Arc · 7 posts

The Local Coding Harness

One week, one question: can a 24GB laptop running only local models safely fix real code? The honest answer went from "it cannot produce a single valid patch" to "two local labs autonomously fix each other's code" in seven days, not by making the models smarter, but by repeatedly moving trust from model instruction-following to mechanical verification.

Reading order

  1. 1
    The Model That Couldn't Write a Patch

    local-llm-lab-notes

    A local model can diagnose a bug correctly and still fail to ship the fix. Four live runs, two models, two scenarios, three genuinely different ways of getting the diff syntax wrong.

  2. 2
    Falsify Your Favorite Hypothesis First

    local-llm-lab-notes

    Every plausible explanation for the malformed-diff pattern got tested against stored evidence instead of assumed, and every one of them was insufficient or flatly wrong, before the real cause finally cleared.

  3. 3
    Stop Teaching the Model diff Syntax

    local-llm-lab-notes

    After three eliminated hypotheses, the fix wasn't a better prompt. It was removing the requirement that the model author diff syntax at all, and the first live attempt afterward finally produced a clean-applying patch.

  4. 4
    The Bug Behind the Bug

    local-llm-lab-notes

    Fixing the diff-authoring problem let a patch survive for the first time in the workflow's history, and immediately tripped a second bug that had been invisible since day one.

  5. 5
    A Review Panel on One Model Is One Reviewer in Four Hats

    local-llm-lab-notes

    A hardware ceiling forced a four-persona review panel onto one resident model. Measuring it honestly, against both a bugs-seeded fixture and a genuinely clean one, found a real instruction-following ceiling the panel couldn't tighten past.

  6. 6
    Make the Reviewer Quote Its Evidence

    local-llm-lab-notes

    The same trust shift that fixed diff generation fixed the review panel too: a second-pass verifier that must quote an exact line from the diff, mechanically checked, not trusted on its word.

  7. 7
    Two Labs, No Cloud, One Fix

    local-llm-lab-notes

    A cross-repo path-conflation fix turned a harness that only worked on itself into one that could fix a genuinely foreign repository. The same day, a sibling lab's fourth attempt at a real bug closed, entirely through local models.