Microsoft Employee: Harness is 60% of Performance, Models Just 40%
RihardJarc · x · 2026-08-31
A Microsoft AI employee shared insights on model comparisons and industry trends:
- Model Characteristics: Claude excels at complex multi-step reasoning with less rework; OpenAI suits single-step execution but is 20-30% more costly for complex tasks due to iterations.
- Performance: Claude Code's native pairing has a 20-30% higher first-pass acceptance rate than GitHub Copilot, though the gap is expected to close.
- Key Takeaway: Models are becoming commodities; 60% of performance depends on the "harness" (engineering framework), while only 40% relies on the model itself.
- Pricing Trend: The industry is expected to shift from token-based to outcome-based pricing for cost predictability.
- Talent Shift: Junior hiring is down. 65-70% of the dev lifecycle will be autonomous in 3-5 years, making professionals with judgment and agentic skills more valuable.
More from coding & agent
- Vibe-coded dashboard leaked franchise financials before anyone logged in — shauntrennery · 2026-10-16
- Fintech HR built a sourcing agent on 60,000 past applicants — 30 qualified candidates in a month — G0NZ0BOY · 2026-10-16
- Orchestrating a 100-image portrait of humanity for aliens with ChatGPT, Argon and p5.brush — dx8xb · 2026-10-16
- 3x DGX Spark cluster runs GLM-5.3-Flash locally behind a proxy, 1.75M tokens/day at 119W — Teknium · 2026-10-16
- Builder's one rule for trusted desktop agents: read freely, but sending, deleting, buying always ask first — No_Fan5310 · 2026-10-16
- GPT-built unifi-simulator makes MikroTik switches appear inside UniFi Network — Pytlicek · 2026-10-16