Harness engineering gap: Provider harness 39.0% vs Mini-SWE 12.8% on same model
andrew_n_carr · x · 2026-11-20
Andrew Carr argues "harness engineering as a service" is a multi-million-dollar-a-year business opportunity: companies want to save tokens, boost accuracy, and gain repeatability. An attached chart shows the same model scoring 39.0% with a Provider harness vs just 12.8% with the Mini-SWE harness. His take: harness engineering will matter more than prompting ever did, and for longer.
More from coding & agent
- OpenAI hiring for computer-use agents team as Apple lists similar ambition — yungcontent · 2026-11-20
- Bug Hunt chart: score vs subscription cost, Gemini 4 Neon xhigh tops at 63 for $1.80 — PawelHuryn · 2026-11-20
- Measured: Claude Max delivers 46.4x API value on Opus 5.5, beating ChatGPT Pro's 10.25x — PawelHuryn · 2026-11-20
- 10 open-source GitHub repos that turn your docs into queryable knowledge graphs — victor_explore · 2026-11-20
- What happens when two AI agents disagree? The multi-agent conflict problem — EricJose59 · 2026-11-20
- AI turns Mediterranean illustration into interactive 3D world; Three.js wall-crash demo costs $6.98 — Saboo_Shubham_ · 2026-11-20