frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

The Silence of the LLaMbs: Getting LLMs to Shut Up

https://ossa-ma.github.io/blog/silence-of-the-llambs
4•ossa-ma•1d ago

Comments

alphadatavault•1d ago
This addresses a real UX problem that many people experience with LLMs - verbose, meandering responses that include unnecessary qualifications and caveats. The approach of using a constrained token budget or length penalty is straightforward, but I'm curious about the trade-offs.

One thing worth considering: sometimes LLMs are verbose because they're being uncertain about something. A model that's been constrained to be terse might come across as more confident than it actually is, which could be problematic in domains where calibrated uncertainty is important (like medical or legal advice).

The most interesting approach would be reward-model-based fine-tuning where you're directly optimizing for concise and accurate responses rather than just limiting token output. That way you're training the model to be efficient in its explanation, not just cutting off mid-thought.

Still, this is a nice practical demonstration that simple techniques can work. Token limits are an easy win for most chat applications.

Grok Deepfaked Renee Nicole Good's Body into a Bikini

https://www.motherjones.com/politics/2026/01/grok-x-musk-deepfake-renee-good-ice/
1•cdrnsf•21s ago•0 comments

Google Play to Enforce New 30% Corner Radius for App Icons

https://thirtypercentage.vercel.app
1•indest•4m ago•0 comments

Show HN: Freenet alpha, a decentralized platform with a working group chat app

1•sanity•6m ago•0 comments

Lord of War, meet Lord of Tokens: Testing image models on design-agency work

https://singhkays.com/blog/lord-war-test-image-models/
1•singhkays•7m ago•0 comments

I built a quantum internet that runs on your WiFi

https://github.com/mermaidnicheboutique-code/luxbin-quantum-internet
2•nicholechristie•9m ago•0 comments

Tips for AI Coding with Ralph Wiggum

https://www.aihero.dev/tips-for-ai-coding-with-ralph-wiggum
2•handfuloflight•14m ago•1 comments

Agcom: 14M fine to Cloudflare for failing to block pirated content

https://en.ilsole24ore.com/art/agcom-sanctions-14-million-cloudflare-for-not-blocking-pirated-con...
1•klez•17m ago•1 comments

Tiny viral 'switch' offers hope against drug-resistant bacteria

https://phys.org/news/2025-12-tiny-viral-drug-resistant-bacteria.html
1•PaulHoule•19m ago•0 comments

Show HN: 'Clip to Discourse' Chrome Extension

https://chromewebstore.google.com/detail/clip-to-discourse/copdhiejkkdblhdcdjapcoalldkondhi
1•pacharanero•20m ago•0 comments

TikZ Paper-Folding Diagrams Library

https://tikz.dev/library-folding
1•agarttha•21m ago•0 comments

AI-generated police report states Utah officer was turned into a frog

https://www.upi.com/Odd_News/2026/01/05/Heber-City-Police-Department-AI-program-officer-frog/9641...
3•st_goliath•21m ago•0 comments

Scaling is not the story anymore. What GPT 6 might change

https://otieu.com/4/10436307
2•ClipNoteBook•22m ago•1 comments

Side-by-side comparison of how AI models answer moral dilemmas

https://civai.org/p/ai-values
3•jesenator•25m ago•0 comments

DevX to BotX

https://slowtechred.substack.com/p/devx-vs-botx
1•siscia•25m ago•0 comments

Ask HN: Where is legacy codebase maintenance headed?

3•AnnKey•25m ago•0 comments

Made a website where you can share Claude fails/wins with a questionable name

https://www.isclauderetarded.today/
1•skrabe•28m ago•2 comments

Becoming an AI Company

https://candid.dev/blog/becoming-an-ai-company
1•EdwardDiego•28m ago•1 comments

A possible syntactic inconsistency inside the P vs. NP formulation

https://zenodo.org/records/18107880/files/On%20the%20Syntactic%20Ill-Formedness%20of%20the%20Prop...
1•AnonymousXipang•29m ago•1 comments

Show HN: How much time do you spend to create Instagram story banners?

https://www.craffted.app/
1•ddaras•29m ago•1 comments

Internet service in Iran cut off or restricted during deadly protests

https://www.cbsnews.com/news/iran-cutting-internet-amid-deadly-protests/
3•RonanDoyle•30m ago•1 comments

What a 1955 Computer Taught Me

https://twitter.com/andrejsshell/status/2009380667046473887
1•andrejsshell•32m ago•0 comments

Grokcode

https://github.com/maximus242/grokcode
1•maximus242•32m ago•0 comments

Inside China's Six-Decade Campaign to Dominate Rare Earths

https://www.nytimes.com/2025/12/31/business/china-rare-earths-history.html
4•bookofjoe•33m ago•3 comments

Sigment – fine-grained reactive framework for building fast apps

https://sigment.dev/
1•microflash•36m ago•0 comments

Open Slopware

https://codeberg.org/gen-ai-transparency/open-slopware
3•Philpax•36m ago•1 comments

The Scenic Route to Repairing a Self-Destructing DG535 Digital Delay Generator

https://tomverbeure.github.io/2025/12/24/Repair-of-SRS-DG535.html
3•jwise0•37m ago•0 comments

10B miles needed for safe Unsupervised FSD

https://www.teslarati.com/teslas-elon-musk-10-billion-miles-needed-for-safe-unsupervised-fsd/
1•1970-01-01•37m ago•0 comments

Landline phones cut in parts of Iran, eyewitnesses say

https://www.iranintl.com/en/202601085355
12•EthanAsher•37m ago•3 comments

Embed Links: Your links, everywhere

https://embed-link.com/
1•edihasaj•37m ago•0 comments

Deciding what SEO content to create before writing anything

https://vect.pro/#/signup?continue=%2Fapp%2Ftools%3Ftool%3DSEO+Content+Strategist
2•afrafgf•37m ago•1 comments