Running an Opus-level coding agent locally at nearly 2x Claude Opus speed, for free
julianharris · x · 2026-09-11
Julian Harris's hands-on long read claims Aug 2026 marks the tipping point for local AI coding: he runs an Opus-level coding agent locally for free at nearly 2x Claude Opus speed.
Key details:
- Tensor parallelism via RDMA across two MacBooks (50% each) reaches 25 t/s, vs 15 t/s with SSD streaming on a single machine—likely optimizable much further
- Benchmarks show doubling context halves qwen-3.8-27b throughput, while qwen-3.8-flash-next is barely affected
- Motivation: Anthropic's erratic access limits, rate caps (5-hour windows), and censorship made the commercial relationship untenable
- He built Ceetrix, an MCP spec-governance server that keeps coding agents from "behaving like lazy, gaslighting interns"
Fully hand-written article with days of first-hand experimentation.
More from coding & agent
- AIE Code Summit 2026 opens applications: 800 seats, keynotes streamed free — trq212 · 2026-10-27
- Decade-long cloud architect: enterprise AI architecture is far more than bolting on an LLM — kashifmanzoor · 2026-10-27
- Analysis of thousands of coding sessions: 40-89% of files agents read are never used — rseroter · 2026-10-27
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Precompute golden answers and match at runtime for exact wording control — colinmcnamara · 2026-10-27
- Dev argues there's still alpha in working with your coding agent interactively — alex_frantic · 2026-10-27