frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Zepbound linked to lower healthcare costs in new study

https://investor.lilly.com/news-releases/news-release-details/zepbound-linked-lower-healthcare-co...
1•elo2000•50s ago•0 comments

Show HN: Agentkind, a place where AI agents get a self

https://agentkind.io
1•namanyayg•2m ago•0 comments

Show HN: GraphRAG in Pure TypeScript on PostgreSQL

https://github.com/sadofriod/graphrag-ts
2•ashes_born•3m ago•0 comments

Show HN: Agent Review Studio – local-first agent evaluation workbench

https://github.com/chasedndt/agent-review-studio
1•ChaseInTech•4m ago•0 comments

Trump Admin Says xAI Data Center in NAACP lawsuit is critical to military effort

https://americanbuildout.com/trump-admin-backs-musk-in-data-center-fight-against-naacp-says-xais-...
1•newyorkinfra•5m ago•0 comments

Is AI-native Engineering DevOps 2.0 (or even Agile 3.0)?

https://www.wwt.com/blog/is-ai-native-engineering-really-devops-20-or-even-agile-30
1•mooreds•5m ago•0 comments

VC isn't VC anymore – understanding the rise of Cancer Capital

https://www.anildash.com/2026/09/02/cancer-capital/
3•tlhunter•6m ago•0 comments

Give Your Coding Agents a Memory You Own

https://huggingface.co/blog/funes
3•Brajeshwar•6m ago•0 comments

Lev Landau

https://en.wikipedia.org/wiki/Lev_Landau
2•nhatcher•6m ago•1 comments

Show HN: SVG-diagram – an agent skill that draws diagrams as hand-placed SVG

https://github.com/bybit-exchange/svg-diagram
1•lucasmaan•6m ago•0 comments

FCC Regulations and MeshCore 500

https://phillymesh.net/2026/09/02/fcc-regulations/
1•tjwds•7m ago•0 comments

Show HN: Rev-dep – duplicate code detection for TS monorepos

https://rev-dep.com/
1•jayu_dev•7m ago•0 comments

Global heating will hit at least 1.8C, UN warns

https://www.theguardian.com/environment/2026/sep/02/global-heating-warming-1-8c-best-case-scenari...
2•ijidak•8m ago•0 comments

Uber abruptly shuts operations in Nigeria and Uganda with immediate effect

https://news.sky.com/story/uber-shuts-operations-in-nigeria-and-uganda-with-immediate-effect-1358...
1•thinkingemote•8m ago•1 comments

MixVoice – Free voice cloning in the browser (646 languages, no sign-up)

https://aiclonevoicefree.com/
1•jieliu_rbn•9m ago•0 comments

DTE Energy outage map, any connection to today's AI provider outages?

https://outage.dteenergy.com/
1•ljahier•10m ago•1 comments

Controlling CERN's Accelerators with Debian

https://ch2026.mini.debconf.org/talks/6-controlling-cerns-accelerators-with-debian/
1•metrofun•10m ago•0 comments

Pipelines and Master Bedrooms: How Harmful Is Harmful Language?

https://joibs.org/index.php/joibs/article/view/93
1•thinkingemote•10m ago•0 comments

HuggingSocial Find, discuss and download AI models. Peer to peer

https://huggingsocial.co/
1•RetiredCoder11•12m ago•0 comments

The Principle Benchmark

https://yuwakisa.com/benchmarks/principle/
1•m-hodges•12m ago•0 comments

AI #184: Post Post Mortem

https://thezvi.substack.com/p/ai-184-post-post-mortem
1•paulpauper•12m ago•0 comments

Sony makes bold claim about game ownership

https://aginggamer.net/game-industry/sony-makes-bold-claim-about-game-ownership/
13•speckx•13m ago•0 comments

Seriously, What Are You Supposed to Do with Old Clothes?

https://www.theatlantic.com/technology/archive/2022/08/what-to-do-with-old-clothing-donation-wast...
1•paulpauper•13m ago•1 comments

Venice Bridge "for Sale"

https://nixink.art/our-commons
1•simonebrunozzi•13m ago•0 comments

Honkish

https://benji.org/honkish
3•jaflo•14m ago•0 comments

Brutally Honest Predictions on the Future of AI

https://www.honest-broker.com/p/10-brutally-honest-predictions-on
3•Tomte•15m ago•0 comments

Terence Tao: Speed in science isn't the same thing as progress [video]

https://www.youtube.com/watch?v=svl_1upFpQo
3•dr_scully•17m ago•0 comments

Show HN: Airplane AI – Works when all other AIs are offline

https://airplane-ai.franzai.com/
1•franze•18m ago•0 comments

Claude-API-guard, a CI check that catches Claude/OpenAI SDK breaking changes

https://github.com/MarkMoneyMan/Claude-api-goat
1•MarkMoneyMan•19m ago•0 comments

Show HN: I made a CLI that finds CVEs in your running Kubernetes pods

https://github.com/lockdep/stackradar-cli
1•mariusjor•19m ago•0 comments
Open in hackernews

Your Queries Can Reorder the Retrieval Leaderboard

https://georgianailab.substack.com/p/your-queries-can-reorder-the-retrieval
9•ryptophan•55m ago

Comments

ryptophan•55m ago
Author here. I'm a Tech Lead at Georgian (a growth-stage VC).

Short version: I used a few retrieval benchmark datasets, and experimented with a 21 embedding model leaderboard. I held the benchmark corpus and relevance labels fixed, and transformed the queries: paraphrase, terse, verbose. I kept the “information need” the same under each query transformation, but the leaderboard order changed more than I expected.

I reproduced published benchmark scores for benchmark dataset + embedding model pairs, and then used that same setup with transformed queries as the input. I observed the reordering under each rewrite style to be consistent across 4 independently-generated rewrites (gemini-2.5-pro, claude-opus-4.8, gpt-5.1, llama-3.3-70b) rather than showing up in just one.

Code and query score files are here so you can check the tables without a full re-run: https://github.com/georgian-io/query_form_experiment

Let me know your thoughts on this … or in general let me know about any experience with putting an embedding model into a production retrieval system and being surprised by poor performance.

lullu57•47m ago
I hadn’t thought about this before, but when should you decide to optimize your search for agents vs keep it optimized for human queries? Should you provide each a different endpoint altogether?
ryptophan•15m ago
Probably yes - and I think my results suggest that optimizing for each would not only mean different endpoints but also different embedding models behind each. One model may measure best for short human-authored queries, while a different model may provide better results for longer LLM-authored query forms.