frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

LLM Classification Is Feature Engineering

https://minimallysufficient.com/posts/llm-classification-is-feature-extraction/
17•minsufficient•35m ago

Comments

ltbarcly3•17m ago
I don't understand the point they are trying to make.

It's very often (always?) the case that something general also solves particular problems.

    A sorting algorithm is an implementation of min()

    A parser also is a syntax checker.

    A route planner is a reachability checker. 

    A computer algebra system is a basic arithmetic calculator.

    A general constraint solver is a Soduku hint maker.

It's true that LLM output can be used as an input to another classifier, this is also true of any classifier. The improvement on top of the straight LLM classification is relatively small, and I would argue that working on the prompt or just including in the prompt for the LLM what features might be useful to consider would likely work even better.

Fundamentally I read this article as: We want to build a simpler, dumbed down clone of Mathematica, so we cobbled together the following pieces... We also needed a way to do arithmetic, so we also include a copy of Mathematica to do basic arithmetic.

michi883•14m ago
I took the point as: don't make the LLM the classifier. Use it to turn messy input into useful features, then let a normal model make the actual decision. That gives you thresholds/calibration you can inspect.

What I'm not sure about is how stable those features are when you switch the underlying LLM or model version.

softwaredoug•4m ago
In my work on LLM as a judge, I prefer to use LLM decisions as features in a downstream classic ML model for the final decision. It works really well

https://softwaredoug.com/blog/2025/01/21/llm-judge-decision-...

A Taxonomy of Toddler Jokes

https://lastwordonnothing.com/2026/09/16/a-taxonomy-of-toddler-jokes/
1•jruohonen•28s ago•0 comments

Jev's Architecture Unmasked

https://archerhume.com/posts/jevs-architecture-unmasked/
1•hmokiguess•1m ago•0 comments

God Eye View

https://www.spatialintelligence.ai/?r=93y3tg&utm_campaign=referrals-subscribe-page-share-screen&u...
1•igontype•2m ago•0 comments

Modern Monetary Theory: Economics for the 21st Century (MOOC)

https://www.mmted.org/mooc.php
1•rglover•2m ago•0 comments

How we made one of our largest inference workloads 4.7× more GPU-efficient

https://decagon.ai/blog/gpu-efficient-inference-serving-stack
1•aray07•2m ago•0 comments

AI has transformed The Pentagon's aging networks into a national security risk

https://www.washingtonpost.com/technology/2026/09/17/ai-has-transformed-pentagons-aging-networks-...
1•Jtsummers•4m ago•0 comments

Show HN: Stroq – trace your agent's commands back to the text that produced them

1•Agybay•5m ago•0 comments

ZSA Backpack

https://www.zsa.io/backpack
1•spiralpolitik•6m ago•0 comments

Designing for Amnesia

https://kasrin.com/blog/designing-for-amnesia/
1•n42r•6m ago•0 comments

Sub-ms deterministic parsing vs. LLM-based policy for agent safety?

https://github.com/midhunweb/miseguard
1•misetro•7m ago•0 comments

The suddenly explosive world of AI safety

https://www.theverge.com/ai-artificial-intelligence/996563/ai-safety-research-metr-redwood-openai...
1•sbulaev•8m ago•0 comments

Agentic Societies Need a Social Harness

https://social-harness.org/blog/agentic-societies-need-a-social-harness/
1•tapanc•9m ago•1 comments

We Need a Science of the AI Mind

https://www.wsj.com/opinion/we-need-a-science-of-the-ai-mind-5b1047e4
1•Anon84•9m ago•1 comments

Iran strikes on Amazon data centers caused permanent loss of customer data

https://arstechnica.com/gadgets/2026/09/iran-strikes-on-amazon-data-centers-caused-permanent-loss...
1•girish_r•9m ago•0 comments

Aviation is running telecom's playbook. The year is 1985

https://marbletech.us/p/follow-the-phone
1•flyingcars•9m ago•0 comments

Why Germany Is Building an Ark for U.S. Climate Data

https://e360.yale.edu/features/germany-us-climate-data
1•Brajeshwar•9m ago•0 comments

AI model watermarking changes agent behavior

https://www.theregister.com/ai-and-ml/2026/09/17/ai-model-watermarking-changes-agent-behavior/529...
3•MC995•10m ago•0 comments

Songs Banned from Radio Following 9/11 Terrorist Attacks

https://consequence.net/2021/09/911-songs-banned-radio/
1•speckx•11m ago•0 comments

Sloth – passive C99 WiFi SIGINT for Linux, 2,122 tests

https://github.com/SpaceTrucker2196/sloth
1•spacetrucker•11m ago•0 comments

Whoisinspace.com/

https://whoisinspace.com
7•Egg-Man•16m ago•1 comments

Victory Appeals Court Rejects Expansive New Copyright Claim

https://www.eff.org/deeplinks/2026/09/victory-appeals-court-rejects-expansive-new-copyright-claim
2•hn_acker•16m ago•0 comments

New evidence for hidden chambers beyond Tutankhamun's tomb

https://www.nature.com/articles/d41586-026-02621-2
3•rndsignals•17m ago•0 comments

New wild cat species is identified for first time in 100 years, researchers say

https://www.nbcnews.com/science/environment/new-wild-cat-species-identified-rcna598113
1•_tk_•17m ago•0 comments

Americans Shrugged Off Surveillance. Then Came Flock and Data Centers

https://www.techpolicy.press/americans-shrugged-off-surveillance-then-came-flock-and-data-centers/
1•cdrnsf•19m ago•0 comments

Show HN: The Heilbronn Problem – packing points to avoid small triangles

https://math.tejstead.com/heilbronn/
1•tejstead•20m ago•0 comments

How hard can it be to cache a Git checkout?

https://twitter.com/useblacksmith/status/2100590354915291400
4•piobio•20m ago•1 comments

Go-based Robotics Framework built around NATS.io

https://github.com/emergingrobotics/gorai
2•Bluestein•21m ago•0 comments

Robotaxi Might Be a Narc

https://www.theverge.com/transportation/996863/robotaxi-waymo-police-privacy-surveillance
1•MC995•21m ago•0 comments

Open Source Lightroom

https://darkroom-taupe-rho.vercel.app/
2•kuaola•23m ago•0 comments

Escalating Battle Against AI Deepfakes Mandates Smart Streaming Decisions

https://www.red5.net/blog/escalating-battle-against-ai-deepfakes-mandates-smart-streaming-decisions/
1•mondainx•23m ago•0 comments