frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

UUIDs, ULIDs and Sqids: A Practical Deep Dive

https://wendelladriel.com/blog/uuids-ulids-and-sqids-a-practical-deep-dive
1•shaunpud•50s ago•0 comments

Strata – An app that talks me out of dying outdoors

https://strata.highloop.co/
1•cloocher•2m ago•0 comments

The healing power of watching the boys of 'Jackass' do stupid things

https://www.yahoo.com/entertainment/general/article/the-healing-power-of-watching-the-boys-of-jac...
1•herbertl•4m ago•0 comments

Show HN: Tarot.free – Free Tarot Readings

https://tarot.free/
1•nadermx•6m ago•0 comments

The Hard Parts of Real-Time Voice Agent Audio

1•gokuljs•6m ago•0 comments

Binvariants: Register-Level Invaraint-Guided Fuzzing for Binaries

https://github.com/FuturesLab/Binvariants
1•matt_d•9m ago•0 comments

Env.style: Deployment environment disambiguation via favicon

https://www.env.style/
1•handfuloflight•11m ago•0 comments

Cewsco – the AI that keeps up with you

https://cewsco.com/
1•kalkalka•14m ago•1 comments

Show HN: Dedtxt is a quiet, cross-platform plain-text editor

https://dedtxt.app/
1•chrisportka•14m ago•0 comments

Moody Bible Institute breach leaves 2.3M accounts needing salvation

https://www.theregister.com/security/2026/07/06/moody-bible-institute-breach-leaves-23m-accounts-...
2•Bender•16m ago•0 comments

Brit supermarket giant triples down on facial recog to nab shoplifters

https://www.theregister.com/security/2026/07/06/brit-supermarket-giant-triples-down-on-facial-rec...
2•Bender•17m ago•1 comments

GitHub cuts short offer to burn repos on CD after mockery ensues

https://www.theregister.com/devops/2026/07/06/github-cuts-short-offer-to-burn-repos-on-cd-after-m...
1•Bender•17m ago•1 comments

PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models

https://supercomputing-system-ai-lab.github.io/projects/puzzlemoe/
1•matt_d•17m ago•0 comments

Jackrong LLM Fine-Tuning Guide

https://github.com/R6410418/Jackrong-llm-finetuning-guide
1•verdverm•19m ago•0 comments

Claude Sonnet 5: Anthropic's Most Agentic AI Model Arrives at a Reduced Price (2026)

https://lucasaguiar.xyz/en/posts/claude-sonnet-5-2026/
2•isfttr•20m ago•0 comments

How to sequence your own DNA at home

https://bradleywoolf.com/links-1/sequencing-my-own-dna-at-home
2•bilsbie•21m ago•0 comments

Supporting Ordering in Schema.org

https://blog.schema.org/2026/06/17/supporting-ordering-in-schema-org/
1•tingletech•22m ago•0 comments

Fixed-Point Theorems and the Ethics of Transparency (2025)

https://arxiv.org/abs/2509.06055
1•measurablefunc•24m ago•0 comments

Apple's agent security skills updated for Xcode 27 beta 3

https://mastodon.social/@PeteGoldsmith/116875450861776782
1•theraven•26m ago•0 comments

Show HN: GeoLens – a self-hosted catalog and map builder for spatial data

https://getgeolens.com/
1•ishiland•28m ago•0 comments

Video: Rio Rancho police arrest man accused of destroying Flock cameras

https://www.krqe.com/news/crime/video-rio-rancho-police-arrest-man-accused-of-destroying-license-...
1•evo_9•32m ago•1 comments

Show HN: Capsule – E2E encrypted file sharing

https://withcapsule.dev/
1•seanathan•34m ago•0 comments

Small AI Models Gain Traction In places with unreliable networks

https://spectrum.ieee.org/small-language-models-ai-pharmaceuticals
2•sscaryterry•35m ago•0 comments

Show HN: InstantVideos.org – short documentaries in ~30 seconds

https://instantvideos.org/
1•pw•36m ago•1 comments

Wikipedia Is Battling for the Soul of the Internet

https://www.nytimes.com/2026/07/05/business/media/wikipedia-ai-elon-musk.html
3•cdrnsf•37m ago•0 comments

Merlin the duck, a Mexico City streetside regular turned World Cup mascot

https://www.sfgate.com/sports/article/meet-merlin-the-duck-a-mexico-city-streetside-22306478.php
1•gnabgib•44m ago•0 comments

Samsar: Model-Composable 1-Shot Generative Video Harness and Software Factory

https://github.com/samsarone/samsar
1•proy24•49m ago•1 comments

Procedural Trick Before Recess Pushes EU Towards Capitulation on "Chat Control"

https://www.patrick-breyer.de/en/procedural-trick-before-summer-recess-pushes-eu-parliament-towar...
2•iamnothere•51m ago•0 comments

Show HN: LLM Thought Visualization

https://github.com/ninjahawk/Subtext
1•ninjahawk1•1h ago•0 comments

NSA and IETF: Fairness

https://blog.cr.yp.to/20260706-fairness.html
2•WatchDog•1h ago•0 comments
Open in hackernews

Ternlight – 7 MB embedding model that runs in browser (WASM)

https://ternlight-demo.vercel.app/
43•soycaporal•1h ago

Comments

soycaporal•1h ago
Hobby project, I wanted to "ship a useful model in a web browser". so I distilled a small sentence encoder from MiniLM with ternary quantization-aware training. Also wrote the inference engine from scratch and shipped in Rust → WASM SIMD.

It's an embeddings model, not an LLM: text goes in, a 384-dim vector comes out, and cosine similarity between two vectors tells you how related the texts are — regardless of shared words ("reset my password" ↔ "I forgot my password" → 0.88). Used for semantic search, FAQ/intent matching, and clustering. Running it on-device means search-as-you-type semantic search is performant with no API dependencies.

Demo (2k React docs, fully on-device): https://ternlight-demo.vercel.app

Two tiers on npm: - @ternlight/base (7 MB, ~5 ms/embed, more capable embedings) - @ternlight/mini (5 MB wire, ~2.5 ms/embed).

Bundled for Node and browsers.

Repo - see technical details (MIT, training pipeline included): https://github.com/soycaporal/ternlight

Curious if this is something useful, what are the use cases for on-device embeddings.

fellowniusmonk•39m ago
Awesome! Besides size, how does this compare to gte-small?
soycaporal•32m ago
gte-small outscores all-MiniLM-L6 on MTEB (~61 vs ~56 avg per the GTE paper). MiniLM is ternlight's teacher (ternlight holds 0.84 Spearman fidelity to teacher). I haven't run a head-to-head yet; STS-B/MTEB numbers are on the roadmap. Also on the roadmap is to distill gte-small as teacher.
aetherspawn•37m ago
Can the 30 second embedding time be done beforehand and sent to the browser?

Inference is nice and quick after that.

soycaporal•35m ago
yes, you could run a 1 time indexing run on the server side, and just ship the embeddings to frontend
dirteater_•34m ago
This is cool!

but also maybe you could put a button on the landing page to trigger the demo because it's a bit startling to hear my fans go crazy when opening a webpage.

esafak•27m ago
What we need is a W3C LLM API.
yesidoagree•17m ago
If it was like Math (Math.round, Math.PI, etc.) it could be Language, as in:

    Language.complete('the quick brown fox jumped over the lazy') 
and maybe even static methods on Image

    Image.generate('a spaceship flying toward a planet')
soycaporal•9m ago
I think standardizing the runtime is pretty effective, it then open up portability
wazzup_im•18m ago
I added an offline search engine to app.wazzup.im/search (no login or payment required).

First search downloads the model from the internet and subsequent runs are from the cache.

The model is very small so it's not the best for everything but it's good for basic math and coding.

Give it a try.

CobrastanJorji•9m ago
Great, now my websites are gonna push entire LLMs onto my browser in order to use my CPU to make inferences about my shopping habits or whatever.