frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

What are you looking for when reviewing LLM generated code?

2•gnunicorn•1h ago
For work I often have to review multi thousand file changes in PR mostly generated by AI/Llama/claude code. It's mainly rust with many unit tests, so there is a lot tooling helps verify the code already to help a lot. But I struggle with these reviews, finding me to just glance over at times unable to comprehend the changes in any higher level capacity (e.g. architecture). Which I think is mainly due to so many changes of these thousands of lines happening that I lack any coherent model of what things look like in the first place. And the LLM generated descriptive commit messages or the PR description and documentation of what they do or even graphite doesn't really help. I wouldn't even be sure I could verify over thousands of lines in multiple commits whether these changes end up doing the architecture described.

So, I wonder, how do other people experience this? How do you review all that code that these generated and what are you looking for? Especially when CI/tooling takes care of all the lints and compile problems? Or do you just yolo that code now, accepting the occasional hallucination?

Comments

JuicerSocial•1h ago
For me personally I tried to keep this PRs without too many files at once plus we use skills to create a workflow.

For example when dev is trying to push something for review we ask them to:

run /simplify skill from claude, then do /code-review on ultra effort, address any issues it finds and then ask for review.

After that we review it manually if there is anything suspicious, we test it in UI...

It doesn't matter if it works

https://henry.codes/writing/it-doesnt-matter-if-it-works/
1•tobr•51s ago•0 comments

Fallow – Deterministic codebase intelligence for TypeScript and JavaScript

https://github.com/fallow-rs/fallow
1•modinfo•3m ago•0 comments

From Australia to Europe, countries move to curb children's social media access

https://www.reuters.com/legal/government/australia-europe-countries-move-curb-childrens-social-me...
2•geox•3m ago•0 comments

Five Papers That Show Where AI Research Is Heading

https://www.youtube.com/watch?v=3rWSvrFahIY
2•root-parent•6m ago•0 comments

Show HN: Memorypad – A note editor for your daily notes, in Markdown

https://memorypad.io/
1•Malfunction92•6m ago•0 comments

The Third Generation of Apple's Foundation Models

https://machinelearning.apple.com/research/introducing-third-generation-of-apple-foundation-models
1•fragebogen•8m ago•0 comments

Tuningfork – LLM agent grounding rules derived from human reality-testing

https://github.com/T-Chartrand/tuningfork
1•T-Chartrand•9m ago•0 comments

The agent principal-agent problem

https://crawshaw.io/blog/agent-principal-agent
1•tosh•12m ago•0 comments

Lamport Signature

https://en.wikipedia.org/wiki/Lamport_signature
1•teleforce•16m ago•0 comments

The Hanging Gardens Problem (2024)

https://ar-ms.me/thoughts/the-hanging-gardens-problem/
1•tosh•16m ago•0 comments

Rails: The Sharp Parts. Callbacks Are Not Invariants

https://baweaver.com/writing/2026/06/13/rails-sharp-parts-callbacks-are-not-invariants/
1•doppp•16m ago•0 comments

Rails: The Sharp Parts. Queries, Read Models, and Batching

https://baweaver.com/writing/2026/06/14/rails-sharp-parts-queries-read-models-and-batching/
1•doppp•16m ago•0 comments

Summary of Key Changes in OMB's Proposed Federal Financial Assistance Rule

https://elizabethginexi.substack.com/p/summary-of-key-changes-in-ombs-proposed
1•Anon84•16m ago•0 comments

Show HN: CLI for crawling documentation sites into Markdown with defuddle

https://github.com/artemnistuley/docrawl
1•nistuley•19m ago•0 comments

Cyber leaders defend Anthropic's banned model

https://www.axios.com/2026/06/15/anthropic-fable-security-leaders-trump-admin
1•giuliomagnifico•20m ago•0 comments

Batta – plan-phase security reviews for AI coding agents (OSS)

https://github.com/Dolevco/batta-ai
1•dolevco1•20m ago•0 comments

IT Wage Slave Simulator

https://knowledgeworker.alexmeub.com/
1•ed_mercer•22m ago•0 comments

The Anthropic Fable Mess

https://www.cautiousoptimism.news/the-anthropic-fable-mess-explained/
1•handoflixue•24m ago•0 comments

How to pass Series A security due diligence (before it catches you off guard)

https://faultlinesec.com
2•albernaz•26m ago•0 comments

"is just a state change" [video]

https://www.youtube.com/shorts/Wl2bFXEDdJ0
1•modinfo•26m ago•0 comments

Is Vibe Coding Dead? Even Karpathy Is Moving On

https://www.forbes.com/sites/jodiecook/2026/06/12/is-vibe-coding-already-dead-even-karpathy-is-mo...
3•indigodaddy•32m ago•0 comments

SvGrid – A Svelte 5 data grid with a headless core and an MCP server

https://svgrid.com/
1•boikom•33m ago•0 comments

SAT-Based Approaches for the General High School Timetabling Problem (2017) [pdf]

https://dbai.tuwien.ac.at/staff/musliu/emird.pdf
3•tosh•34m ago•0 comments

AI SQL Advisor integrated into a famous database tool

https://github.com/Wisser/Jailer
1•rwisser•37m ago•0 comments

Psilocybin Unlocks Lost Memories in an Alzheimer's Patient

https://neurosciencenews.com/psilocybin-memory-alzheimers-30084/
1•giuliomagnifico•38m ago•0 comments

Services by Lifecycle

https://www.michaelnygard.com/blog/2018/01/services-by-lifecycle/
1•alex_brajkovic•39m ago•0 comments

Anthropic's Safety Superpower

https://stratechery.com/2026/anthropics-safety-superpower/
4•swolpers•43m ago•0 comments

Ask HN: Why are Spec-kit specs like that

1•pretorian_paul•45m ago•0 comments

Philippines now China's 2nd biggest market for solar panels

https://www.philstar.com/business/2026/05/29/2531211/philippines-now-chinas-2nd-biggest-market-so...
5•doener•47m ago•1 comments

Neuroscientists discover cognitive benefits of reading physical comic books

https://www.psypost.org/neuroscientists-discover-previously-unknown-cognitive-benefits-of-reading...
2•emot•48m ago•1 comments