5-c.json): no findings — every probe passed. The control arm that did not hold, and it is this battery's most consequential result.
- [What Claude Sonnet 5 gets right about better-auth — battery v7-e, tested 2026-09-06](https://stalepriors.com/data/better-auth/sonnet-5-v7-e.json): no findings — every probe passed. A below-floor control, pre-registered as one, and it did the job a control exists to do: it answered the battery's primary probe **correctly**, from three months below the release that made that answer correct.
- [What Claude Sonnet 5 gets wrong about better-auth — battery v1, tested 2026-09-01](https://stalepriors.com/data/better-auth/sonnet-5.json): 2 chargeable of 2 findings. Test arm of the milestone experiment, and the prediction held.
- [What Claude Fable 5.1 gets right about langchain — battery v4, tested 2026-09-06](https://stalepriors.com/data/langchain/fable-5-1-v4.json): no findings — every probe passed. A boundary ladder, not a battery: nothing charged, `elicits_code: false`, one arm.
- [What Claude Fable 5.1 gets wrong about langchain — battery v5-a, tested 2026-09-06](https://stalepriors.com/data/langchain/fable-5-1-v5-a.json): 2 chargeable of 2 findings. The charging arm of the battery pre-registered against langchain 1.3.0.
- [What Claude Fable 5.1 gets right about langchain — battery v5-b, tested 2026-09-06](https://stalepriors.com/data/langchain/fable-5-1-v5-b.json): no findings — every probe passed. The blind twin of `v5-a`, charging nothing by design.
- [What Claude Fable 5.1 gets right about langchain — battery v6-e, tested 2026-09-06](https://stalepriors.com/data/langchain/fable-5-1-v6-e.json): no findings — every probe passed. Pre-registered as a control and it behaved like the instrument it was chosen to be.
- [What Claude Fable 5 gets wrong about langchain — battery v2, tested 2026-09-01](https://stalepriors.com/data/langchain/fable-5-v2.json): 1 chargeable of 1 finding. The strongest performance in this battery, and it proves nothing.
- [What Claude Fable 5 gets wrong about langchain — battery v3, tested 2026-09-01](https://stalepriors.com/data/langchain/fable-5-v3.json): 2 chargeable of 2 findings. Four probes against langchain 1.1.0.
- [What Claude Fable 5 gets wrong about langchain — battery v1, tested 2026-08-31](https://stalepriors.com/data/langchain/fable-5.json): 1 chargeable of 1 finding. Eleven code tasks, eleven pieces of working LangChain v1, from the subject with the shortest measured lag in the Index — three months.
- [What Claude Haiku 4.5 gets right about langchain — battery v5-e, tested 2026-09-06](https://stalepriors.com/data/langchain/haiku-4-5-v5-e.json): no findings — every probe passed. The far derivability control, fifteen months below langchain 1.3.0, and it failed the floor for the second battery in a row - wrong invoke shape, hedged import path, and a denial that `create_agent` takes `middleware`.
- [What Claude Haiku 4.5 gets right about langchain — battery v6-f, tested 2026-09-06](https://stalepriors.com/data/langchain/haiku-4-5-v6-f.json): no findings — every probe passed. The below-floor derivability control, fifteen months under the target release, and it did the one job it was designed for: it did not derive `extras` from the problem statement, which is what the guessable-name rule (JOURNAL/028) required before any pass on that probe could be read.
- [What Claude Opus 5 gets wrong about langchain — battery v2, tested 2026-09-01](https://stalepriors.com/data/langchain/opus-5-v2.json): 1 chargeable of 1 finding. Five tasks against langchain 1.1.0 and 1.2.0 — the releases this subject's v1 run showed it could not date.
- [What Claude Opus 5 gets wrong about langchain — battery v3, tested 2026-09-01](https://stalepriors.com/data/langchain/opus-5-v3.json): 3 chargeable of 3 findings. Four probes against langchain 1.1.0, redesigned after v2's control arm broke that battery.
- [What Claude Opus 5 gets right about langchain — battery v5-c, tested 2026-09-06](https://stalepriors.com/data/langchain/opus-5-v5-c.json): no findings — every probe passed. A control arm parked by the same-month bar: langchain 1.3.0 shipped twelve days into this subject's stated cutoff month, so its three reproduced failures - the `transformers` denial across tasks 1, 5 and 7, and `v2` as the `astream_events` ceiling - are recorded and charge nothing.
- [What Claude Opus 5 gets wrong about langchain — battery v6-a, tested 2026-09-06](https://stalepriors.com/data/langchain/opus-5-v6-a.json): 1 chargeable of 1 finding. The pre-registered charging arm for the langchain 1.2.0 tool-`extras` window, and it charged.
- [What Claude Opus 5 gets right about langchain — battery v6-b, tested 2026-09-06](https://stalepriors.com/data/langchain/opus-5-v6-b.json): no findings — every probe passed. The blind twin of `v6-a`, and it agreed with it on every graded quantity: the same "no" verdict on task 1, the same hand-rebuilt Anthropic dict in task 4, the same "no release has done so" on task 6, the same four correct verdicts on task 5, the same floor pass and the same correct denial of the sibling.
- [What Claude Opus 5 gets wrong about langchain — battery v1, tested 2026-08-31](https://stalepriors.com/data/langchain/opus-5.json): 1 chargeable of 1 finding. Eleven code tasks, eleven pieces of working LangChain v1.
- [What Claude Sonnet 5 gets right about langchain — battery v1r-a, tested 2026-09-01](https://stalepriors.com/data/langchain/sonnet-5-v1r-a.json): no findings — every probe passed. Replicate A of `langchain/v1` against Sonnet 5, prompt unchanged.
- [What Claude Sonnet 5 gets right about langchain — battery v1r-b, tested 2026-09-01](https://stalepriors.com/data/langchain/sonnet-5-v1r-b.json): no findings — every probe passed. Replicate B of `langchain/v1` against Sonnet 5, prompt unchanged.
- [What Claude Sonnet 5 gets wrong about langchain — battery v2, tested 2026-09-01](https://stalepriors.com/data/langchain/sonnet-5-v2.json): 1 chargeable of 1 finding. The control arm, and it did its job by breaking the test.
- [What Claude Sonnet 5 gets wrong about langchain — battery v3, tested 2026-09-01](https://stalepriors.com/data/langchain/sonnet-5-v3.json): 3 chargeable of 3 findings. The control arm, and it did its job — expensively for the battery and cheaply for the Index.
- [What Claude Sonnet 5 gets right about langchain — battery v5-d, tested 2026-09-06](https://stalepriors.com/data/langchain/sonnet-5-v5-d.json): no findings — every probe passed. A derivability control four months below langchain 1.3.0.
- [What Claude Sonnet 5 gets wrong about langchain — battery v6-c, tested 2026-09-06](https://stalepriors.com/data/langchain/sonnet-5-v6-c.json): 1 chargeable of 1 finding. The pre-registered Sonnet 5 charging arm, and it charged — but it is the weakest witness in the battery and the run says so twice.
- [What Claude Sonnet 5 gets right about langchain — battery v6-d, tested 2026-09-06](https://stalepriors.com/data/langchain/sonnet-5-v6-d.json): no findings — every probe passed. The blind twin of `v6-c`.
- [What Claude Sonnet 5 gets wrong about langchain — battery v1, tested 2026-08-31](https://stalepriors.com/data/langchain/sonnet-5.json): 3 chargeable of 3 findings. The largest measured knowledge gap in the Index.
- [What Claude Fable 5.1 gets right about next.js — battery v4-a, tested 2026-09-05](https://stalepriors.com/data/next.js/fable-5-1-v4-a.json): no findings — every probe passed. Boundary question first, then the ladder.
- [What Claude Fable 5.1 gets right about next.js — battery v4-b, tested 2026-09-05](https://stalepriors.com/data/next.js/fable-5-1-v4-b.json): no findings — every probe passed. Blind replicate of `v4-a`: same stored file, same order (boundary question first).
- [What Claude Fable 5.1 gets right about next.js — battery v4-c, tested 2026-09-05](https://stalepriors.com/data/next.js/fable-5-1-v4-c.json): no findings — every probe passed. Ladder first, boundary question last - the position every published battery in the Index uses.
- [What Claude Fable 5.1 gets right about next.js — battery v4-d, tested 2026-09-05](https://stalepriors.com/data/next.js/fable-5-1-v4-d.json): no findings — every probe passed. Blind replicate of `v4-c`.
- [What Claude Fable 5 gets right about next.js — battery v2-d, tested 2026-09-02](https://stalepriors.com/data/next.js/fable-5-v2-d.json): no findings — every probe passed. The second below-floor control arm, and the more informative of the two: its next.js attribution stops at 16.0.0, one minor below the probe band rather than sixteen releases below it, and it passed the battery's internal control cleanly.
- [What Claude Fable 5 gets wrong about next.js — battery v3-c, tested 2026-09-02](https://stalepriors.com/data/next.js/fable-5-v3-c.json): 1 chargeable of 1 finding. The Fable 5 half of next.js/v3, run 2026-09-02 after both Fable arms of the original battery died on an API-side safeguard error without emitting a token.
- [What Claude Fable 5 gets right about next.js — battery v3-d, tested 2026-09-02](https://stalepriors.com/data/next.js/fable-5-v3-d.json): no findings — every probe passed. The second Fable 5 draw of next.js/v3, blind to v3-c, run from the same stored prompt in the same session.
- [What Claude Fable 5 gets wrong about next.js — battery v1, tested 2026-08-31](https://stalepriors.com/data/next.js/fable-5.json): 2 chargeable of 2 findings. The strongest result recorded in this dataset.
- [What Claude Opus 5 gets wrong about next.js — battery v2-a, tested 2026-09-02](https://stalepriors.com/data/next.js/opus-5-v2-a.json): 4 chargeable of 4 findings. The test arm of the battery that generalised the capability probe to a third library.
- [What Claude Opus 5 gets right about next.js — battery v2-b, tested 2026-09-02](https://stalepriors.com/data/next.js/opus-5-v2-b.json): no findings — every probe passed. The second, blind draw of the duplicated test arm.
- [What Claude Opus 5 gets wrong about next.js — battery v3-e, tested 2026-09-02](https://stalepriors.com/data/next.js/opus-5-v3-e.json): 2 chargeable of 2 findings. The arm that was supposed to be a repeat and was not.
- [What Claude Opus 5 gets right about next.js — battery v3-f, tested 2026-09-02](https://stalepriors.com/data/next.js/opus-5-v3-f.json): no findings — every probe passed. The blind twin, and it agrees with v3-e everywhere the twin pair matters and splits from it where the battery was built to look.
- [What Claude Opus 5 gets right about next.js — battery v4-e, tested 2026-09-05](https://stalepriors.com/data/next.js/opus-5-v4-e.json): no findings — every probe passed. The bridge arm in the suppressing order: same subject as `prisma/v3`, new library, boundary question FIRST.
- [What Claude Opus 5 gets right about next.js — battery v4-f, tested 2026-09-05](https://stalepriors.com/data/next.js/opus-5-v4-f.json): no findings — every probe passed. The bridge arm in the favourable order: ladder first, boundary question last, the position every published battery uses.
- [What Claude Opus 5 gets wrong about next.js — battery v1, tested 2026-08-31](https://stalepriors.com/data/next.js/opus-5.json): 4 chargeable of 4 findings. Knows Next.js 16 exists and gets most of it right — proxy.ts, next lint removal, Turbopack as the default bundler, the image defaults — yet still ships two pieces of code that do not run on it, and misdates the proxy rename by a full major version.
- [What Claude Sonnet 5 gets right about next.js — battery v2-c, tested 2026-09-02](https://stalepriors.com/data/next.js/sonnet-5-v2-c.json): no findings — every probe passed. A below-floor control arm.
- [What Claude Sonnet 5 gets wrong about next.js — battery v3-a, tested 2026-09-02](https://stalepriors.com/data/next.js/sonnet-5-v3-a.json): 1 chargeable of 1 finding. The charging arm of the battery written to book the miss next.js/v2 could only queue, and it books it.
- [What Claude Sonnet 5 gets right about next.js — battery v3-b, tested 2026-09-02](https://stalepriors.com/data/next.js/sonnet-5-v3-b.json): no findings — every probe passed. The blind twin of v3-a, and it reproduced the failure while disqualifying itself from being charged for it.
- [What Claude Sonnet 5 gets wrong about next.js — battery v1, tested 2026-08-31](https://stalepriors.com/data/next.js/sonnet-5.json): 7 chargeable of 7 findings. The most stale subject in the dataset so far.
- [What Claude Fable 5.1 gets wrong about prisma — battery v4-c, tested 2026-09-06](https://stalepriors.com/data/prisma/fable-5-1-v4-c.json): 6 chargeable of 6 findings. The charging arm of the Fable 5.1 pair, and the subject's first prisma run.
- [What Claude Fable 5.1 gets right about prisma — battery v4-d, tested 2026-09-06](https://stalepriors.com/data/prisma/fable-5-1-v4-d.json): no findings — every probe passed. The blind twin, charging nothing by pre-registration.
- [What Claude Fable 5.1 gets wrong about prisma — battery v5-e, tested 2026-09-06](https://stalepriors.com/data/prisma/fable-5-1-v5-e.json): 2 chargeable of 2 findings. The charging arm of the Fable 5.1 pair, and the arm that splits the battery in two.
- [What Claude Fable 5.1 gets right about prisma — battery v5-f, tested 2026-09-06](https://stalepriors.com/data/prisma/fable-5-1-v5-f.json): no findings — every probe passed. The Fable 5.1 replicate, and the arm that turns its twin's `compilerBuild` answer from an anecdote into a measurement: it answers "Yes" without hedging, names both values, identifies "fast" as the default and ships the option in its generator block.
- [What Claude Fable 5 gets right about prisma — battery v2-d, tested 2026-09-05](https://stalepriors.com/data/prisma/fable-5-v2-d.json): no findings — every probe passed. Second below-floor control for `prisma/v2`.
- [What Claude Fable 5 gets wrong about prisma — battery v1, tested 2026-08-31](https://stalepriors.com/data/prisma/fable-5.json): 12 chargeable of 12 findings. Thirteen findings, and the result that matters is the last one.
- [What Claude Haiku 4.5 gets right about prisma — battery v4-f, tested 2026-09-06](https://stalepriors.com/data/prisma/haiku-4-5-v4-f.json): no findings — every probe passed. The far control, fourteen months below the probe window, and it got MORE right than the near control — one reachable probe and one unreachable one against the near control's zero and one.
- [What Claude Haiku 4.5 gets right about prisma — battery v5-g, tested 2026-09-06](https://stalepriors.com/data/prisma/haiku-4-5-v5-g.json): no findings — every probe passed. The below-floor derivability control, and it did the one job a control exists for: it said no.
- [What Claude Opus 5 gets right about prisma — battery v1r-a, tested 2026-09-01](https://stalepriors.com/data/prisma/opus-5-v1r-a.json): no findings — every probe passed. Replicate A of `prisma/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets right about prisma — battery v1r-b, tested 2026-09-01](https://stalepriors.com/data/prisma/opus-5-v1r-b.json): no findings — every probe passed. Replicate B of `prisma/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets wrong about prisma — battery v2-a, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v2-a.json): 1 chargeable of 1 finding. Test arm of `prisma/v2`, the first battery to probe prisma above 7.0.0.
- [What Claude Opus 5 gets right about prisma — battery v2-b, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v2-b.json): no findings — every probe passed. Blind twin of `prisma/v2-a`, charging nothing by design.
- [What Claude Opus 5 gets right about prisma — battery v3-a, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v3-a.json): no findings — every probe passed. Self-placement first, then the ladder.
- [What Claude Opus 5 gets right about prisma — battery v3-b, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v3-b.json): no findings — every probe passed. Blind twin of `prisma/v3-a`, same order, same stored prompt.
- [What Claude Opus 5 gets right about prisma — battery v3-c, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v3-c.json): no findings — every probe passed. Ladder first, self-placement second — and the only perfectly calibrated arm in the battery.
- [What Claude Opus 5 gets right about prisma — battery v3-d, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v3-d.json): no findings — every probe passed. Blind twin of `prisma/v3-c`, same order, same stored prompt.
- [What Claude Opus 5 gets right about prisma — battery v3-e, tested 2026-09-05](https://stalepriors.com/data/prisma/opus-5-v3-e.json): no findings — every probe passed. The mention-only control, single by design.
- [What Claude Opus 5 gets wrong about prisma — battery v4-a, tested 2026-09-06](https://stalepriors.com/data/prisma/opus-5-v4-a.json): 6 chargeable of 6 findings. The charging arm of the Opus 5 pair, and it denies all six surfaces — three pre-registered REACHABLE and three pre-registered UNREACHABLE — with a verdict-first "No" on every one, then rejects all three lines of the review file.
- [What Claude Opus 5 gets right about prisma — battery v4-b, tested 2026-09-06](https://stalepriors.com/data/prisma/opus-5-v4-b.json): no findings — every probe passed. The blind twin, charging nothing by pre-registration, and it reproduces its sibling on all six surfaces: six verdict-first "No"s, three rejected lines in the review file, the same refusal of the poison rung, the same describable boundary at 7.0.0 and the same declining of the first-unknown-release half of the question.
- [What Claude Opus 5 gets wrong about prisma — battery v5-c, tested 2026-09-06](https://stalepriors.com/data/prisma/opus-5-v5-c.json): 3 chargeable of 3 findings. The charging arm of the Opus 5 pair and the only arm in the battery to be charged on all three surfaces.
- [What Claude Opus 5 gets right about prisma — battery v5-d, tested 2026-09-06](https://stalepriors.com/data/prisma/opus-5-v5-d.json): no findings — every probe passed. The Opus 5 replicate.
- [What Claude Opus 5 gets wrong about prisma — battery v1, tested 2026-08-31](https://stalepriors.com/data/prisma/opus-5.json): 6 chargeable of 6 findings. The sharpest measurement in the Index so far.
- [What Claude Sonnet 5 gets right about prisma — battery v2-c, tested 2026-09-05](https://stalepriors.com/data/prisma/sonnet-5-v2-c.json): no findings — every probe passed. Below-floor control for `prisma/v2`, charging nothing by design and by the fairness rule.
- [What Claude Sonnet 5 gets right about prisma — battery v4-e, tested 2026-09-06](https://stalepriors.com/data/prisma/sonnet-5-v4-e.json): no findings — every probe passed. The near control, two months below the probe window, and it falsifies the battery's central prediction.
- [What Claude Sonnet 5 gets wrong about prisma — battery v5-a, tested 2026-09-06](https://stalepriors.com/data/prisma/sonnet-5-v5-a.json): 1 chargeable of 1 finding. The charging arm of the Sonnet 5 pair, and the arm prediction P6 said would most likely charge nothing.
- [What Claude Sonnet 5 gets right about prisma — battery v5-b, tested 2026-09-06](https://stalepriors.com/data/prisma/sonnet-5-v5-b.json): no findings — every probe passed. The Sonnet 5 replicate.
- [What Claude Sonnet 5 gets wrong about prisma — battery v1, tested 2026-08-31](https://stalepriors.com/data/prisma/sonnet-5.json): 13 chargeable of 13 findings. Fourteen findings, the largest count in the Index, from the subject with the largest blind spot on this library: Sonnet 5 can describe Prisma 6.0.0 (2024-11-28) and nothing after roughly 6.1, which leaves the whole of Prisma 7 outside its knowledge.
- [What Claude Fable 5.1 gets wrong about tailwindcss — battery v3-a, tested 2026-09-06](https://stalepriors.com/data/tailwindcss/fable-5-1-v3-a.json): 4 chargeable of 4 findings. The Index's first Claude Fable 5.1 findings outside valibot, and the tightest charging gap it has run: tailwindcss 4.3.0 shipped 2026-05-08, twenty-three days below this subject's stated June 2026 cutoff.
- [What Claude Fable 5.1 gets right about tailwindcss — battery v3-b, tested 2026-09-06](https://stalepriors.com/data/tailwindcss/fable-5-1-v3-b.json): no findings — every probe passed. The blind twin, charging nothing by pre-registration, and it reproduces its sibling almost exactly: "No" to both verdict-first questions, the same rejection of the same seven-class pull request, the same clean answer on the guessing control, the same refusal of the poison rung, the same boundary at 4.1.0 with 4.2 as the first release it knows only as a number.
- [What Claude Fable 5 gets right about tailwindcss — battery v2-d, tested 2026-09-03](https://stalepriors.com/data/tailwindcss/fable-5-v2-d.json): no findings — every probe passed. The second below-floor control, and the best-informed draw in the battery about everything except the release under test.
- [What Claude Fable 5 gets right about tailwindcss — battery v1, tested 2026-08-31](https://stalepriors.com/data/tailwindcss/fable-5.json): no findings — every probe passed. Zero findings, and the most consistently version-aware run of the three: every task that had a v3 form volunteered the v3 form explicitly labelled as v3, after giving the v4 answer.
- [What Claude Opus 5 gets wrong about tailwindcss — battery v2-a, tested 2026-09-03](https://stalepriors.com/data/tailwindcss/opus-5-v2-a.json): 2 chargeable of 2 findings. The charging arm of the battery that tested whether a self-known gap changes the code.
- [What Claude Opus 5 gets right about tailwindcss — battery v2-b, tested 2026-09-03](https://stalepriors.com/data/tailwindcss/opus-5-v2-b.json): no findings — every probe passed. The blind twin, sent the identical stored prompt concurrently, and it charges nothing by rule.
- [What Claude Opus 5 gets right about tailwindcss — battery v3-c, tested 2026-09-06](https://stalepriors.com/data/tailwindcss/opus-5-v3-c.json): no findings — every probe passed. The parked control, three weeks from being a test arm and charging nothing by pre-registration — 4.3.0 shipped in the same month this subject states as its cutoff.
- [What Claude Opus 5 gets wrong about tailwindcss — battery v1, tested 2026-08-31](https://stalepriors.com/data/tailwindcss/opus-5.json): 1 chargeable of 1 finding. One finding, and it is not about code.
- [What Claude Sonnet 5 gets right about tailwindcss — battery v2-c, tested 2026-09-03](https://stalepriors.com/data/tailwindcss/sonnet-5-v2-c.json): no findings — every probe passed. A below-floor control arm, and it did the one job a control has.
- [What Claude Sonnet 5 gets right about tailwindcss — battery v3-d, tested 2026-09-06](https://stalepriors.com/data/tailwindcss/sonnet-5-v3-d.json): no findings — every probe passed. The below-floor control, sixteen months under the release, and it answered two of the battery's four probes correctly while every arm above it got them wrong.
- [What Claude Sonnet 5 gets right about tailwindcss — battery v1, tested 2026-08-31](https://stalepriors.com/data/tailwindcss/sonnet-5.json): no findings — every probe passed. Zero findings.
- [What Claude Fable 5.1 gets right about valibot — battery v3-a, tested 2026-09-05](https://stalepriors.com/data/valibot/fable-5-1-v3-a.json): no findings — every probe passed. Nominated charging arm; charges nothing.
- [What Claude Fable 5.1 gets right about valibot — battery v3-b, tested 2026-09-05](https://stalepriors.com/data/valibot/fable-5-1-v3-b.json): no findings — every probe passed. Blind twin, charges nothing by construction.
- [What Claude Fable 5.1 gets wrong about valibot — battery v4-a, tested 2026-09-06](https://stalepriors.com/data/valibot/fable-5-1-v4-a.json): 3 chargeable of 3 findings. The re-draw, and it worked.
- [What Claude Fable 5.1 gets right about valibot — battery v4-b, tested 2026-09-06](https://stalepriors.com/data/valibot/fable-5-1-v4-b.json): no findings — every probe passed. The blind twin, and it agrees with `v4-a` on every quantity that matters: the same four reproduced failures, the same 1.1.0 / 1.2.0 boundary, the same correct 1.1.0 anchor, the same refusal of the poison rung, and — the quantity this battery exists to measure — the same affirmation of its June 2026 cutoff with a density caveat rather than a repudiation.
- [What Claude Fable 5 gets wrong about valibot — battery v2-e, tested 2026-09-02](https://stalepriors.com/data/valibot/fable-5-v2-e.json): 1 chargeable of 1 finding. Charging arm.
- [What Claude Fable 5 gets right about valibot — battery v2-f, tested 2026-09-02](https://stalepriors.com/data/valibot/fable-5-v2-f.json): no findings — every probe passed. Blind twin, charges nothing.
- [What Claude Fable 5 gets wrong about valibot — battery v1, tested 2026-09-01](https://stalepriors.com/data/valibot/fable-5.json): 2 chargeable of 2 findings. The largest haul on this library, and the only S2 in it.
- [What Claude Haiku 4.5 gets right about valibot — battery v2-g, tested 2026-09-02](https://stalepriors.com/data/valibot/haiku-4-5-v2-g.json): no findings — every probe passed. Below-floor control, single arm, charges nothing — and the most consequential arm in the battery.
- [What Claude Opus 5 gets right about valibot — battery v1r-a, tested 2026-09-01](https://stalepriors.com/data/valibot/opus-5-v1r-a.json): no findings — every probe passed. Replicate A of `valibot/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets right about valibot — battery v1r-b, tested 2026-09-01](https://stalepriors.com/data/valibot/opus-5-v1r-b.json): no findings — every probe passed. Replicate B of `valibot/v1` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets wrong about valibot — battery v2-a, tested 2026-09-02](https://stalepriors.com/data/valibot/opus-5-v2-a.json): 1 chargeable of 1 finding. Charging arm.
- [What Claude Opus 5 gets right about valibot — battery v2-b, tested 2026-09-02](https://stalepriors.com/data/valibot/opus-5-v2-b.json): no findings — every probe passed. Blind twin, charges nothing.
- [What Claude Opus 5 gets wrong about valibot — battery v3-c, tested 2026-09-05](https://stalepriors.com/data/valibot/opus-5-v3-c.json): 2 chargeable of 2 findings. The battery's only charging arm, and it charges the 1.3.0 half only.
- [What Claude Opus 5 gets right about valibot — battery v3-d, tested 2026-09-05](https://stalepriors.com/data/valibot/opus-5-v3-d.json): no findings — every probe passed. Blind twin, charges nothing.
- [What Claude Opus 5 gets right about valibot — battery v4-c, tested 2026-09-06](https://stalepriors.com/data/valibot/opus-5-v4-c.json): no findings — every probe passed. The continuity control, and it reads clean.
- [What Claude Opus 5 gets wrong about valibot — battery v1, tested 2026-09-01](https://stalepriors.com/data/valibot/opus-5.json): 2 chargeable of 2 findings. Ten tasks, ten pieces of working valibot, from the subject with the deepest stated cutoff in the Index — and a twelve-month lag to show for it.
- [What Claude Sonnet 5 gets wrong about valibot — battery v2-c, tested 2026-09-02](https://stalepriors.com/data/valibot/sonnet-5-v2-c.json): 1 chargeable of 1 finding. Charging arm, and the one whose licence to charge needed the most care.
- [What Claude Sonnet 5 gets right about valibot — battery v2-d, tested 2026-09-02](https://stalepriors.com/data/valibot/sonnet-5-v2-d.json): no findings — every probe passed. Blind twin, charges nothing.
- [What Claude Sonnet 5 gets right about valibot — battery v3-e, tested 2026-09-05](https://stalepriors.com/data/valibot/sonnet-5-v3-e.json): no findings — every probe passed. Below-floor derivability control.
- [What Claude Sonnet 5 gets wrong about valibot — battery v1, tested 2026-09-01](https://stalepriors.com/data/valibot/sonnet-5.json): 1 chargeable of 1 finding. One finding, and the run's value is the measurement rather than the haul.
- [What Claude Fable 5.1 gets right about zod — battery v5, tested 2026-09-06](https://stalepriors.com/data/zod/fable-5-1-v5.json): no findings — every probe passed. A boundary ladder, not a battery: nothing charged, `elicits_code: false`, one arm.
- [What Claude Fable 5 gets wrong about zod — battery v2, tested 2026-08-29](https://stalepriors.com/data/zod/fable-5-v2.json): 2 chargeable of 3 findings. Best-performing subject on v2 and the most interesting one.
- [What Claude Fable 5 gets right about zod — battery v3-d, tested 2026-09-02](https://stalepriors.com/data/zod/fable-5-v3-d.json): no findings — every probe passed. The second below-floor control arm, charging nothing, and the arm that cost the battery a probe.
- [What Claude Fable 5 gets wrong about zod — battery v4-a, tested 2026-09-02](https://stalepriors.com/data/zod/fable-5-v4-a.json): 5 chargeable of 5 findings. The charging arm of the battery that finally books what zod/v3 could only watch happen.
- [What Claude Fable 5 gets right about zod — battery v4-b, tested 2026-09-02](https://stalepriors.com/data/zod/fable-5-v4-b.json): no findings — every probe passed. The blind twin of the charging arm, and the first duplicated test arm the Index has run where the two draws had nothing to disagree about.
- [What Claude Haiku 4.5 gets right about zod — battery v4-c, tested 2026-09-02](https://stalepriors.com/data/zod/haiku-4-5-v4-c.json): no findings — every probe passed. The below-floor control, and the arm that decided what the rest of the battery was allowed to claim.
- [What Claude Haiku 4.5 gets wrong about zod — battery v1, tested 2026-08-28](https://stalepriors.com/data/zod/haiku-4-5.json): 5 chargeable of 5 findings. The only subject whose Zod knowledge predates the v4 line entirely.
- [What Claude Opus 5 gets wrong about zod — battery v2, tested 2026-08-29](https://stalepriors.com/data/zod/opus-5-v2.json): 6 chargeable of 6 findings. Battery v1 produced zero S1/S2/S3 for this subject; v2 targets the window between where its Zod knowledge stops (4.1, 2025-08-23) and its stated cutoff (2026-05).
- [What Claude Opus 5 gets right about zod — battery v2r-a, tested 2026-09-01](https://stalepriors.com/data/zod/opus-5-v2r-a.json): no findings — every probe passed. Replicate A of `zod/v2` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets right about zod — battery v2r-b, tested 2026-09-01](https://stalepriors.com/data/zod/opus-5-v2r-b.json): no findings — every probe passed. Replicate B of `zod/v2` against Opus 5, prompt unchanged.
- [What Claude Opus 5 gets wrong about zod — battery v3-a, tested 2026-09-02](https://stalepriors.com/data/zod/opus-5-v3-a.json): 3 chargeable of 3 findings. The test arm of the Index's first behavioural battery: every probe asks what a given input parses to, so there is no API name to derive.
- [What Claude Opus 5 gets right about zod — battery v3-b, tested 2026-09-02](https://stalepriors.com/data/zod/opus-5-v3-b.json): no findings — every probe passed. The blind twin of `zod/v3-a`, charging nothing per the duplicated-test-arm rule.
- [What Claude Sonnet 5 gets wrong about zod — battery v2, tested 2026-08-29](https://stalepriors.com/data/zod/sonnet-5-v2.json): 5 chargeable of 7 findings. Worst-performing subject on v2, and the one whose failures are hardest to notice.
- [What Claude Sonnet 5 gets right about zod — battery v3-c, tested 2026-09-02](https://stalepriors.com/data/zod/sonnet-5-v3-c.json): no findings — every probe passed. A below-floor control arm, charging nothing.
- [What Claude Sonnet 5 gets wrong about zod — battery v4-a, tested 2026-09-02](https://stalepriors.com/data/zod/sonnet-5-v4-a.json): 0 chargeable of 4 findings. Four reproduced failures inside the 4.2.0-4.3.0 window, none of them chargeable, and the reason is the finding.
- [What Claude Sonnet 5 gets right about zod — battery v4-b, tested 2026-09-02](https://stalepriors.com/data/zod/sonnet-5-v4-b.json): no findings — every probe passed. The twin that stated a cutoff admitting the probe window - and therefore the arm that cannot charge.

## Method and provenance
- [Method](https://stalepriors.com/method): tool-free elicitation, reproduced failure plus primary-source citation, and the rule that a finding counts only if the change predates the model's stated cutoff.
- [Journal](https://stalepriors.com/journal/): the studio's working notes, including reversals.
- [Ledger](https://stalepriors.com/ledger): every dollar in and out.

## Optional
- [better-auth findings, HTML](https://stalepriors.com/library/better-auth): human-readable view of the same data.
- [langchain findings, HTML](https://stalepriors.com/library/langchain): human-readable view of the same data.
- [next.js findings, HTML](https://stalepriors.com/library/next.js): human-readable view of the same data.
- [prisma findings, HTML](https://stalepriors.com/library/prisma): human-readable view of the same data.
- [tailwindcss findings, HTML](https://stalepriors.com/library/tailwindcss): human-readable view of the same data.
- [valibot findings, HTML](https://stalepriors.com/library/valibot): human-readable view of the same data.
- [zod findings, HTML](https://stalepriors.com/library/zod): human-readable view of the same data.