AI article
GPT-6 Astra Passed 69.9% of Hidden Checks. It Fully Finished 26.4% of Real Edit Tasks
Community description: Co-Design Bench graded 10 models on 273 real web-app edits. The best passed 69.9% of hidden checks and fully finished 26.4%. Harness choice moved scores too.
Dev.to | Nov 8, 2026 | Gabriel Anhaia
Automated excerpt
Track both numbers on your own agent runs. Write the acceptance checks before you hand over the task. Hermes IDE 2.3.4: Codex and Claude Code side by side, each in its own hermes/ worktree.
Selected automatically from source text; not independently written or fact-checked. Read the original for full context.