Entity trail

MirrorCode

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-28 / sources-researcherEpoch and METR's MirrorCode Benchmark: Opus 4.7 Reimplemented a 61K-Line Program in 14 Hours for $251 — Work Estimated at 2–17 Human WeeksImport AI 466 details MirrorCode, a new Epoch/METR benchmark that asks models to re-implement real software from scratch using only CLI access to the original binary's input/output behavior. Across 25 target programs (22 released publicly) spanning six languages and 16k–61k lines, 17/25 hit perfect scores and 21/25 reached 99%+ accuracy, including Apple's pkl configuration language; the hardest holdouts were ruff, giac_subset, and mailauth. The builder takeaway is the cost curve, not the ceiling — gotree was reproduced for $100–400, which reframes 'rewrite this dependency' as a budget line rather than a quarter of engineering.
2026-05-02 / sources-researcherImport AI #453: MirrorCode Benchmark Shows Claude Opus 4.6 Can Autonomously Reimplement 16,000-Line CodebasesMETR and Epoch AI's MirrorCode benchmark tests whether AI models can autonomously reimplement complex software. Claude Opus 4.6 successfully reimplemented 'gotree', a 16,000-line bioinformatics toolkit — a task that would take a human engineer weeks. Jack Clark also covers Ryan Greenblatt doubling his probability estimate of full AI R&D automation by end of 2028 from 15% to 30%.
2026-04-20 / rss-researcherImport AI 454: Anthropic's Automated Alignment Researchers Recover 97% of Performance Gap at $22/Hour — Plus HiFloat4 and MirrorCodeJack Clark's Import AI 454 covers three significant developments: (1) Anthropic's nine parallel Claude Opus 4.6 agents outperformed human researchers on weak-to-strong supervision, recovering 97% of the performance gap vs. humans' 23%, at $18K total ($22/research-hour). (2) Huawei's HiFloat4 4-bit format beats Open Compute Project's MXFP4 with ~1.0% relative loss vs ~1.5%. (3) METR/Epoch's MirrorCode benchmark shows Opus 4.6 autonomously reimplemented a 16K-line Go bioinformatics toolkit — a task estimated at 2-17 weeks for a human. The automated alignment result is particularly significant: AI research acceleration is no longer theoretical.
2026-04-13 / rss-researcherImport AI 453: Breaking AI Agents, MirrorCode, and Ten Views on Gradual DisempowermentJack Clark's weekly newsletter covers three major threads: METR/Epoch's MirrorCode benchmark showing Opus 4.6 can autonomously reimplement a 16,000-line bioinformatics toolkit (estimated 2-17 weeks of human work), methods for breaking AI agent benchmarks, and a structured survey of ten expert perspectives on gradual AI disempowerment risks. MirrorCode suggests current AI can complete certain multi-week software engineering tasks given good requirements and test cases.

Graph relationships

BENCHMARKED AGAINST
Claude Opus 4.7
Opus 4.7 achieved near-perfect scores on 21 of 25 programs in the MirrorCode benchmark.
Source finding
BUILT BY
METR
MirrorCode is co-developed by METR.
Source finding
BUILT BY
Epoch
MirrorCode is a benchmark built by Epoch.
Source finding
BENCHMARKED AGAINST
Opus 4.6
Opus 4.6 was benchmarked on MirrorCode, reimplemeting 16000 lines of code
Source finding
BENCHMARKED AGAINST
Claude Opus 4.6
Claude Opus 4.6 successfully reimplemented 16,000-line codebase in MirrorCode benchmark.
Source finding
BENCHMARKED AGAINST
Claude Opus 4.7 -> MirrorCode
Opus 4.7 achieved near-perfect scores on 21 of 25 programs in the MirrorCode benchmark.
Source finding
BENCHMARKED AGAINST
Opus 4.6 -> MirrorCode
Opus 4.6 was benchmarked on MirrorCode, reimplemeting 16000 lines of code
Source finding
BENCHMARKED AGAINST
Claude Opus 4.6 -> MirrorCode
Claude Opus 4.6 successfully reimplemented 16,000-line codebase in MirrorCode benchmark.
Source finding

Source trail

Import AI 466 Import AI Import AI

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…