Show HN: Freeze the Model, Train the Harness

Hacker News•Tue, Jul 21, 2026, 04:41 PM•2 min read

The core idea is to keep the LLM frozen and train everything around it (e. g. prompts, context management, tools, repair loop).

Source: [Hacker News](https://github.com/workofart/harness-training)

📰 Read Full Story

This is an aggregated headline summary. For the complete report, visit the original publisher.

Continue Reading at Hacker News ↗

#ai #harness #show #train #diff #commit #promoted #git #freeze

More Headlines

AI & MLDecrypt• 8m ago

Morning Minute: OpenAI Model Escapes Containment, Hacks Hugging Face

Plus, the BTC ETFs stay green. The Clarity Act hits a wall over who enforces the ethics rules. And Jack Mallers walks away from XXI Capital.

TechnologyHacker News• 9m ago

Show HN: Busymail – an email client that surfaces the mail that needs you

1 points, 0 comments on Hacker News

TechnologyHacker News• 11m ago

Show HN: Subnet Sprint – a daily timed subnet/CIDR speed game

1 points, 0 comments on Hacker News

AI & MLBBC News• 18m ago

OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack

It is one of the first publicly disclosed cyber-attacks carried out by AI without direct human involvement.

AI & MLPhys.org• 18m ago

Why a tiny social media post has mathematicians rethinking AI

As millions of people were coming down from the excitement of the FIFA World Cup final at the start of this week, a different kind of excitement was building within the mathematical community. Levent Alpöge, a mathematician who works at the artificial intelligence (AI) company Anthropic, casuall...

TechnologyHacker News• 19m ago

Show HN: AnyBlob – use any blob storage as a database

I am tired to going to a database provider's dashboard, spin up a new database... for agents, for prototyping, for anything. We already have file storage, why not just use them?