Early Tester: Fable 5.2 Built a 3,100-City Water Map in One Go, but Still Makes Silly Errors
alliekmiller · x · 2026-11-10
AI researcher alliekmiller shared her experience as an early tester of Fable 5.2, calling it the model she trusts more than any other she has tested:
- Built a water safety map of 3,100 US cities from FEMA, NOAA and USGS data, with two demo videos
- Created and ran a monetizable ADHD-themed YouTube channel end to end
- Built psychological profiles of her stalkers to check if they were the same person (output withheld for privacy)
- Built a 3D hot dog kiosk game as a to-do command center; filled out 31 personality tests on her behalf with longitudinal analysis
Her verdict: the model is "unbelievable at handling complexity" — "there is no such thing as a needle in a haystack anymore" — but it still makes weirdly obvious errors (colors, units, overlapping text) that secondary agents failed to catch. She feels she "could be more ambitious."
More from coding & agent
- 8 concrete game-feel parameters to hand Claude: hitstop, flash, shake, with frame counts — algo_diver · 2026-11-10
- MARS trains RL agents to search raw corpora via shell commands, no retriever needed — _reachsumit · 2026-11-10
- Claude Plus Blender MCP Codes a 931-Part, 24,260-Rivet Jet Model — gaganghotra_ · 2026-11-10
- ENGRAM paper bakes agent experience into LLM weights, deletes memory only after verified tests — _reachsumit · 2026-11-10
- Tuned Claude skills win 95% of blind tests, beating one-shot prompt-written skills — every · 2026-11-10
- Mollick's Fable 5.2 experiment: 25,000 agents autonomously build 450+ linked 3D worlds — emollick · 2026-11-10