Periodic Harness vs. vanilla Claude Code: a striking diff comparison
zainhas · x · 2026-09-16
The author shares a screenshot comparing the diff quality of Periodic Labs' harness against vanilla Claude Code, calling the gap striking. Periodic Harness is the lab's in-house agent execution framework, which the team claims beats off-the-shelf coding harnesses like Claude Code and Codex for scientific analysis workloads. Full benchmark details are in their FrontierXRD evaluation writeup.
More from coding & agent
- 10 agent skills with 3.49M downloads form a full dev workflow, led by mattpocock/skills — Roger_M_Taylor · 2026-09-16
- Perplexity built a DynamoDB replacement in 2 months with 2 engineers and hundreds of AI agents — inductionheads · 2026-09-16
- He makes $8,000/month building Claude skills for businesses at ~$1,200/hour — Roger_M_Taylor · 2026-09-16
- Celesto open-sources disposable full macOS desktops for AI agents on Apple Silicon — aniketmaurya · 2026-09-16
- Dev builds TrendsMCP: one API aggregating 40+ trend sources after pytrends collapse — Proper_Design_2616 · 2026-09-16
- AI connector value lies in secrets store and personal context, not payments — jeff_weinstein · 2026-09-16