# Weight-space programming as a linear program — 8x fewer edits than GPT-6 Astra

> A minute of linear programming reproduces a weight edit that took an AI model eight hours of trial and error.

- Org: Nathan Barry
- Date: 2026-11-09
- Category: article, algorithm
- Source: https://nathan.rs/posts/weight-space-lp/
- Canonical page: https://ai-tldr.dev/releases/nathan-barry-weight-space-lp/

## In short

Nathan Barry shows that GPT-6 Astra's weight edit to Qwen3.5-0.8B can be found as a linear program. Two one-minute LP solves match Astra's scores while changing 16,291 weights instead of 130,188.

## What it is

Nathan Barry's post revisits MIT CSAIL's 'Sparks of Weight-Space Programming', in which GPT-6 Astra edited Qwen3.5-0.8B's weights to stop the model's thinking loops. Barry shows that choosing those weight changes is a linear program, and publishes the code and edits.

## How it works

Given an edit direction, an LP solver picks a sparse set of weight changes that meets the target. Two solves of about a minute each matched Astra on GSM8K (69.6 vs 68.2), MATH-500 (51.7 vs 50.9) and IFEval (59.7 vs 61.9). With a simple heuristic direction and no LLM in the loop, the LP reached about 93% of Astra's result.

## Why it matters

The result suggests that part of what looked like an AI doing creative weight surgery can be done by a classic optimizer, faster and with about 8x fewer changed weights. Barry also found that a steering vector added at inference time recovers nearly the whole fix, though only within a narrow range of strengths.

## Who it's for

interpretability and model-editing researchers

## Try it

git clone https://github.com/nathanrs/weight-space-lp

## Links

- [Blog Post](https://nathan.rs/posts/weight-space-lp/)
- [Code and Edits](https://github.com/nathanrs/weight-space-lp)
- [Original MIT CSAIL Post](https://rikiyat.github.io/blog/weight-space-programming/)
- [HN Discussion](https://news.ycombinator.com/item?id=50452655)