frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Ask HN: How does one build an LLM router with a labelled set of 100 tasks?

1•zodwick•13m ago
I already spent the time building rubrics for a set of ~100 tasks on various criterias (accuracy, thoroughness communication style ). Then ran these on a few models and now have a labelled-set where I know for each task which is the cheapest/fastest model that gets a minimum threshold ( say 80 ).

How does one go on to build a router based on this ? There appears to be no good literature / open source projects on this.

I tried a bunch of approaches with the help of claude: - trying to use Jev via a serious of probe Yes/No questions ( does task contain x, does the world have y ...etc ). But this didn't yield good results. - kNN on embeddings of the task - in retrospect this was set to fail from the tasks, as now I think we do need some sort of NN to figure out the world/task complexity. - classifying prompt-complexity - something like https://catalog.ngc.nvidia.com/orgs/nvidia/nemo/models/prompt-task-and-complexity-classifier/-

The main issue from a conceptual standpoint is that one cannot fully predict how hard a task is until you are into the task - like you might encounter unknown nuances or issues in later stages, but classification has to happen at start. So I am starting to doubt if this is even possible tbh

Aura: 1.3B speech-to-video model that runs in real time on a single H100 GPU

https://huggingface.co/kyutai/AURA-1.3B
1•armcat•53s ago•0 comments

Finite-time blow-up for the Navier–Stokes equations

https://github.com/tristanbuckmaster/navier-stokes/
1•trotro•2m ago•0 comments

Agentic Workflow Governance with MFA

https://studio.sinapsdata.com.br/
1•kardassnp•5m ago•0 comments

Show HN: Sandscale – Firecracker microVMs for AI coding agents

https://sandscale.dev
2•Sanjay05•7m ago•0 comments

TypeScript Is Enough

https://twitter.com/dashersw/status/2123724189651612042
1•arbayi•7m ago•0 comments

Show HN: Online curling game/SIM with Three.js

https://www.cloudcurling.com
1•nrjames•7m ago•0 comments

You trust your bank. But who else has your information?

https://inclinq.medium.com/chase-visa-and-mastercard-a-vendor-breach-can-expose-your-data-without...
1•inclinq•7m ago•0 comments

Show HN: Tikivela – Online multiplayer football career game

https://tikivela.com/
1•faridmovsumov•7m ago•0 comments

Next wave: data, AI agents, and no-code automation

https://aws.amazon.com/marketplace/pp/prodview-pvfpfkgk2vdba
1•sadrk•7m ago•0 comments

EU confirms AI-assisted cyberattack on European Commission

https://apnews.com/article/cyberattack-european-union-artificial-intelligence-ai-79705095e8141cd6...
2•datascientist•8m ago•0 comments

Show HN: Formbook – MCP server for horse racing data

https://mcp.formbook.io
1•beginnings•8m ago•0 comments

Signal restricts screenshots to block AI from reading messages

https://www.reuters.com/legal/litigation/signal-restricts-screenshots-block-ai-agents-reading-mes...
2•laurex•10m ago•2 comments

Slopstrap – Make your site look hastily vibecoded

https://slopstrap.mikelacher.com/
2•mlacher•12m ago•0 comments

Programming in English

https://pvs-studio.com/en/blog/posts/1428/
1•thunderbong•12m ago•0 comments

Top AI models have mostly stopped inventing citations, Stanford study

https://thinkpol.ca/2026/11/19/top-ai-models-have-mostly-stopped-inventing-citations-stanford-ret...
1•bazlightyear•13m ago•0 comments

Ask HN: How does one build an LLM router with a labelled set of 100 tasks?

1•zodwick•13m ago•0 comments

How Long Does It Take to Read the Bible?

https://jesussavescities.com/read/how-long-does-it-take-to-read-the-bible/
2•titusblair•14m ago•0 comments

Debian APT to drop legacy features, citing advances in AI security research

https://lists.debian.org/debian-devel-announce/2026/11/msg00001.html
1•khalidkhair•14m ago•0 comments

The Dubious Idea of AI-Run Companies

https://read.misalignedmag.com/the-dubious-idea-of-ai-run-companies-b26c49c77dc5
1•lcubw•14m ago•1 comments

Riemann Hypothesis Is Proven

https://twitter.com/sytelus/status/2123797987625431512
1•sytelus•16m ago•0 comments

COO – Covert Ornithological Operations

https://coo.mx/
1•dirkk0•17m ago•0 comments

Tinyeval – 1 file evals framework, language agnostic

https://github.com/evalship-hq/tinyeval
1•gghootch•19m ago•0 comments

Show HN: Daily Games Today – DLE Games Like Krillion

https://dailygames.today/
1•sparkmesh•20m ago•0 comments

US fishers see offshore wind as a 'godsend'. Trump wants to shut it down

https://www.theguardian.com/environment/2026/nov/20/offshore-wind-farms-trump-fishers
1•dzonga•20m ago•0 comments

Open-source RASP for Android and iOS apps, no source code changes

https://github.com/Ostorlab/rasp
1•AbirJelti•21m ago•0 comments

Show HN: Preflight-agent – five checks your AI-agent setup needs before it runs

https://github.com/jasveerai88/preflight-agent
1•SinghularityAI•22m ago•0 comments

Tristan Buckmaster Claims Proof of Unforced Navier Stokes and Unforced Euler

https://mastodon.social/@tristanbuckmaster/117649035783718539
1•iamronaldo•23m ago•0 comments

Show HN: HyprMac: Keyboard-driven tiling window manager for macOS

https://hyprmac.app/
1•zachtgray•25m ago•0 comments

Text, Filters, and Now Vectors in One Postgres Index

https://www.paradedb.com/blog/filter-aware-vector-search
1•philippemnoel•27m ago•0 comments

jj explained in 3D, every step compared to Git

https://brajeshwar.com/misc/jj/
2•Brajeshwar•27m ago•1 comments