This is a personal project I&#x27;ve been working on for a long time now.I stumbled upon the color quantization problem while doing something related for work. I then found an interesting paper for which I could find no implementations online, and the thing went from &quot;let&#x27;s implement this paper&quot; to getting pretty obsessed with the whole thing.It&#x27;s at an early, eaaaarly stage. There&#x27;s a lot of work to be done, and it&#x27;s a memory hog, but generally speaking works quite well, and the output is for the most part very high quality, so I&#x27;m happy to share it as beta.

Show HN: High End Color Quantizer

Companies building software in the US were hit hard a few years ago when the tax code stopped allowing deduction of software dev expenses. Now they have to be amortized over several years.HN has had many discussions about this, including The time bomb in the tax code that&#x27;s fueling mass tech layoffs - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44180533">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44180533</a> - (927 comments) a few days ago. Other threads are listed at <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44203869">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44203869</a>.There&#x27;s currently a major effort to get this change reversed. One of the people working on it is YC&#x27;s Luther Lowe (<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;user?id=itsluther">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;user?id=itsluther</a>). Luther has been organizing YC alumni to urge lawmakers to support this reversal. I asked him if we could do that on Hacker News too. He said yes—hence this thread :)If you&#x27;re a US taxpayer and if you agree that software dev expenses should be deductible like they used to be, please sign this letter to the relevant committee members: <a href="https:&#x2F;&#x2F;docs.google.com&#x2F;forms&#x2F;d&#x2F;1DkRGeef2e_tU2xf3TyEyd2JLlmZH8sGs6hhuVkokgpw&#x2F;viewform?edit_requested=true" rel="nofollow">https:&#x2F;&#x2F;docs.google.com&#x2F;forms&#x2F;d&#x2F;1DkRGeef2e_tU2xf3TyEyd2JLlmZ...</a>.(If you&#x27;re not a US person, please don&#x27;t sign the letter, since lawmakers will only listen to feedback from taxpayers and we don&#x27;t want to dilute the signal.)I&#x27;m sure not everyone here agrees with us—HN is a big community, there&#x27;s no total agreement on anything—but this issue has as close to a community consensus as HN gets, so I think it makes sense to add our voices too.Luther will be around to answer questions and hopefully HN can contribute to getting this done!

Tell HN: Help restore the tax deduction for software dev in the US (Section 174)

Slides: <a href="https:&#x2F;&#x2F;docs.google.com&#x2F;presentation&#x2F;d&#x2F;1f_vKBJMaD68ifBO2j83lBly9Zdk-2bsvj_DIHXxvcuk" rel="nofollow">https:&#x2F;&#x2F;docs.google.com&#x2F;presentation&#x2F;d&#x2F;1f_vKBJMaD68ifBO2j83l...</a>

Faster, easier 2D vector rendering [video]

Containerization is a Swift package for running Linux containers on macOS

Apple announces Foundation Models and Containerization frameworks, etc

'Proof' Review: Finding Truth in Numbers

Animate a mesh across a sphere's surface

Our Website: <a href="https:&#x2F;&#x2F;onlook.com&#x2F;">https:&#x2F;&#x2F;onlook.com&#x2F;</a>
Our Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;onlook-dev&#x2F;onlook">https:&#x2F;&#x2F;github.com&#x2F;onlook-dev&#x2F;onlook</a>Apply here: <a href="https:&#x2F;&#x2F;www.ycombinator.com&#x2F;companies&#x2F;onlook&#x2F;jobs&#x2F;e4gHv1n-founding-engineer-fullstack-ai">https:&#x2F;&#x2F;www.ycombinator.com&#x2F;companies&#x2F;onlook&#x2F;jobs&#x2F;e4gHv1n-fo...</a>or email daniel [at] onlook.comWe&#x27;re Onlook, the Cursor for Designers.We made an open-source (18k stars) visual editor for code that lets anyone craft changes to codebases using AI on an infinite canvas. We just had a blowout launch of our latest product on HN, 10xing our traffic and making us the #1 trending Github repo in the world for 3 days in a row.We’re looking to bring on Onlook’s first Founding Engineer to work with us in our incredibly cool office in downtown San Fransisco. You’re a full-stack engineer with experience in AI who is ready to build a generational design-led company.The RoleThis role requires autonomy – you’ll lay the foundation for an uncompromising open-source visual IDE that tens-of-thousands of engineers and designers around the world love, and help craft our AI.The most important things we look for:• Olympic-level dedication – you want to be the best in the world at what you do.
• Ownership – you like autonomy and control over the destiny of the company.
• Speed – you’re comfortable shipping and iterating quickly with feedback.
• Craft – you’re opinionated and are willing to defend your opinions.Ideally, you:• Are looking for a fast-paced, intense startup environment.
• Are willing to put in long hours and go the extra mile.
• Are comfortable with any part of the stack, front-end, back-end, or database.
• Believe in open source and are ok with your work being very public.If you’d like to stand out, please share a project or piece of work that you’re most proud of. We love seeing people’s work.Working at OnlookWe’re based in San Francisco, the startup capital of the world, and work closely together in person. We have a very cool office near South Park – it has ultra tall ceilings, tons of natural light, great restaurants and groceries stores just a block away, a full-kitchen, and even a shower. Employees get bespoke black bomber jackets with the Onlook crest embroidered on them, their own standing desk, monitor, budget for tech, and more swag.You must be willing to move to San Francisco if you are not already here.Onlook’s values are a reflection of our founding team and the cultures of the companies we’ve come from. Speed, technical excellence, brand, and attention to detail are what uniquely set Onlook apart.

Onlook (YC W25) Is Hiring an engineer in SF

Why does C++ think my class is copy-constructible when it can't be?

Successful people set constraints rather than chasing goals

Sly Stone has died

Micrographia (1665) [pdf]

Implementing DOES> in Forth, the entire reason I started this mess

Encapsulated Co–Ni alloy boosts high-temperature CO2 electroreduction

Why agents are bad pair programmers

Hello HN!Showing off the first version of Munal OS, an experimental operating system I have been writing in Rust on and off for the past few years.<a href="https:&#x2F;&#x2F;github.com&#x2F;Askannz&#x2F;munal-os">https:&#x2F;&#x2F;github.com&#x2F;Askannz&#x2F;munal-os</a>It&#x27;s an unikernel design that is compiled as a single EFI binary and does not use virtual address spaces for process isolation. Instead, applications are compiled to WASM and run inside of an embedded WASM engine.Other features:* Fully graphical interface in HD resolution with mouse and keyboard support* Desktop shell with window manager and contextual radial menus* PCI and VirtIO drivers* Ethernet and TCP stack* Customizable UI toolkit providing various widgets, responsive layouts and flexible text rendering* Embedded selection of applications including:<pre><code> * A web browser supporting DNS, HTTPS and very basic HTML

 * A text editor

 * A Python terminal
</code></pre>
Checkout the README for the technical breakdown.Demo video: <a href="https:&#x2F;&#x2F;streamable.com&#x2F;5xqjcf" rel="nofollow">https:&#x2F;&#x2F;streamable.com&#x2F;5xqjcf</a>

Show HN: Munal OS: a graphical experimental OS with WASM sandboxing

Apple introduces a universal design across platforms

A Primer on Molecular Dynamics

Go is a good fit for agents

Hey HN! We&#x27;re Shreyash and Bhavnick. We&#x27;re building Chonkie (<a href="https:&#x2F;&#x2F;chonkie.ai">https:&#x2F;&#x2F;chonkie.ai</a>), an open-source library for chunking and embedding data.Python: <a href="https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie">https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie</a>TypeScript: <a href="https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie-ts">https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie-ts</a>Here&#x27;s a video showing our code chunker: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;Xclkh6bU1P0" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;Xclkh6bU1P0</a>.Bhavnick and I have been building personal projects with LLMs for a few years. For much of this time, we found ourselves writing our own chunking logic to support RAG applications. We often hesitated to use existing libraries because they either had only basic features or felt too bloated (some are 80MB+).We built Chonkie to be lightweight, fast, extensible, and easy. The space is evolving rapidly, and we wanted Chonkie to be able to quickly support the newest strategies. We currently support: Token Chunking, Sentence Chunking, Recursive Chunking, Semantic Chunking, plus:- Semantic Double Pass Chunking: Chunks text semantically first, then merges closely related chunks.- Code Chunking: Chunks code files by creating an AST and finding ideal split points.- Late Chunking: Based on the paper (<a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2409.04701" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2409.04701</a>), where chunk embeddings are derived from embedding a longer document.- Slumber Chunking: Based on the &quot;Lumber Chunking&quot; paper (<a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2406.17526" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2406.17526</a>). It uses recursive chunking, then an LLM verifies split points, aiming for high-quality chunks with reduced token usage and LLM costs.You can see how Chonkie compares to LangChain and LlamaIndex in our benchmarks: <a href="https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie&#x2F;blob&#x2F;main&#x2F;BENCHMARKS.md">https:&#x2F;&#x2F;github.com&#x2F;chonkie-inc&#x2F;chonkie&#x2F;blob&#x2F;main&#x2F;BENCHMARKS....</a>Some technical details about the Chonkie package: - ~15MB default install vs. ~80-170MB for some alternatives. - Up to 33x faster token chunking compared to LangChain and LlamaIndex in our tests. - Works with major tokenizers (transformers, tokenizers, tiktoken). - Zero external dependencies for basic functionality. - Implements aggressive caching and precomputation. - Uses running mean pooling for efficient semantic chunking. - Modular dependency system (install only what you need).In addition to chunking, Chonkie also provides an easy way to create embeddings. For supported providers (SentenceTransformer, Model2Vec, OpenAI), you just specify the model name as a string. You can also create custom embedding handlers for other providers.RAG is still the most common use case currently. However, Chonkie makes chunks that are optimized for creating high quality embeddings and vector retrieval, so it is not really tied to the &quot;generation&quot; part of RAG. In fact, We&#x27;re seeing more and more people use Chonkie for implementing semantic search and&#x2F;or setting context for agents.We are currently focused on building integrations to simplify the retrieval process. We&#x27;ve created &quot;handshakes&quot; – thin functions that interact with vector DBs like pgVector, Chroma, TurboPuffer, and Qdrant, allowing you to interact with storage easily. If there&#x27;s an integration you&#x27;d like to see (vector DB or otherwise), please let us know.We also offer hosted and on-premise versions with OCR, extra metadata, all embedding providers, and managed vector databases for teams that want a fully managed pipeline. If you&#x27;re interested, reach out at shreyash@chonkie.ai or book a demo: <a href="https:&#x2F;&#x2F;cal.com&#x2F;shreyashn&#x2F;chonkie-demo" rel="nofollow">https:&#x2F;&#x2F;cal.com&#x2F;shreyashn&#x2F;chonkie-demo</a>.We&#x27;re eager to hear your feedback and comments! Thanks!

Launch HN: Chonkie (YC X25) – Open-Source Library for Advanced Chunking

Scientific Papers: Innovation or Imitation?

Marines being mobilized in response to LA protests

AI Saved My Company from a 2-Year Litigation Nightmare

Doctors could hack the nervous system with ultrasound

Bruteforcing the phone number of any Google user

Show HN: A MCP server and client implementing the latest spec

Container: Apple's Linux-Container Runtime

What methylene blue can (and can’t) do for the brain

Most feedback tools are built like people actually want to report bugs. They don’t. Unless you make it dead-simple, or better yet - a little fun.After shipping a few SaaS products, I noticed a pattern: Bugs? Yes. Bug reports? No.Not because users didn’t care but because reporting bugs is usually a terrible experience.Most tools want users to:* Fill out a long form* Enter their email* Describe a bug they barely understand* Maybe sign in or create an account* Then maybe submit itLet’s be real: no one’s doing that. Especially not someone just trying to use your product.So I built Bugdrop.app - It’s a little draggable bug icon that users can drop right on the issue, type a quick note, and they’re done. No logins. No forms. Just context-rich feedback that your team can actually use — with screenshots, browser info, even console logs if they hit an error.And weirdly? People actually use it. Even non-technical users click it just because &quot;the little bug looked fun.&quot;I didn’t want to build another &quot;feedback suite&quot;. I just wanted something lightweight, fast, and so stupidly simple that people actually report stuff. If you&#x27;ve ever had a user say “something’s broken” and then ghost you forever, you probably get where I’m coming from.What I’m most proud of? People are actually using it. And their users? They’re actually reporting stuff. Even non-technical ones.Would love to hear if you’ve faced similar problems, and if this feels like something that would’ve helped in your own projects. Not trying to sell you anything — just sharing something I built to scratch my own itch.

Show HN: Most users won't report bugs unless you make it stupidly easy

Discrete Mathematics: An Open Introduction [pdf]