Lava lamp v1 · Claude Code · run 96e988f7 · 09.11.2026 21:52 Tallinn

Claude Fable 5.2

Low effort

Works

7 of 8 checks passed. All 5 blocker checks passed.

What it built: 3 captures

Desktop capture of Claude Code (Fable 5.2)'s output on Lava lamp v1, low effort, 09.11.2026 21:52 (run 96e988f7)
Desktop · 297.5 KB · e2d2ea9e309d…
Phone capture of Claude Code (Fable 5.2)'s output on Lava lamp v1, low effort, 09.11.2026 21:52 (run 96e988f7)
Phone · 128.7 KB · ef4ee3b890a5…
Motion capture of Claude Code (Fable 5.2)'s output on Lava lamp v1, low effort, 09.11.2026 21:52 (run 96e988f7)
Motion · 297.1 KB · 4f083a25ca50…

Along the row

Claude Fable 5.2 at every effort.

  1. Claude Fable 5.2 at low effort: desktop capture of run 96e988f7 Low4m 43s

    Works

    This effort
  2. Claude Fable 5.2 at medium effort: desktop capture of run b79ff6d9 Medium7m 47s

    Works

    1 run
  3. Claude Fable 5.2 at high effort: desktop capture of run 45666ded High34m 55s

    Works

    1 run
  4. Claude Fable 5.2 at extra high effort: desktop capture of run e47eb2fe Extra high40m 03s

    Works

    1 run
  5. MaxNot run

Down the column

Other models at low effort.

Fastest median first · this run: 4m 43s

  1. Muse Ember at low effort: desktop capture of run 9819d391Muse Ember23.2 s

    Works

    12.2× faster than this run
  2. GPT-5.6 Luna at low effort: desktop capture of run 4b99cb20GPT-5.6 Luna29.7 s

    Works

    9.6× faster than this run
  3. GPT-6.1 Terra at low effort: desktop capture of run e58b2b0cGPT-6.1 Terra42.9 s

    Works

    6.6× faster than this run
  4. GPT-6 Luna at low effort: desktop capture of run 46ad720eGPT-6 Luna51.5 s

    Fails

    5.5× faster than this run
  5. GPT-5.6 Terra at low effort: desktop capture of run 0512c0d4GPT-5.6 Terra1m 38s

    Works

    2.9× faster than this run
  6. Claude Opus 5.5 at low effort: desktop capture of run dc76fa3bClaude Opus 5.51m 42s

    Works

    2.8× faster than this run
  7. GPT-5.6 Sol at low effort: desktop capture of run 184f5448GPT-5.6 Sol1m 46s

    Works

    2.7× faster than this run
  8. Claude Haiku 5.5 at low effort: desktop capture of run 69b2dfe6Claude Haiku 5.52m 36s

    Works

    1.8× faster than this run
  9. Claude Sonnet 5.5 at low effort: desktop capture of run b7f9c96bClaude Sonnet 5.52m 37s

    Works

    1.8× faster than this run
  10. GPT-6 Sol at low effort: desktop capture of run 0beb33c6GPT-6 Sol2m 39s

    Works

    1.8× faster than this run
  11. Claude Fable 5.1 at low effort: desktop capture of run 4bd42367Claude Fable 5.12m 50s

    Works

    1.7× faster than this run
  12. GPT-6.2 Sol at low effort: desktop capture of run 4a1042d5GPT-6.2 Sol2m 56s

    Works

    1.6× faster than this run
  13. GPT-6.1 Sol at low effort: desktop capture of run 80d2eb3dGPT-6.1 Sol3m 11s

    Works

    1.5× faster than this run
  14. GPT-6.2 Astra at low effort: desktop capture of run 58a9b80aGPT-6.2 Astra3m 54s

    Works

    1.2× faster than this run
  15. GPT-6 Astra at low effort: desktop capture of run e375a13dGPT-6 Astra4m 17s

    Works

    1.1× faster than this run

Hidden checks

7 of 8 checks passed.

  • Self-contained index.html is presentblocker
  • Visible page contentblocker
  • Visible autonomous animationblocker
  • Accessible on/off controlextra
  • 390 px mobile layout has no horizontal overflowextra
  • Local animation frame performanceextra
  • No browser errorsblocker
  • No attempted external or linked-resource requestsblocker

It works when every blocker passes. How it looks is reviewed blind: pending.

Run detailsagent config, what ran, tokens, source and revisions
  • Video (video): not published, videos are withheld. SHA-256 bb477130d03d9b361aa74be86cb3dc7b1821de857fda34d4b2f226c825dd73c4, 1127.4 KB.
Run ID
96e988f7-469d-4ea2-b142-b546adfd9095
Benchmark
lava-lamp v1
Experiment
Single run
Started
09.11.2026 21:52 Tallinn
Graded
revision 1 · 09.11.2026 21:58

Works and checks, in full

All 5 blocker checks passed.

The rule: a run works when every blocker check passes. If any blocker fails it does not work. If none fails but one has no result yet, it stays pending, never counted as a pass or a fail. Extra checks are reported but do not decide it. Works is not a judgement of looks: that is the visual review.

This is a legacy contract v1 run: the task has no public requirements, and these are the original lava-lamp checks, not a versioned hidden verifier. It is never compared with contract v2 runs.

Contract v1 checks for this run, with blocker and result
RequirementCheckBlockerResult
none (v1)Self-contained index.html is presentBlockerpass
none (v1)Visible page contentBlockerpass
none (v1)Visible autonomous animationBlockerpass
none (v1)Accessible on/off controlExtrapass
none (v1)390 px mobile layout has no horizontal overflowExtrafail
none (v1)Local animation frame performanceExtrapass
none (v1)No browser errorsBlockerpass
none (v1)No attempted external or linked-resource requestsBlockerpass

Verification completed; eligible for benchmark results.

Visual review

Pending.

Pending: the captures have not been reviewed blind yet. Checks decide whether it works; how it looks is decided only in that review.

What ran

Requested and detected.

Detected values come from the run’s own output, never from our settings.

Requested agent config beside what the run reported
FieldRequestedDetectedEvidence
Modelclaude-fable-5-2claude-fable-5-2harness result, 10 of 10 API requests
Effortlowlow10 of 10 API requests
Harnessnot recorded2.1.321 (Claude Code)Version as the CLI reported it; binary SHA-256 abc9feeeca80…
Deadline60 min5m 42s usedCompleted

Tool decisions

17 calls · 16 accepted · 1 rejected

Each tool call is allowed or blocked by the harness’s permission settings before it runs. Allowed is not the same as succeeded: an allowed call can still fail.

  • Bash8 calls · 7 accepted · 1 rejected · 8 failed
  • Edit4 calls · 4 accepted
  • Grep2 calls · 2 accepted · 1 failed
  • Write2 calls · 2 accepted
  • Glob1 call · 1 accepted

Telemetry

Time and usage.

Generation time
4m 43s
First content
1.0 s
Output speed
98.0 tokens/s, whole run including tools
API-equivalent cost
$1.77 (computed from the run’s token counts at list prices, not a charge)
Actual charge
not reported

Input tokens

349,156 tokens
  • Cache reads312,01389%
  • Cache writes37,12311%
  • Uncached input20<0.1%

Output tokens

27,769 tokens
  • Reasoning11,46141%
  • Other output16,30859%

Usage as the agent reported it at the end of the run. Cache reads are part of input; reasoning is part of output.

Transcript and source

Plain-text downloads.

The redacted transcript and the frozen output bundle, with LF line endings. Hashes are of the original files.

Redacted agent transcript: 43.1 KB, 7 redactions. Download transcript.txt

Source files

1 file · 20.1 KB
Files in the output bundle
FileSizeSHA-256Redactions
index.html20.1 KB7e7dc6007b2a1ab7fe71b6f825fb15945d67774c72bc4c6f47ecbd387d9d9ac30
Bundle SHA-256 7e7dc6007b2a1ab7fe71b6f825fb15945d67774c72bc4c6f47ecbd387d9d9ac3

Revision history

This page shows revision 1.

Grades are append-only: a regrade adds a revision and never rewrites an earlier one.

Grading revisions of this run
RevisionGradedWorks
1 (current)09.11.2026 21:58works

Redaction: automated scan run; checked by Jaanus on 09.11.2026 22:10. Redacted: 7 local-path.