frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Show HN: Prompt injection detector beats ProtectAI by 19% accuracy, 8.9x smaller

https://huggingface.co/hlyn/prompt-injection-judge-deberta-70m
1•Karan047•1m ago•0 comments

The Downfall and Enshittification of Microsoft in 2026

https://caio.ca/blog/the-downfall-and-enshittification-of-microsoft.html
1•birdculture•2m ago•0 comments

Nuclear Energy Heresy with Daniel Chen [video]

https://www.youtube.com/watch?v=HZq_V-UKLj0
1•leonidasrup•3m ago•1 comments

Show HN: Access OpenClaw's workspace files from anywhere and any device

https://github.com/RageDotNet/openclaw-webdav
1•gregatragenet3•3m ago•0 comments

Show HN: MCP-fence – MCP firewall I built and tried to break (6 audit rounds)

https://www.npmjs.com/package/mcp-fence
1•yjcho9317•4m ago•0 comments

Our servers are experiencing high traffic please try again in a minute

https://discuss.ai.google.dev/t/how-to-resolve-this-issue-our-servers-are-experiencing-high-traff...
1•maarut•5m ago•0 comments

Show HN: I've build a hermes agent helper website

https://hermes-agent.us
1•mixfox•6m ago•0 comments

The Oldschool PC Font Pack

https://int10h.org/oldschool-pc-fonts/
1•petercooper•7m ago•0 comments

Are file systems all you need?

https://onyx.app/blog/file-search-vs-hybrid-search
1•Weves•8m ago•0 comments

Cyclotron: The Streaming Multiprocessor Abstraction Is Broken [pdf]

https://capra.cs.cornell.edu/latte26/paper/latte26-final28.pdf
1•matt_d•10m ago•0 comments

The Worst of Us

https://www.ianbetteridge.com/the-worst-of-us/
1•speckx•11m ago•0 comments

Wamp, WinAmp style native audio player for macOS

https://github.com/wishval/wamp
1•vnorilo•13m ago•0 comments

Easy Management

https://easy-manage-biz.com
1•charlmarajh•15m ago•0 comments

Tom Brady becomes 'chief wellness officer' at GLP-1 weight-loss shot company

https://www.independent.co.uk/news/world/americas/tom-brady-emed-weightloss-shot-company-b2899015...
2•randycupertino•17m ago•0 comments

AI doesn't know how to interact with touchscreens

https://blog.allada.com/give-an-llm-an-api-and-itll-thrive-give-it-a-touchscreen-and-it-struggles/
1•allada•17m ago•0 comments

Juan Benet Podcast Episode 1: Max Hodak, Founder and CEO of Science Corp

https://www.juanbenetpodcast.com/p/max-hodak-restoring-sight-growing
1•nettynol•18m ago•0 comments

Show HN: My Hyperliquid Trading Terminal

https://www.aulico.com
1•rovinarov•18m ago•0 comments

Show HN: TUI-use: Let AI agents control interactive terminal programs

https://github.com/onesuper/tui-use
3•dreamsome•18m ago•1 comments

Show HN: I bootstrapped a foundational text-to-speech model from scratch

https://tontaube.ai/
1•vincenttjona•19m ago•0 comments

Space Propulsion Made Easy: Eat Beans

https://www.npr.org/sections/krulwich/2010/09/16/129908529/space-propulsion-made-easy-eat-beans
1•thunderbong•19m ago•0 comments

Show HN: One click to deploy AI platforms and other open source tools

https://hyp.app
2•dashtio•20m ago•0 comments

Pgfmt – a PostgreSQL specific SQL formatter

https://github.com/gmr/pgfmt
2•whalesalad•22m ago•0 comments

Akamai: AI bot traffic surged 300% in 2025, hitting publishers hardest

https://www.akamai.com/resources/state-of-the-internet/publishing-ai-botnet-report
1•speckx•22m ago•0 comments

AI Experience Engineering

https://raqibul.com/writing/ai-experience-engineering
1•raqib-hayder•23m ago•0 comments

Improving LLM citation accuracy with agentic highlighting tools for local files

https://old.reddit.com/r/LLMDevs/comments/1sfd6ga/annotation_update_just_pushed_improved_note/
1•ieuanking•24m ago•0 comments

Next Grok model training with 10T parameter model

https://twitter.com/i/status/2041754402239975479
3•ramshanker•24m ago•2 comments

Bonsai 8B: a 1-bit LLM that fits in 1.15GB

https://firethering.com/bonsai-8b-1bit-llm/
4•steveharing1•25m ago•1 comments

AI agents as CRDT peers – building collaborative AI with Yjs

https://electric-sql.com/blog/2026/04/08/ai-agents-as-crdt-peers-with-yjs
2•samwillis•26m ago•0 comments

Confidential Inference

https://confidentialinference.net/
1•rzk•26m ago•0 comments

OneLivePage

https://www.onelive.page/
1•erii•26m ago•1 comments
Open in hackernews

Show HN: routr - a fast local replacement for DuckDuckGo bangs

https://t128n.github.io/routr/
2•trbndev•11mo ago

Comments

trbndev•11mo ago
routr is a lightweight, offline search router inspired by DuckDuckGo's bangs. It's fully browser-based (no server requests needed other than the initial one) and routes queries like `!gh tokio` to GitHub, `!yt` to YouTube, `!hn` to Hacker News, etc.

What sets it apart:

- Runs entirely locally (no network, no tracking)

- Lets you define your own bangs and destinations

- Optionally rewrites queries for better precision (with AI)

Example:

Typing `article about german war in the stone age !!g` becomes: `site:en.wikipedia.org "german" AND "war" AND "stone age"`

More on query rewriting and performance:

https://tl128n.github.io/writings/2025-05-02_optimizing_sear...

Leftium•11mo ago
That link is broken. Correct link: https://t128n.github.io/writings/2025-05-02_optimizing_searc...
trbndev•11mo ago
Ah damn what a bummer that I can‘t edit the original comment anymore. Thank you for pointing it out… didn‘t even see that extra `l` there in the url
Leftium•11mo ago
> Typing `article about german war in the stone age !!g` becomes: `site:en.wikipedia.org "german" AND "war" AND "stone age"`

I wonder how much of that could be done without AI, just (locally) using https://www.npmjs.com/package/compromise

---

Also I couldn't find your source for bang data; Kagi maintains an updated list: https://github.com/kagisearch/bangs

I'm working on an enhanced version of bangs. Since there are so many "dead" bangs, I'm thinking of using a bang whitelist and suggestions from the full list if not found in the whitelist.

trbndev•11mo ago
What an interesting library. Thanks for mentioning it here… will definitely look into it!

—-

The bangs are at

https://github.com/t128n/routr/blob/main/src/sw/routes.ts

Mostly ripped from duckduckgo.com/bang.js and then pretty-formatted with Prettier… with some sane-defaults added like a bang for ChatGPT or T3 Chat

Leftium•11mo ago
Related project (also works locally):

- https://unduck.link/

- https://github.com/t3dotgg/unduck

- https://youtu.be/_DnNzRaBWUU

---

Note unduck works without service workers: https://github.com/t3dotgg/unduck/blob/91a66c25de8b0e7ac5ba5...

Is there any advantage to using service workers?

edit: Ah... unduck doesn't work from the browser search bar.

trbndev•11mo ago
Unduck actually also uses service workers to cache the app locally

https://github.com/t3dotgg/unduck/blob/91a66c25de8b0e7ac5ba5...

—-

> Is there any advantage to using service workers?

With service workers we can intercept the `fetch` request, before the index.html gets rendered/loaded. One should (in theory) be redirected some micro-seconds faster than doing the redirect in the index.html js