Slack has raised our charges by $195k per year

Nvidia buys $5B in Intel

KDE is now my favorite desktop

Want to piss off your IT department? Are the links not malicious looking enough?

Apple: SSH and FileVault

Grief gets an expiration date, just like us

Learn Your Way: Reimagining Textbooks with Generative AI

Flipper Zero Geiger Counter

This map is not upside down

TernFS – An exabyte scale, multi-region distributed filesystem

The Sagrada Família takes its final shape

Luau – Fast, small, safe, gradually typed scripting language derived from Lua

U.S. already has the critical minerals it needs, according to new analysis

David Lynch LA House

I built this yesterday to help understand papers I&#x27;m interested in. It&#x27;s using the gemini 2.5 flash lite model, but you can run it yourself[1] and switch to 2.5 pro for better results.Happy to answer any questions or take suggestions on how I can improve it!1. <a href="https:&#x2F;&#x2F;github.com&#x2F;montanaflynn&#x2F;asxiv" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;montanaflynn&#x2F;asxiv</a>

Show HN: Asxiv.org – Ask ArXiv papers questions through chat

Gemini in Chrome

Hey HN, Henry &amp; Roman here, we are building Cactus (<a href="https:&#x2F;&#x2F;cactuscompute.com&#x2F;">https:&#x2F;&#x2F;cactuscompute.com&#x2F;</a>), an AI inference engine specifically designed for phones.We&#x27;re seeing a major push towards on-device AI, and for good reason: on-device AI decreases latency from &gt;1sec to &lt;100ms, guarantees privacy by default, works offline, and doesn&#x27;t rack up a massive API bill at scale.Also, tools and agentic designs make small models really good beyond benchmarks. This has been corroborated by other papers like <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2506.02153" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2506.02153</a>, and we see model companies like DeepMind aggressively going into smaller models with Gemma3 270m and 308m. We found Qwen3 600m to be great at tool calls for instance.Some frameworks already try to solve this but in my previous job, they struggled in production compared to research and playgrounds:- They optimise for modern devices but 70% of phones today are low-mid budget.- Bloated app bundle sizes and battery drain are serious concerns for users.- Phone GPU battery drain is unacceptable, NPUs are preferred, but few phones have those for now.- Some are platform-specific, requiring different models and workflows for different operating systems.At Cactus, we’ve written kernels and inference engine for running AI locally on any phone, from the ground-up.Cactus is designed for mobile devices and their constraints. Every design choice like energy efficiency, accelerator support, quantization levels, supported models, weight format, and context management were determined by this. We also provide minimalist SDKs for app developers to build agentic workflows in 2-5 lines of code.We made a Show HN post when we started the project to get the community&#x27;s thoughts (<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44524544">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=44524544</a>). Based on your feedback, we built Cactus bottom-up to solve those problems, and are launching the Cactus Kernels, Cactus Graph and Cactus Engine, all designed for phones and tiny devices.CPU benchmarks for Qwen3-600m-INT8 :- 16-20 toks&#x2F;sec on Pixel 6a &#x2F; Galaxy S21 &#x2F; iPhone 11 Pro- 50-70 toks&#x2F;sec on Pixel 9 &#x2F; Galaxy S25 &#x2F; iPhone 16.- Time-to-first-token is as low as 50ms depending on prompt size.On NPUs, we see Qwen3-4B-INT4 run at 21 toks&#x2F;sec.We are open-source (<a href="https:&#x2F;&#x2F;github.com&#x2F;cactus-compute&#x2F;cactus" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;cactus-compute&#x2F;cactus</a>). Cactus is free for hobbyists and personal projects, with a paid license required for commercial use.We have a demo app on the App Store at <a href="https:&#x2F;&#x2F;apps.apple.com&#x2F;gb&#x2F;app&#x2F;cactus-chat&#x2F;id6744444212">https:&#x2F;&#x2F;apps.apple.com&#x2F;gb&#x2F;app&#x2F;cactus-chat&#x2F;id6744444212</a> and on Google Play at <a href="https:&#x2F;&#x2F;play.google.com&#x2F;store&#x2F;apps&#x2F;details?id=com.rshemetsubuser.myapp">https:&#x2F;&#x2F;play.google.com&#x2F;store&#x2F;apps&#x2F;details?id=com.rshemetsub...</a>.In addition, there are numerous apps using Cactus in production, including AnythingLLM (<a href="https:&#x2F;&#x2F;anythingllm.com&#x2F;mobile" rel="nofollow">https:&#x2F;&#x2F;anythingllm.com&#x2F;mobile</a>) and KinAI (<a href="https:&#x2F;&#x2F;mykin.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;mykin.ai&#x2F;</a>). Collectively they run over 500k weekly inference tasks in production.While Cactus can be used for all Apple devices including Macbooks due to their design, for computers&#x2F;AMD&#x2F;Intel&#x2F;Nvidia generally, please use HuggingFace, Llama.cpp, Ollama, vLLM, MLX. They&#x27;re built for those, support x86, and are all great!Thanks again, please share your thoughts, we’re keen to understand your views.

Launch HN: Cactus (YC S25) – AI inference on smartphones

OpenTelemetry collector: What it is, when you need it, and when you don't

AI tools are making the world look weird

Rupert's snub cube and other Math Holes