frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Laguna XS.2 and M.1

https://poolside.ai/blog/laguna-a-deeper-dive
40•tosh•1h ago

Comments

rohitpaulk•1h ago
Been testing these via their "pool" agent. It's fast, and the agent adheres to the ACP spec pretty well (better than codex, opencode etc.) so it's a good experience in Zed.
throwaw12•1h ago
Has anyone tried these models?

I like their honesty in benchmarks, looks like Qwen3.6 35B is outperforming their Laguna M.1 225B model

kingjimmy•56m ago
the color-codes make those benchmarks charts impossible to understand. very pretty though.
data-ottawa•52m ago
For what it's worth, the bars correspond in order with the legend. Plus there’s hover text.
franksiem•53m ago
Felt like they would never come out of stealth mode but very nice to see it materialized into something competitive.
refulgentis•18m ago
What makes them distinctive?
throwaw12•18m ago
Not sure if this is competitive, look at the numbers for Qwen3.6
jaen•43m ago
For similarly sized models, not looking very good on the slightly-less-benchmaxxed Terminal-Bench 2.0:

  Laguna XS.2  33B-A3B params: 30.6
  Qwen 3.6     35B-A3B       : 51.5
  Devstral 2   123B          : 31.2
Quite a huge lead for Qwen... well, at least it's catching up to other smaller Western labs.
megavon•33m ago
Need to look at SWEBench-Pro, it's super competitive. Suspect they'll catch up given the longer-tail on TB scores.
jaen•17m ago
Just by the (lack of) inter-model variance, I don't think SWEBench-Pro does a very good job of representing model capability. Terminal-Bench seems more challenging and separates the wheat from the chaff.

Also, *ops work, which in my experience can actually be more complicated than SWE is underrepresented there obviously.

speedgoose•42m ago
Please update the charts. Consider using textures or filling patterns.

I usually score pretty well in colour perception tests but distinguishing between those two purples made me doubt myself.

matthewfcarlson•21m ago
My phone is in grayscale to make it less interesting (I still watch way too many videos in grayscale but it helps) so I’m right with you
esafak•11m ago
They're not winning any popular benchmark so is there some niche this excels?

How to Keep Your Brain Sharp: A Practical Playbook Beyond the Basics

https://tim.blog/2026/04/24/how-to-keep-your-brain-sharp/
1•AlexArias•59s ago•0 comments

Claude Design Is 404ing

https://claude.ai/design
1•xena•1m ago•0 comments

Phaser: Create 2D games for the web – free, open source, and AI-ready

https://phaser.io/
1•doener•1m ago•0 comments

Wild GPT-image-2 use cases

https://medium.com/@HungryMinded/5-wild-use-cases-for-gpt-image-2-d9b803c1113c
2•hungryminded•2m ago•0 comments

Amtaitfy – Let Me Google That for You, but the AI Is Wrong on Purpose

https://amtaitfy.com/
1•meghneelgore•7m ago•0 comments

Nvidia Nemotron 3 Nano Omni

https://blogs.nvidia.com/blog/nemotron-3-nano-omni-multimodal-ai-agents/
1•tosh•8m ago•0 comments

Height hunt: a quest to find and visit every possible low bridge / height restri

https://adamtownsend.com/heighthunt/
1•fanf2•8m ago•0 comments

Shots Fired by Google Cloud CEO Thomas Kurian

https://twitter.com/tanayj/status/2048838842031956395
1•jmintz•8m ago•0 comments

Woman's Talkspace therapy app sessions exposed in court

https://www.proofnews.org/womans-talkspace-therapy-app-sessions-exposed-in-court/
1•pavel_lishin•9m ago•0 comments

The Guard Act Isn't Targeting Dangerous AI–It's Blocking Everyday Internet Use

https://www.eff.org/deeplinks/2026/04/guard-act-isnt-targeting-dangerous-ai-its-blocking-everyday...
2•hn_acker•9m ago•0 comments

GPT-Engineer: Precursor to Lovable.dev

https://github.com/antonosika/gpt-engineer
1•doener•11m ago•0 comments

Ask HN: Site that tracks AI subscription token amount?

1•yukIttEft•11m ago•0 comments

Show HN: Inter-session messaging between Claude Code sessions

https://github.com/yilunzhang/claude-code-inter-session
1•skysniper•11m ago•0 comments

OpenAI Models on Amazon Bedrock

https://aws.amazon.com/bedrock/openai/
2•jaredwiener•12m ago•1 comments

Distilling a Tiny Model for Fast Interpretability

https://ethanfast.substack.com/p/a-tiny-model-for-fast-interpretability
1•unignorant•14m ago•0 comments

Apple Weather App Down

https://9to5mac.com/2026/04/28/apple-weather-down-iphone-app-experiencing-issues-right-now/
1•bear_with_me•14m ago•0 comments

Bounce Update: PDS Provider Migrations

https://blog.anew.social/bounce-pds-provider-migrations/
2•Kye•20m ago•0 comments

Google DeepMind Paper Argues LLMs Will Never Be Conscious

https://www.404media.co/google-deepmind-paper-argues-llms-will-never-be-conscious/
1•Brajeshwar•20m ago•1 comments

Why So Many Mayors Are Quitting

https://thewalrus.ca/why-so-many-mayors-are-quitting/
1•speckx•21m ago•0 comments

BookStack Moves from GitHub to Codeberg

https://github.com/BookStackApp/BookStack/issues/4551
14•RadiozRadioz•21m ago•0 comments

Ryzen Saved AMD from Bankruptcy – 10 Years of CPUs Tested [video]

https://www.youtube.com/watch?v=EZeiaK0T3Jk
2•mariuz•22m ago•0 comments

How Semiconductors Were Made in America

https://www.siliconimist.com/p/semiconductors-made-in-america
3•johncole•23m ago•1 comments

Once I Understood Where AI Is Heading, I Stopped Being Anxious About It

https://ai.gopubby.com/once-i-understood-where-ai-is-heading-i-stopped-being-anxious-about-it-849...
2•swolpers•24m ago•0 comments

Buying, Selling on eBay Disrupted Worldwide for more than 24 hours

https://www.sanjoseinside.com/business/buying-selling-on-ebay-disrupted-worldwide-for-nearly-two-...
1•j79•24m ago•1 comments

Universal Transformers Need Memory: Depth-State Trade-Offs in Adaptive Recursive

https://arxiv.org/abs/2604.21999
1•che_shr_cat•28m ago•0 comments

Show HN: Art Coding Lab – Learn Creative Coding Through Micro Challenges

https://artcodinglab.com/
1•absurdwebsite•29m ago•1 comments

GraphCompose – declarative PDF layout engine for Java (MIT)

https://github.com/DemchaAV/GraphCompose
1•demchaav•30m ago•0 comments

Show HN: I built a dating SIM that prepares you for your date

https://claude.ai/public/artifacts/98750067-546b-4c9e-ab62-68cae2941329
2•danish00111•33m ago•0 comments

Study Finds a Third of New Websites Are AI-Generated

https://www.404media.co/study-finds-a-third-of-new-websites-are-ai-generated/
2•Brajeshwar•36m ago•1 comments

GB Electricity Bills

https://www.electricitybills.uk/
2•kieranmaine•36m ago•1 comments