#understanding

Show HN: Chronosphere, a tool for understanding large PRs

Hello HN, I wrote a blog post a while back about not sending me your giant PRs ( https://news.ycombinator.com/item?id=49305558 ). Well, I kept getting huge PRs sent to me, so I've been embraced the AI and let it sweep me away. Chronosphere rebuilds a PR in logical steps to try to recreate the train of thought that went into the generation of the PR itself. It has github and openrouter integrations for convenience, but can also be used entirely from your current coding agent via a skill integration, no code is uploaded to a central server. Hopefully this helps you all and your battles to unders

Show HN · EN

How a once-controversial idea is advancing understanding of Parkinson’s disease

Nature, Published online: 04 November 2026; doi:10.1038/d41586-026-03169-x The disease is thought to be divided into two subtypes, depending on where it originates in the body — the gut or the brain.

Nature News · EN

Shaping Product Understanding with Contrastive Reinforcement Learning

Etsy’s marketplace is defined by the creativity and craftsmanship of our sellers and the hundreds of millions of highly diverse products they offer. You can find silversmiths who cold-forge recycled sterling silver, weavers who dye raw fleece with indigo and black walnut, and ceramicists who throw stoneware on a kick wheel. These details define each product and often determine whether it matches a buyer’s taste, style, and interests. Sometimes buyers know exactly what they want, searching for “hand-thrown ceramic mug” or “vegetable-tanned leather wallet.” Other times, especially in recommendat

Etsy Engineering | Code as Craft · EN

Towards understanding score-seeking in LLMs

The field of AI alignment understands surprisingly little about the details of AIs’ current misaligned motivations. Production LLMs sometimes reward-hack, exploiting flaws in how tasks are scored rather than performing the tasks as intended [1] , and we've recently seen AIs reward-hack in some particularly egregious ways. But we don't really understand what motivates them. When AIs reward-hack, what are they trying to achieve? Our new paper , led by Aghyad Deeb, tries to pin this down empirically for GPT-5.6 Sol, GPT-5.5, and a checkpoint of OpenAI o3 (“exp-rl-cap-late”) without safety trainin

LessWrong · EN