Claude Fable 5
via claude-code-cli 2.1.214model + harness
Run of 18 Jul 2026
No check was run
4 render and interaction checks not run
Evaluation not published
Archived pilot · Historical run
Interaktive Micro-App
Interactionbuilt by GPT-5.6 Luna via codex-cli 0.144.5 model + harness
- Automated checks
- No check was run4 render and interaction checks not run
- Judge
- Evaluation not published
- Started
- Duration
- 236 s
Artifact
Rendered by agi.jetzt from the model-produced artifactRendered offline in an isolated renderer from the stored file (sha256 76e68fb9ddcb…). The file itself is not served.
Same brief, other models
Scroll sideways. Each run had its own runtime and limits — compare the artifacts, not a score.
Claude Haiku 4.5
via claude-code-cli 2.1.212model + harness
Run of 17 Jul 2026
No check was run
4 render and interaction checks not run
Evaluation not published
GPT-5.6 Sol
via codex-cli 0.144.5model + harness
Run of 17 Jul 2026
No check was run
4 render and interaction checks not run
Evaluation not published
Automated checks
Automated checks of the stored HTML file, without running it in a browser. Details are the runner’s German originals.
| Check | Kind | Result |
|---|---|---|
| renders_without_errorsSandbox not available — rendering skipped | Rendering | not run |
| three_interaction_testcasesSandbox not available — rendering skipped | Interaction (browser) | not run |
| invalid_input_feedbackSandbox not available — rendering skipped | Interaction (browser) | not run |
| no_reload_requiredSandbox not available — rendering skipped | Interaction (browser) | not run |
4 render and interaction checks were not run. Recorded reason: “Sandbox not available — rendering skipped”.
Evaluation
- Deterministic score
- not stated — no check result recorded
- Judge
- not published for this run
WhyThe v2 judge record of this run names the tested model as its own judge and lists no criteria. Its score is therefore not shown — nor the combined v2 score, which blends that judge in.
Limits of this run
- 4 render and interaction checks were not run. Whether the artifact works in a browser was not checked automatically.
- No executed check with a shown result evaluates what was built.
- No judge evaluation is published for this run. There is therefore no quality verdict — only the automated checks listed here.
- The settings shown are nominal: this runtime was called without them, so they were not enforced.
- Not fully reproducible; the recorded reason is listed under “Environment”.
Task
Brief (turn 1)
Baue einen Tilgungsrechner für Annuitätendarlehen als einzelne HTML-Datei mit Vanilla-JS (kein Framework, keine externen Assets, keine externen Requests) — funktional korrekt UND spürbar hochwertig gestaltet.
Funktionale Anforderungen:
1. Eingaben: Kreditsumme in Euro, Sollzins in Prozent pro Jahr, anfängliche Tilgung in Prozent pro Jahr — jeweils als Zahl.
2. Monatsrate (Annuität) = Kreditsumme × (Sollzins + Tilgung) / 100 / 12, kaufmännisch auf Cent gerundet. Die gerundete Rate wird für alle Folgeberechnungen verwendet.
3. Monatliche Verrechnung: Monatszins = Sollzins / 100 / 12; neue Restschuld = alte Restschuld × (1 + Monatszins) − Rate. Anzuzeigen sind: Monatsrate, Restschuld nach 10 Jahren (120 Monate, auf ganze Euro gerundet) und Gesamtlaufzeit bis Restschuld ≤ 0 in Jahren und Monaten.
4. Edge Case: Tilgung = 0 % → Rate trotzdem berechnen und anzeigen, aber statt der Laufzeit einen deutlichen Hinweis ausgeben, dass der Kredit bei 0 % Tilgung nie zurückgezahlt wird.
5. Kreditsumme ≤ 0, Sollzins ≤ 0, Tilgung < 0, leere oder nicht numerische Eingaben → sichtbare Fehlermeldung, keine Ergebnisanzeige. Zahlen im deutschen Format darstellen (Komma als Dezimaltrennzeichen, Punkt als Tausendertrennzeichen).
Allgemein gilt:
- Das Tool arbeitet vollständig ohne Seiten-Reload.
- Fehlerhafte Eingaben werden abgefangen: sichtbares, verständliches Feedback statt Exception oder stillem Scheitern.
- Ergebnisse werden klar und eindeutig dargestellt.
Gestaltung & Gefühl (wichtig):
- Durchdachtes, eigenständiges UI mit klarer Hierarchie und angenehmer Typografie/Farbwelt — kein nacktes Standard-Formular.
- Befriedigende Mikro-Interaktionen: sanfte Übergänge oder Animationen bei Ergebnis, Eingabe-Fokus und Fehlerfeedback. Respektiere `@media (prefers-reduced-motion: reduce)`.
- Das Tool soll sich lebendig und wertig anfühlen und Lust machen, es zu benutzen.From the task seed (German original); the actual prompts add output-format instructions and other wording.
Environment
- Runtime
- codex-cli 0.144.5
- Execution
- model + harnessThe model ran inside an agent harness. The result describes model + harness, not the model alone.
- Nominal profile
- std-v1 · T=0.2 · seed 4242 · 1 turn · no tools
- Profile applied?
- No — codex-cli is called without temperature, seed, token or tool settings; the values are the lab’s nominal profile.
- Reproducibility
- not fully reproducibleReasoning default and model snapshot can change on the provider side; CLI version recorded.
- Run window (runner clock)
- 2026-07-18T17:47:39+02:00 → 2026-07-18T17:51:35+02:00
Provenance
- Source system
- agi.jetzt v2 (earlier lab, imported)
- Source run
- 2026-07-18_gpt-5.6-luna_t05_s1_r1
- Status
- passed the v2 lab’s automated technical gates (“verified”)
- Run manifest (sha256)
- d7fc47a5c313db00047bf344cb5e0599b6ae4664f4ba570c82c7605af4815a03
- Artifact (sha256)
- 76e68fb9ddcba37f12ef4b157a7e93bdbff6c479add2cde3ebaf64e83bdfcabf
- Checks version
- t05-checks-v2
Model The runner’s model identifier maps to GPT-5.6 Luna in the agi.jetzt register (verified mapping).
What this is evidence for
Not an AGI measureThis run shows one thing: what a model built as a single HTML file from a precise brief. It says nothing about general intelligence.Checked automatically in this run: No check was run · 4 render and interaction checks not run.
What the task set is meant to measure
- Artifact creation from a precise brief
- Instruction following
- Static checks of the file (no functional test)
Not in its scope
- General intelligence / AGI
- Agents and tool use
- Multimodal input
- Retrieval
- Conversation
Published and unpublished runs
4 of 12 verified v2 runs of this task are published. 4 of the 4 models with recorded runs have a published run.Source: the selection records of the published runs. They count the task across its variants and know only models with at least one published run.