Development

A benchmark tests rebuilding whole projects from scratch and passing every test

development2026-08-04

Claim

On MirrorCode, which tests whether agents can rebuild whole software projects from scratch and pass every test, Claude Fable 5 solves 64% to GPT-5.6 Sol's 20%.

Reported metric: 64% vs 20%

Reporting date: 2026-08-04 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

Welcome to August 4, 2026 (theinnermostloop.substack.com)

Knowledge graphopen in browser →

Reported in

Welcome to August 4, 2026

Actor

Evidence for

Supersedes

A model reaches a credible score at a hundredth of the frontier's price

Referenced by

Welcome to August 4, 2026

Cite or correct this record

Citation formatCitation text

A benchmark tests rebuilding whole projects from scratch and passing every test
Innermost Loop KB — secondary record
Reporting date: 2026-08-04
KB permalink: https://nicholsn.github.io/innermost-loop-kb/developments/2026-08-04-rebuilding-whole-projects-from-scratch
Claim: On MirrorCode, which tests whether agents can rebuild whole software projects from scratch and pass every test, Claude Fable 5 solves 64% to GPT-5.6 Sol's 20%.
Attributed sources:
- Welcome to August 4, 2026 — https://theinnermostloop.substack.com/p/welcome-to-august-4-2026
This citation does not imply independent verification of linked-source claims.

Suggest a correction — opens a GitHub draft for you to review and submit.