frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Dear Europe: Germany has shown the way forward

https://blog.documentfoundation.org/blog/2026/03/23/dear-europe/
1•doener•47s ago•0 comments

Pentagon Adopts New Limits for Journalists After Court Loss

https://www.nytimes.com/2026/03/23/business/media/pentagon-closes-journalists-work-area.html
1•doener•1m ago•0 comments

RocksDB development finds a CPU bug

https://rocksdb.org/blog/2026/02/17/cpu-bug.html
1•eatonphil•4m ago•0 comments

Classification: Accuracy, recall, precision, and related metrics

https://developers.google.com/machine-learning/crash-course/classification/accuracy-precision-recall
1•teleforce•4m ago•0 comments

Personal and Business Tax Returns via MCP

1•unmarkdown•6m ago•0 comments

Show HN: AI Morning Briefing – weather, stocks, news, and a quote emailed daily

https://ia2ai.ai/beta
1•ia2ai_official•6m ago•1 comments

XMMS Resuscitated: XMMS ported to gtk4 with gstreamer/pipewire

https://gitlab.com/cschalle/xmms-resuscitated
1•marcodiego•6m ago•0 comments

SettleTheVibe – Let strangers vote on who's right in your real-life disputes

https://settlethevibe.com
1•misterinfo•8m ago•0 comments

Big Nuclear's Big Mistake – Linear No-Threshold - Kyle Hill [video]

https://www.youtube.com/watch?v=gzdLdNRaPKc
1•zeristor•9m ago•1 comments

Study: Workers who fall for 'corporate bullshit' may be worse at their jobs

https://www.theguardian.com/business/2026/mar/23/corporate-speak-study
1•DamnInteresting•9m ago•0 comments

Evercookie

https://en.wikipedia.org/wiki/Evercookie
1•chistev•9m ago•0 comments

Windows 3.1 tiled background .bmp archive

https://github.com/andreasjansson/win-3.1-backgrounds
2•justsomehnguy•10m ago•0 comments

The Belonging Industrial Complex

https://asseenonbyochuko.substack.com/p/the-belonging-industrial-complex
1•herbertl•10m ago•0 comments

Show HN: Remember psDooM, try ps HID Combat

https://github.com/akoerner/kprox
3•alhadrad•12m ago•0 comments

Quantum‑Grade Entropy

https://realrandom.co/
1•structuredPizza•12m ago•0 comments

SQL's TypeScript moment will help agents just as much as it'll help humans

https://roundup.getdbt.com/p/sql-typescript-and-agents
1•craneca0•16m ago•0 comments

In sports vision analysis cloud compute isn't best for speed, cost, connectivity

https://www.sportsbusinessjournal.com/Articles/2026/03/23/sports-tech-isnt-always-cloudy/
1•Austin_Conlon•18m ago•0 comments

Show HN: Dgs-CLI – 63-command CLI for D-Link DGS-1100 switches via Selenium

https://github.com/bobberb/dgs-cli
1•ShellackGobln7•19m ago•0 comments

4chan's hamster-filled reminder US cos needn't comply with UK speech regulations

https://reason.com/2026/03/23/4chan-sends-hilarious-hamster-filled-reminder-that-u-s-companies-ne...
1•jrosenblatt•20m ago•0 comments

Supreme Court declines to review press freedom case

https://www.npr.org/2026/03/23/nx-s1-5757440/supreme-court-press-freedom
2•Jimmc414•20m ago•1 comments

Rasmus Andersson Launches Playbit macOS App

https://twitter.com/rsms/status/2036210025258164264
1•colesantiago•23m ago•1 comments

IRIX 3dfx Voodoo driver and glide2x IRIX port

https://sdz-mods.com/index.php/2026/03/23/irix-3dfx-voodoo-driver-glide2x-irix-port/
3•zdw•26m ago•0 comments

Steps toward great agent experience every API provider can take today

https://www.stainless.com/blog/steps-toward-great-agent-experience-every-api-provider-can-take-today
1•dgellow•29m ago•0 comments

Is Argon2 better than Bcrypt?

https://pilcrowonpaper.com/blog/14
1•troupo•29m ago•0 comments

Ask HN: Karpathy auto research on a miata ecu?

1•alex7o•30m ago•0 comments

Welcome to the AI agent arms race

https://www.axios.com/2026/03/23/openclaw-agents-nvidia-anthropic-perplexity
1•indigodaddy•32m ago•0 comments

GSA extends comments on AI clause after industry pushback

https://fedscoop.com/gsa-extends-comments-on-sweeping-ai-clause-after-industry-pushback/
1•petethomas•34m ago•0 comments

The Builder's Dilemma: Why the World Rewards Hype and How to Win with Real Value

https://medium.com/@pierreneter/the-builders-dilemma-why-the-world-rewards-hype-and-how-to-win-wi...
2•pierreneter•35m ago•0 comments

A Claude plugin for the full software development lifecycle

https://github.com/shamkhall/sdlc
1•shamkhal•36m ago•0 comments

Show HN: Npx poke-PC – Give your Poke agent a full Docker box to play in

https://github.com/calganaygun/poke-pc
1•calganaygun•38m ago•0 comments
Open in hackernews

OpenMath: Ontology-Guided Neuro-Symbolic Inference

https://arxiv.org/abs/2602.17826
3•marcelolabre•1h ago

Comments

marcelolabre•1h ago
Hi HN,

Standard RAG retrieves semantic noise when confronted the logic of specialist domains like maths. I wanted to see if we could treat language models more like compilers by anchoring them to a structural ground truth.

I built a neuro-symbolic pipeline that grounds smaller models (<9B parameters, like Gemma2 and Qwen2.5-Math) in the OpenMath ontology using hybrid retrieval and cross-encoder reranking.

Evaluating on the MATH 500 benchmark revealed a severe bottleneck. When retrieval succeeds, reasoning and convergence improve. But the semantic gap between natural language and formal definitions is massive. When retrieval fails, the injected irrelevant ontological context actively degrades performance, hitting a hard "context utilization ceiling" in smaller models.

The paper and pipeline code are open. I ran these experiments locally without hyperscaler compute. I would love the community's technical feedback.

I’m continuing with the research now toward solving the retrieval quality bottleneck.