frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Show HN: LocalGPT – A local-first AI assistant in Rust with persistent memory

https://github.com/localgpt-app/localgpt
125•yi_wang•4h ago•35 comments

Haskell for all: Beyond agentic coding

https://haskellforall.com/2026/02/beyond-agentic-coding
53•RebelPotato•3h ago•10 comments

SectorC: A C Compiler in 512 bytes (2023)

https://xorvoid.com/sectorc.html
247•valyala•12h ago•49 comments

Speed up responses with fast mode

https://code.claude.com/docs/en/fast-mode
165•surprisetalk•11h ago•155 comments

Software factories and the agentic moment

https://factory.strongdm.ai/
195•mellosouls•14h ago•350 comments

Total surface area required to fuel the world with solar (2009)

https://landartgenerator.org/blagi/archives/127
18•robtherobber•4d ago•5 comments

Brookhaven Lab's RHIC concludes 25-year run with final collisions

https://www.hpcwire.com/off-the-wire/brookhaven-labs-rhic-concludes-25-year-run-with-final-collis...
73•gnufx•10h ago•59 comments

LLMs as the new high level language

https://federicopereiro.com/llm-high/
62•swah•4d ago•113 comments

Hoot: Scheme on WebAssembly

https://www.spritely.institute/hoot/
180•AlexeyBrin•17h ago•35 comments

Stories from 25 Years of Software Development

https://susam.net/twenty-five-years-of-computing.html
171•vinhnx•15h ago•17 comments

Vocal Guide – belt sing without killing yourself

https://jesperordrup.github.io/vocal-guide/
319•jesperordrup•22h ago•97 comments

First Proof

https://arxiv.org/abs/2602.05192
134•samasblack•14h ago•77 comments

Vouch

https://twitter.com/mitchellh/status/2020252149117313349
62•chwtutha•2h ago•10 comments

Show HN: I saw this cool navigation reveal, so I made a simple HTML+CSS version

https://github.com/Momciloo/fun-with-clip-path
82•momciloo•12h ago•16 comments

Wood Gas Vehicles: Firewood in the Fuel Tank (2010)

https://solar.lowtechmagazine.com/2010/01/wood-gas-vehicles-firewood-in-the-fuel-tank/
31•Rygian•2d ago•7 comments

Why there is no official statement from Substack about the data leak

https://techcrunch.com/2026/02/05/substack-confirms-data-breach-affecting-email-addresses-and-pho...
14•witnessme•1h ago•4 comments

Al Lowe on model trains, funny deaths and working with Disney

https://spillhistorie.no/2026/02/06/interview-with-sierra-veteran-al-lowe/
104•thelok•14h ago•22 comments

Show HN: A luma dependent chroma compression algorithm (image compression)

https://www.bitsnbites.eu/a-spatial-domain-variable-block-size-luma-dependent-chroma-compression-...
40•mbitsnbites•3d ago•4 comments

FDA intends to take action against non-FDA-approved GLP-1 drugs

https://www.fda.gov/news-events/press-announcements/fda-intends-take-action-against-non-fda-appro...
112•randycupertino•7h ago•233 comments

Start all of your commands with a comma (2009)

https://rhodesmill.org/brandon/2009/commands-with-comma/
577•theblazehen•3d ago•208 comments

Homeland Security Spying on Reddit Users

https://www.kenklippenstein.com/p/homeland-security-spies-on-reddit
59•duxup•1h ago•13 comments

The AI boom is causing shortages everywhere else

https://www.washingtonpost.com/technology/2026/02/07/ai-spending-economy-shortages/
304•1vuio0pswjnm7•18h ago•482 comments

I write games in C (yes, C) (2016)

https://jonathanwhiting.com/writing/blog/games_in_c/
189•valyala•12h ago•173 comments

Microsoft account bugs locked me out of Notepad – Are thin clients ruining PCs?

https://www.windowscentral.com/microsoft/windows-11/windows-locked-me-out-of-notepad-is-the-thin-...
144•josephcsible•10h ago•178 comments

Selection rather than prediction

https://voratiq.com/blog/selection-rather-than-prediction/
34•languid-photic•4d ago•15 comments

Learning from context is harder than we thought

https://hy.tencent.com/research/100025?langVersion=en
233•limoce•4d ago•125 comments

OpenCiv3: Open-source, cross-platform reimagining of Civilization III

https://openciv3.org/
904•klaussilveira•1d ago•276 comments

Where did all the starships go?

https://www.datawrapper.de/blog/science-fiction-decline
150•speckx•4d ago•235 comments

Show HN: Look Ma, No Linux: Shell, App Installer, Vi, Cc on ESP32-S3 / BreezyBox

https://github.com/valdanylchuk/breezydemo
303•isitcontent•1d ago•39 comments

Reinforcement Learning from Human Feedback

https://rlhfbook.com/
118•onurkanbkrc•16h ago•5 comments
Open in hackernews

Show HN: Cursor for Userscripts

https://github.com/chebykinn/browser-code
56•mifydev•1w ago
I’ve been experimenting with embedding an Claude Code/Cursor-style coding agent directly into the browser.

At a high level, the agent generates and maintains userscripts and CSS that are re-applied on page load. Rather than just editing DOM via JS in console the agent is treating the page, and the DOM as a file.

The models are often trained in RL sandboxes with full access to the filesystem and bash, so they are really good at using it. So to make the agent behave well, I've simulated this environment.

The whole state of a page and scripts is implemented as a virtual filesystem hacked on top of browser.local storage. URL is mapped to directories, and the agent starts inside this directory. It has the tools to read/edit files, grep around and a fake bash command that is just used for running scripts and executing JS code.

I've tested only with Opus 4.5 so far, and it works pretty reliably. The state of the file system can be synced to the real filesystem, although because Firefox doesn't support Filesystem API, you need to manually import the fs contents first.

This agent is really useful for extracting things to CSV, but it's also can be used for fun.

Demo: https://x.com/ichebykin/status/2015686974439608607

Comments

Esophagus4•1w ago
Awesome! So the agent has access to the DOM/JS running in the browser?

That’s one of my biggest headaches writing user scripts currently: I write the script in an IDE with Claude then copy it to the browser / manually test it in the browser, then copy the results back to Claude or tell it what went wrong.

Looking forward to trying this.

mifydev•1w ago
Yup, full access to DOM! Still needs a lot of optimizations, but the trick is that the agent reads the DOM as file, so it can grep parts of it naturally.
Zekio•1w ago
to my knowledge all the major userscript extensions, at least allow watching for file changes so you don't have to copy it manually, so you can just refresh the page to test
Akranazon•1w ago
I'm working on a version of this, https://www.quillmonkey.com/ so you got ahead of me. I imagine there are many versions of this coming. Interesting what set of tools you went with.
mifydev•1w ago
Oh that's cool! I've just used wxt to pack extension for firefox and chrome and just used typescript and plain anthropic api. My goal is to make this run fully inside the browser, without any helper binaries, like I've seen with others.
Akranazon•1w ago
Your project seems pretty close to where mine was a couple weeks ago, where I was focused on a BYOK solution (user-entered Anthropic API key). I saw there was another similar extension already released in the app store (RobotMonkey) which hooks up to their own backend service, and offers subscriptions. For my project, I think that's the right way to go.

It's funny what details about our designs are similar through accident. And what other things are completely different. I can show you my design potentially.

Representing websites in a virtual filesystem is creative and definitely makes it easier for the agent to collect information about the page. But I'm confused between the `Bash` and the `Edit` tools. It seems like one uses the chrome executeScript API, and the other updates the file system. But if it's just doing file writes, are those edits visible in the browser, and persistent across sessions?

mifydev•1w ago
Backend service is definitely way to go if you want to serve models for the user.

So Bash and Edit tools are a bit weird, Bash tool is essentially JS execution, and Edit tool automatically generates a script that performs the edits on the page. These tools are needed for the model to explore the page, whatever it does at the end it creates a separate script that will be applied on the page load.

Akranazon•1w ago
Oh neat. So the edit tool is like a convenient API/wrapper for it to eg add HTML to some element? I guess theoretically that can also be achieved through Bash as well, but the tool fits closer to an interface we know exiting agents are good at.
rahimnathwani•1w ago
It would be cool if you could make this work with Gemini Flash, with keys from AI Studio. I imagine that would expand the set of people who would try it out, because they could use 'free' keys and not worry about unexpected bills.
mifydev•1w ago
That's a good point, I'll add support for other models shortly.
_false•1w ago
Love the decision to edit DOM directly. More LLM tools should carefully consider their training environments instead of treating LLMs like AI Gods.
cxr•1w ago
<https://web.archive.org/web/20120601101804/https://blog.mozi...>

<https://sites.coecis.cornell.edu/leshed/files/2015/12/coscri...>

<https://ofb.net/~tlau//research/papers/p97-lin.pdf>

<https://github.com/jeffnichols-ibm/coscripter-extension>

mrandish•1w ago
Cool! Just this week I've been playing around using web chatbots to write userscript. So far, I've just been cutting/pasting code from ViolentMonkey's basic edit window - which is a clunky workflow and a real editor like VSCode would be great. Even though the script is quite short, I've already found I definitely need auto-rollback due to AI regressions. So this sounds very useful!

I use Firefox (on Win) but I guess it might be worth using Chrome just for write/test iteration and then copy over to FF. Looking just now it appears ViolentMonkey on Chrome died in the Manifest V3 apocalypse but is still on Edge.

onurkanbkrc•1w ago
Installation should be easier. Why do I need to build the extension or download a release instead of installing it directly from the Chrome Web Store?
mifydev•1w ago
I've submitted it to web store, but I'm sure that the review will be very long. The extension requires a lot of permissions, with this kinds of things personally I'll trust more if i can build from source.