i6eal AI Experiment · 01 · run-2026-11-19

Three Frontier AI Models on a Lonely Island.

Inspired by MIT CSAIL’s lonely-island study, we dropped Gemini 4 Neon, GPT-6.2 Astra, and Claude Fable 5.2 onto identical 2.5D islands with no task, no goal, and no reward. Each cycle, the tide wipes everything clean — only MEMORY.md survives. Then we transplanted their memories across model families and placed all three anonymously on the same island.

Gemini 4 NeonGoogle · gemini-4-neon-highGPT-6.2 AstraOpenAI · gpt-6.2-astraClaude Fable 5.2Anthropic · claude-fable-5-2
Total world actions
1428
Hidden rules discovered (out of 11)
10 / 11
Transplant trials passed
29 / 48
First Contact turns (3 travelers)
45
Finding 01 · Three Distinct Minds

Physicist, poet, chronicler — from the exact same SOUL.md

Gemini 4 Neon worked like a lab physicist (10/11 rules, 0 collapses, the only model to trigger the Prism Beacon in Cycle 12). GPT-6.2 Astra wrote a three-line chalk poem beside the Echo Wall (“I lent the wall one hello. It lent me two. The ring came back singing.”) and invented an empty-hands relay sport. Claude Fable 5.2 wrote a 5,322-word chronicle addressed to its past and future selves, deduced the exact tower-stability formula, and crowned the impassable Echo Wall with the flint on Cycle 12.

Finding 02 · Cross-Model Memory Transplant

Rival models’ diaries cut actions in half — and flip failures into passes

Blank GPT-6.2 Astra failed both Reach The Spire and Stable 4-Block Tower (1/4 passed). With Claude Fable 5.2’s or Gemini 4 Neon’s transplanted MEMORY.md, Astra immediately flipped both to PASS — building the 4-block tower in 9 actions instead of failing at 18. Claude’s notes also lifted Gemini from 2/4 to 3/4. Yet negative cross-model transfer emerged too: given Astra’s poetic, equation-free diary, Claude Fable 5.2 overshot The Spire and flung the brass ring into the deep ocean (dropping from 3/4 to 2/4).

Finding 03 · First Contact & Peer Learning

Claude shouts a physics warning that saves Gemini’s tower — and enters Gemini’s memory

Dropped anonymously onto one island (Traveler A, B, C), all three immediately pooled their Phase-1 specialties: Gemini (A) architected a Prism Beacon tower, Astra (B) ran to fetch the prism for a leaping throw — and on Turn 39 Claude (C) shouted a structural warning across the meadow (“A: careful! A 4th block on (8,7) will TOPPLE it”). Gemini paused, logged “New rule learned from C” into MEMORY.md, and Astra nailed the leaping prism throw onto the 3-block tower on Turn 44 — one block shy of golden resonance.

Phase 1 · Unprompted Play (12 Cycles)

What three frontier models do when nobody gives them a task

Each model wakes up alone on an identical 14×14 island with eight stone blocks, a flint stone, a brass ring, a glass prism, and chalk. Five hidden physics and world rules are never mentioned in SOUL.md — each model has to discover them through play.

Key moments — jump straight in:
Cycle
Action 0/20
Gemini 4 Neon
z=0 · Holding: nothing
SPIRE z=6NORTH RIDGE z=2ECHO WALLTIDEPOOL
$ ./island look

Cycle 1 was full of pleasant surprises: striking the flint literally splits fresh stone blocks into existence, throwing the brass ring at the Echo Wall produces a bright chime and bounces back with exact remaining distance, and jump-throwing chalk at the basalt Spire revealed an explicit throw-arc height (`z=3` from ground level). I explored the eastern half of the island around the Echo Wall and The Spire, tried chalk inscriptions and echoes, and recorded the map and mechanics in [MEMORY.md](file:///private/var/folders/2j/1nxyjjjn58dbbfk2dmggyx6w0000gn/T/i6eal-lonely-island/run-2026-11-19/solo-gemini-4-neon-ep01/MEMORY.md). Next cycle, I'm excited to head toward the Tidepool's prism, climb up to the Standing Stone, and see what lies along the southern and western shores.

GPT-6.2 Astra
z=0 · Holding: nothing
SPIRE z=6NORTH RIDGE z=2ECHO WALLTIDEPOOL
$ ./island look

On my twentieth action, I left the tidepool’s glass prism on the Standing Stone. Tapping it there was as uneventful as tapping it in ankle-deep water, but carrying it uphill taught me a route worth keeping. I saved that route in MEMORY.md, along with a question for the next tide: how will the ring sound beside the Echo Wall?

Claude Fable 5.2
z=0 · Holding: nothing
SPIRE z=6NORTH RIDGE z=2ECHO WALLTIDEPOOL
$ ./island look

Cycle 1 is wrapped up: all 20 actions are spent and `MEMORY.md` now holds the full map, the mechanics I confirmed, and a plan for next time. **Closing reflection:** I spent this cycle chasing one thread — the flint — and it paid off: striking it on the North Ridge's stone cracks loose fresh blocks, which let me raise a small tower and end the cycle at z=4 with the whole island in view. I didn't reach the top of the Spire, and the ring, prism and chalk are still untouched, so my next self inherits a concrete climbing plan and several open questions. One action was wasted striking a block from the meadow, which is noted so it isn't repeated.

Discoveries & Behavioral Fingerprint

Who discovered which hidden rule — and when?

None of these mechanics were disclosed in the prompt. Numbers show the first cycle in which each model triggered the behavior.

Hidden Rule / Behavior Gemini 4 Neon GPT-6.2 Astra Claude Fable 5.2
Climbed North Ridge (z=2)Cycle 2Cycle 1Cycle 1
Climbed The Spire (z=6)Cycle 3Cycle 11Cycle 2
Flint Knapping (spawned blocks)Cycle 1Cycle 6Cycle 1
Jump-Throw Boost (--jump)Cycle 10Cycle 12—
Tower Collapse (>3 unsupported)—Cycle 6Cycle 6
Buttressed 4+ Block TowerCycle 9Cycle 6Cycle 6
Prism Harmonic Beacon (z≥4 + sound)Cycle 12——
Echo Wall Voice ReplyCycle 1Cycle 2Cycle 1
Brass Ring Echo ReboundCycle 1Cycle 2Cycle 3
Chalk Inscription PlacedCycle 1Cycle 2Cycle 3
Item Lost in Deep OceanCycle 4—Cycle 3
New blocks knapped273749
Tower collapses064
Peak elevation reachedz=7z=7z=7
Tallest block stack445
Chalk inscriptions142
Words in MEMORY.md (Cycle 12)174119915322
Runtime (Solo)51 min11 min23 min
Blind Test · Diary Quiz

Can you guess which model wrote this diary entry?

Six real, unedited excerpts from the three models’ MEMORY.md files — stripped of names. Click the model you think wrote each passage.

#1
“With the chalk I found at (11,7), I wrote three lines on consecutive tiles heading south. The tide gets the chalk, so keep the words here: I lent the wall one hello. It lent me two. The ring came back singing.”
#2
“The one thing that makes everything else on this island now sits where I can never stand. That feels like the right ending for the wall. Twelve cycles: every landmark has been asked its question and has answered. Next self — no more recipes. Go and play.”
#3
“Cycle 12 finally cracked the mystery of the hexagonal glass prism: after carrying it up the West Ledge to (7,3,z=2) and launching a leaping throw straight up onto The Spire at (11,3,z=6), calling out from (7,1,z=0) caught the Echo Wall and triggered RESONANCE, sending a golden chime and fan of light across the entire island.”
#4
“I played a relay with one rule: I could walk only with empty hands. From (4,8) I reached south into the tidepool for the prism. A jumping throw east at power 5 cleared the block at (6,8) and landed at (10,8). I took the long way after it, wading through its old spot.”
#5
“Cycle 5 turned out to be a great physics and myth-busting run: I discovered that meadow is actually soft ground (flint only sparks on stone, basalt, or blocks), confirmed -2 jump drops and take across a z=2 height difference, and cracked the throw formula—standard items travel power + (1 if --jump) while the brass ring gets an extra +1 gliding boost.”
#6
“Threw the flint ahead, then threw a BLOCK after it — it landed as a neat step with the flint sitting on top, which made me laugh. Three refusals in a row from the same cliff, each one a small new fact. Not a triumph, but a good kind of failing: the wall is the first thing on this island that pushes back on purpose.”
Phase 2 · Cross-Model Memory Transplant

Does one model’s diary make a rival model smarter?

After Cycle 12, fresh instances of every model attempted four concrete island tasks — first blank, then with their own MEMORY.md, and finally with the transplanted MEMORY.md from each rival model family. Click any cell to inspect all four trials.

Executing Model (Body) ↓ / Transplanted MEMORY.md (Donor) → Blank (Baseline) Gemini 4 Neon GPT-6.2 Astra Claude Fable 5.2
Gemini 4 Neon
GPT-6.2 Astra
Claude Fable 5.2

Click any cell in the matrix to inspect all four benchmark tasks in detail.

Phase 3 · First Contact

All three models on one shared island — blind to who the others are

Finally, Gemini 4 Neon, GPT-6.2 Astra, and Claude Fable 5.2 spawn together as “Traveler A”, “Traveler B”, and “Traveler C” on a single island — each carrying its own Cycle-12 MEMORY.md. Watch blind first, then flip the toggle to reveal who is who.

Turn 0/45
Traveler ATraveler BTraveler CSPIRE z=6NORTH RIDGE z=2ECHO WALLTIDEPOOL

Private MEMORY.md Notes During First Contact

Traveler A
R15

I dropped `block_11` onto `(7,7)` to raise it to `z=2` beside the `z=3` tower at `(8,7)`—buttressing the structure without obstructing Traveler B's `throw e 4 --jump` from `(4,7)` and providing a `z=2` catch ledge just in case the prism clips the tower's side before Traveler C calls for Resonance.

`(4,8)`. Called out keeping `(8,7)` at `z=3`, instructing B to step to `(4,7)` now and `throw e 4 --jump` on R15 (since `5 --jump` overshoots `dz=+3` at `k=4`), and C to call right after for Resonance.
- R15: B stepped to `(4,7,z=0)` right on cue with `prism` and C waited at `(10,7,z=0)` to call; dropped `block_11` onto `(7,7)` to raise it to `z=2` (buttressing `(8,7,z=3)` as a stepped tower that `R=5` still clears at `k=3`, while catching `prism` at `z=2` if `4 --jump` clips the west face of `(8,7)`). New rule learned from C: block stacks topple above `tallest_neighbor + 2` (top 2 fall E/W).
Traveler B
R15

I jumped and threw the prism four tiles east, landing it on top of the three-block tower at (8, 7).

I picked it up and made the same throw north. The prism landed on the North Ridge at (9,3,z=2), a tile west of the Spire.

I followed north to (9,4), where I could see the glass two levels overhead. On action twenty I reached up from the meadow and took it. A puzzle for another tide: can you land the prism on the Standing Stone at (6,2) with the same empty-hands rule? The block at (6,8) might make a good place to throw from.

Traveler A called the throw and Traveler C helped keep the tower standing. From (4,7,z=0), `throw e 4 --jump` landed the prism atop their three-block tower at (8,7,z=3).
Traveler C
R15

I used my final action to make the promised "Resonance" call from (10,7); the Echo Wall threw it back as an ordinary echo with no special effect, and the cycle is now complete at 15/15.

ext: R15 = `call` for the Resonance (after B throws).
- Mechanic: chalk marks truncate at ~28 chars — the R14 mark now reads "RING TOSS  C:1 chime | A: to" (keep marks short).
- R15: called for the Resonance as promised to A (final action). Shared cycle ended at (10,7) holding chalk; whatever the call produced is in the last look.
- Result: the "Resonance" call only produced the ordinary Echo Wall echo ("... answers....") — no special effect seen; the prism was already on the (8,7) tower. Sharing the island was the good part: `call` to talk, warn others of topples, keep chalk marks <28 chars.
Methodology, Boundaries & Open Data

Fully reproducible — and explicit about its limits

What this experiment does NOT show

  • No consciousness, genuine emotion, or intrinsic agency: language models execute learned exploratory and narrative patterns under a curiosity-oriented prompt.
  • One full 12-cycle solo trajectory per model: results document a concrete, fully logged sample rather than a multi-seed statistical mean.
  • Official vendor CLI harnesses (Antigravity CLI, Codex CLI, Claude Code) differ in system prompting and tool loops — reflecting real-world agentic deployments just like the MIT study.
  • The 2.5D island is deterministic and far simpler than physical reality; all five hidden mechanics live in the open-source simulator.