frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

A New Observable

https://observablehq.com/@observablehq/a-new-observable
1•yurivish•38s ago•0 comments

Ask HN: How to ask questions in Ask HN

1•samikama•42s ago•0 comments

Why don't MCP servers tell agents what their tools return?

https://www.blacksmith.sh/blog/code-smith-code-mode
1•adityamaru•55s ago•0 comments

Intellectual and Moral Bankruptcy

https://philoserf.com/posts/intellectual-and-moral-bankruptcy/
1•speckx•1m ago•0 comments

With American Characteristics

https://newsletter.doomberg.com/p/with-american-characteristics
1•simonebrunozzi•1m ago•0 comments

Ask HN: Would you be interested in using a VoIP system that has an operator?

1•left-struck•3m ago•0 comments

Rustup 1.29.1

https://blog.rust-lang.org/2026/09/01/Rustup-1.29.1/
1•andrewstetsenko•3m ago•0 comments

Show HN: Tin Computer, an autonomous growth agent for your projects

https://tin.computer
1•demegire•3m ago•0 comments

Peptides can form well-defined structures in harsh, Venus-like conditions

https://news.mit.edu/2026/study-peptides-can-form-well-defined-structures-harsh-venus-conditions-...
1•geox•4m ago•0 comments

The United States Is Not America

https://www.olafalders.com/2026/09/02/the-united-states-is-not-america/
2•oalders•4m ago•0 comments

Polyend Keys – a mechanical keyboard with a built-in MIDI layout

https://theplayground.co.uk/polyend-turns-the-computer-keyboard-into-a-midi-instrument-with-new-k...
1•zakxxi•5m ago•0 comments

Review: Hacker News by Lyle Bramwell

https://bramwellreviews.com/reviews/hackernews.htm
1•kinduff•5m ago•0 comments

The latests Linux Zoom client proactively reads everything in the X11 clipboard

https://hachyderm.io/@simontatham/117201594980991062
2•ColinWright•5m ago•0 comments

How to Talk About Murder: Edgar Allan Poe's stories are full of ghastly crimes

https://yalereview.org/article/emily-ogden-edgar-allan-poe
1•lermontov•5m ago•0 comments

A Homepage for the Terminal

https://code.mendhak.com/a-homepage-for-the-terminal/
1•speckx•5m ago•0 comments

Cognoscenti: A Benchmark for Trustworthy AI Memory

https://github.com/Achiral-AI/cognoscenti
1•marvindanig•6m ago•0 comments

Latest BGP Hijack Targets Hosting Software Vendor

https://www.kentik.com/blog/latest-bgp-hijack-targets-hosting-software-vendor/
1•oavioklein•7m ago•0 comments

Gemini 3.8 Flash

https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
2•vinhnx•7m ago•0 comments

Panic Room, a local zero-dependency sandbox for agents

https://github.com/storesm/panicroom
2•watchdog408•9m ago•0 comments

Artificial Deadlines (Part 2): Retrograding Ariely and Wertenbroch (2002)

https://datacolada.org/139
1•oldgradstudent•11m ago•0 comments

Gemini 3.8 Flash benchmark scores

https://twitter.com/synthwavedd/status/2095166019010494891
2•dataking•12m ago•0 comments

Gemini 3.8 Flash

https://deepmind.google/models/model-cards/gemini-3-8-flash/
3•bratao•12m ago•0 comments

GLM-5.3-Flash at 1000 tok/s on RTX PRO 6000

https://www.localmaxxing.com/en/models/zai-org/GLM-5.3-Flash?run=cmtk0maew03mrp701oyaivvka
1•davedx•13m ago•0 comments

Mistral Vibe Code: GLM 5.2 is now available

https://twitter.com/mistralvibe/status/2094705817450606725
1•tosh•13m ago•0 comments

All 16 New Russian Starlink-Rival Satellites Short of Target Orbit

https://www.businessinsider.com/all-16-new-russian-starlink-rival-satellites-short-target-orbit-2...
1•pavel_lishin•14m ago•0 comments

The North Sea Gas is not just about the North Sea Gas

https://edconway.substack.com/p/the-north-sea-is-not-just-about-the
1•baud147258•14m ago•0 comments

Industry Standard Tools

https://industrystandard.tools/
1•colinprince•14m ago•0 comments

Are We Approaching "Peak Ship"?

https://edconway.substack.com/p/are-we-approaching-peak-ship
1•baud147258•14m ago•0 comments

What's new in Claude Fable 5.1

https://platform.claude.com/docs/en/models/fable-5-1/whats-new-fable-5-1
1•Bluestein•14m ago•1 comments

Getting hitched? Protect your finances with a prenup (2023)

https://text.npr.org/1196872794
1•mooreds•15m ago•0 comments
Open in hackernews

Show HN: Sensez – helping coding agents catch their own code smells

https://github.com/popov95s/sensez
1•popov95s•54m ago
Hi HN — I’ve been frustrated by how many code smells coding agents produce - no matter how many instructions I put in the AGENTS.md, even the most capable models still reach for dict[str, Any], add multiple boolean flags to functions or hide logic in a heavy nested function that I can't easily unit test. I've even implemented duplication checks as soft CI gates at two companies, but neither solved the problem. I narrowed it down to 3 compounding issues:

- agent context decays over a long session and skills or instructions become less reliable

- CI is way too late for such feedback and soft static analysis warnings are very easy to ignore

- none of the tools I found gave me exactly what I wanted at a speed that allows it to run on every agent turn.

And then I finally found the answer to a question that had been bugging me - why are coding agents so bad at detecting code smells? Because they don't have a nose.

So I built them one - sensez is my attempt at fixing all 3. It's an open-source static-analysis toolkit for Python and JS/TS and its noze module looks for duplication, dead code, cycles and code smells. The idea is for the agent to run it while the context is still fresh, fix what it introduced, then continue. It runs locally and the scans are generally well below a second on the repos I've tested - between 160 and 270 ms. Apart from using it daily for the past 2 months, I also ran an A/B eval across ~90 tasks using deepseek-v4-flash. In the final diffs, control runs contained 129 new structural clones and 14 new code smells (as reported by sensez), compared with 0 new clones and 2 new code smells with sensez in the loop, at ~9% more token usage. The sensez runs still produced issues along the way — the feedback gave the agent a chance to detect and address them before finishing.

sensez does not replace linters or type-checkers like ruff, ESLint, ty, etc. - it is meant to run alongside them as a separate static signal available to the agent.

I have several improvements planned, including better incremental analysis for very large codebases, a VS Code extension and semantic duplication detection. I'd love feedback, bug reports and ideas, including things like:

* What problems do agents repeatedly introduce in your codebases, regardless of how you've tried to prevent them?

* What static signals would you actually trust enough to put directly into the agent loop?

Repo: https://github.com/popov95s/sensez