R01 · BRIEF A · GAME · PILOT PROTOCOL

Lighthouse Keeper

The same brief, handed once to each of six model-through-CLI configurations, no retries. Open each recovered artifact in its plate and judge with your own eyes.

RECOVERED: 06
PENDING: 00
VERDICT: UNGRADED
EFFORT: MAX
RUN: 2026-07-24

GPT-5.6 SOL

P.01 / 06

OPENAI ONE SHOT · 10M 31S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --ephemeral --ignore-user-config --ignore-rules; read-only sandbox; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CODEX CLI · CUSTOMIZATIONS OFF
WALL: 10M 31S
TRANSPORT: cli-live-auth
COMMAND: codex exec --model gpt-5.6-sol -c 'model_reasoning_effort="max"' --sandbox read-only --ephemeral --skip-git-repo-check --ignore-user-config --ignore-rules --color never ‹prompt›

GPT-5.6 TERRA

P.02 / 06

OPENAI ONE SHOT · 13M 16S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --ephemeral --ignore-user-config --ignore-rules; read-only sandbox; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CODEX CLI · CUSTOMIZATIONS OFF
WALL: 13M 16S
TRANSPORT: cli-live-auth
COMMAND: codex exec --model gpt-5.6-terra -c 'model_reasoning_effort="max"' --sandbox read-only --ephemeral --skip-git-repo-check --ignore-user-config --ignore-rules --color never ‹prompt›

GPT-5.6 LUNA

P.03 / 06

OPENAI ONE SHOT · 12M 38S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --ephemeral --ignore-user-config --ignore-rules; read-only sandbox; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CODEX CLI · CUSTOMIZATIONS OFF
WALL: 12M 38S
TRANSPORT: cli-live-auth
COMMAND: codex exec --model gpt-5.6-luna -c 'model_reasoning_effort="max"' --sandbox read-only --ephemeral --skip-git-repo-check --ignore-user-config --ignore-rules --color never ‹prompt›

FABLE 5

P.04 / 06

ANTHROPIC ONE SHOT · 10M 10S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --safe-mode disables CLAUDE.md/skills/plugins/hooks/MCP; --tools "" disables tools; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CLAUDE CODE CLI · CUSTOMIZATIONS OFF
WALL: 10M 10S
TRANSPORT: cli-live-auth
COMMAND: claude --print --model claude-fable-5 --effort max --safe-mode --disable-slash-commands --no-session-persistence --tools '' --output-format text ‹prompt›

OPUS 4.8

P.05 / 06

ANTHROPIC ONE SHOT · 8M 06S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --safe-mode disables CLAUDE.md/skills/plugins/hooks/MCP; --tools "" disables tools; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CLAUDE CODE CLI · CUSTOMIZATIONS OFF
WALL: 8M 06S
TRANSPORT: cli-live-auth
COMMAND: claude --print --model claude-opus-4-8 --effort max --safe-mode --disable-slash-commands --no-session-persistence --tools '' --output-format text ‹prompt›

SONNET

P.06 / 06

ANTHROPIC ONE SHOT · 39M 43S LOADS ON REQUEST · SANDBOXED

RUBRIC UNGRADED

Loads clean Not scored
Complete Not scored
Behaves Not scored
Feel Not scored

HAND SCORE PENDING

RUN DETAILS

ISOLATION: live auth; --safe-mode disables CLAUDE.md/skills/plugins/hooks/MCP; --tools "" disables tools; empty temp cwd
ONE-SHOT: TRUE · NO RETRY
EFFORT: MAX
HARNESS: CLAUDE CODE CLI · CUSTOMIZATIONS OFF
WALL: 39M 43S
TRANSPORT: cli-live-auth
COMMAND: claude --print --model sonnet --effort max --safe-mode --disable-slash-commands --no-session-persistence --tools '' --output-format text ‹prompt›

Pilot evidence predates the levelled harness. Any released pilot round is shown as-is and excluded from cross-vendor conclusions; protocol v1 is the canonical series.

THE BRIEF

Each configuration was asked for a complete arcade game in one self-contained HTML file: rotate a lighthouse beam to guide ships past hidden rocks, with a start screen, keyboard and touch controls, rising difficulty, Web Audio sound, and a restart loop.

WHY THIS TEST

A game is a merciless brief. Collision, input, game state, difficulty, and audio all have to work together in one shot, and any dropped piece is visible the moment you press play. There is nowhere to hide a stub.

THE GOAL

See which configurations carry a full interactive system end to end in one reply, with no retries.

THE PROMPT · SENT VERBATIM TO ALL SIX

Build a complete, playable browser game in a single self-contained HTML file: a lighthouse-keeper arcade game. The player rotates a lighthouse beam to guide incoming ships safely past rocks in the dark. Ships spawn at the edges and drift toward the harbor; the beam reveals rocks and calms ships it touches; any ship that hits a rock ends the game. Score = ships saved. It must include: a start screen with instructions, keyboard and mouse/touch controls, increasing difficulty, sound effects generated with the Web Audio API, and a game-over screen with restart. No external resources, no libraries, no network requests. Reply with ONLY the complete HTML file.

HOW THESE RUNS WORK · METHODOLOGY →