WebMCP is available for early preview

https://developer.chrome.com/static/blog/webmcp-epp/image/cover.png
WebMCP provides a standard way for websites to expose structured tools, enabling AI agents to interact with them quickly and precisely. This allows agents to perform tasks like booking flights, filing support tickets, and navigating complex data with increased speed and reliability.

If AI writes code, should the session be part of the commit?

https://opengraph.githubassets.com/ce473859d259c8584da9e721e5bcc42bc6db44411c0ee96e44b1daee7f813aab/mandel-macaque/memento
git-memento records AI coding sessions as Git notes, allowing for auditing and amendment of commits. It supports multiple sessions and can sync notes with a remote repository.

Show HN: Timber – Ollama for classical ML models, 336x faster than Python

https://opengraph.githubassets.com/fb58d0f2e35d2f8b148e675034e8357d16a0f7ebe943c0fc1eb900465c604efa/kossisoroyce/timber
Timber compiles trained models into native C and serves them over a local HTTP API. It supports models like XGBoost and CatBoost, and can auto-detect model types.

Little Free Library

https://littlefreelibrary.org/wp-content/uploads/2022/05/Umi-Vaughan-51773-Oakland-CA-1200x901.jpg
This website uses cookies to enhance user experience and understand visitor behavior. Enabling cookies helps improve the website and serve visitors better.

Ghostty – Terminal Emulator

https://ghostty.org/social-share-card.jpg
Ghostty is a fast, feature-rich, and cross-platform terminal emulator that uses platform-native UI and GPU acceleration.

Tove Jansson's criticized illustrations of The Hobbit (2023)

https://tovejansson.com/wp-content/uploads/2023/02/Tove-Jansson-Hobbit_original-vs-printed.jpeg
Tove Jansson illustrated the Swedish translation of J.R.R. Tolkien's The Hobbit in 1960. Her unique illustrations focused on landscapes and locations, differing from the traditional depiction of Tolkien's characters.

Right-sizes LLM models to your system's RAM, CPU, and GPU

https://raw.githubusercontent.com/AlexsJones/llmfit/main/assets/icon.svg
llmfit is a terminal tool that right-sizes LLM models to your system's hardware and recommends suitable models based on quality, speed, fit, and context dimensions. It supports multi-GPU setups, MoE architectures, and dynamic quantization selection.

Why does C have the best file API

The author rants about how programming languages neglect file access, requiring manual parsing and serialization. They praise C's memory mapping feature for efficient binary data access.

When does MCP make sense vs CLI?

The author claims MCP is dying due to its lack of real-world benefits and argues that CLIs are simpler, faster to debug, and more reliable. They believe CLIs are a better abstraction for LLMs to interact with tools.

Running Neural Amp Modeler on embedded hardware

https://www.tone3000.com/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2Fcwdo0dfy%2Fproduction%2F2b7a8625f4000b61e1ac1423229e085b9fcbd51c-1920x1106.png&w=3840&q=75&dpl=dpl_4Su9jzaBLUqeGsKA4j33mHBSHZeR
The authors optimized NeuralAmpModelerCore for embedded hardware, reducing processing time from 5 seconds to 1.5 seconds on the Electrosmith Daisy Seed. They achieved this by improving model size, compute efficiency, and model loading, laying the groundwork for their next-generation NAM architecture.

Show HN: Vibe Code your 3D Models

https://raw.githubusercontent.com/ierror/synaps-cad/main/assets/splash%402x.png
SynapsCAD is a 3D CAD IDE combining OpenSCAD code editor, real-time 3D viewport, and AI assistant for natural language design modification. It's available for Linux, macOS, and Windows with pre-built binaries and supports offline usage with local AI models.

What our DNA reveals about the sex life of Neanderthals

Please enable JS and disable any ad blocker

C64 Copy Protection

A long document was divided into six volumes for easier reading, with Volume 1 providing a technical foundation for the rest. The volumes cover C64 disk copy protection techniques, industrial machinery, and copying scene tools from 1982 to the early 1990s.

Decision trees – the unreasonable power of nested decision rules

https://mlu-explain.github.io/assets/ogimages/ogimage-decision-tree.png
We're farmers with a new plot of land and we're using a Decision Tree to classify trees as Apple, Cherry, or Oak based on Diameter and Height. We've created a tree with a root node at Diameter ≥ 0.45 and subsequent splits at Height ≤ 4.88 and Diameter.

Next-gen spacecraft are overwhelming communication networks

https://bear-images.sfo2.cdn.digitaloceanspaces.com/herman-1683556668-0.png
The aerospace industry is facing a data transmission challenge due to rapidly increasing data volumes from modern spacecraft. Traditional communication methods are struggling to keep pace.

Show HN: I built a zero-browser, pure-JS typesetting engine for bit-perfect PDFs

https://raw.githubusercontent.com/cosmiciron/vmprint/main/documents/readme/manifesto-1.png
If you generate PDFs with headless browsers or HTML-to-PDF tools, you've accepted a compromise: heavy dependencies, memory leaks, and "approximate" layout that shifts across environments. VMPrint offers a stable, high-performance alternative. It composes documents from a versioned JSON instruction stream and guarantees identical layout given identical input, down to the sub-point ...

Long Range E-Bike (2021)

https://jacquesmattheij.com/ebike_1.jpg
The user built a custom 10S17P e-bike battery pack with 170 Samsung E35 cells, increasing the range of their S-Pedelec from 45 km to 180 km. The pack was designed to be safe and efficient, with a capacity of 2150 Wh, and the user hopes to inspire others to consider e-bikes as a commuter car replacement.

Ape Coding [fiction]

https://rsaksida.com/images/og-image.ff8cd797.jpg
Ape coding is a software development practice where humans write source code manually, emphasizing craftsmanship and direct human involvement. It gained popularity as a reaction against AI-written software, but its value now lies in recreation and education, with enthusiasts building large-scale projects like a human-written compiler for the AI-designed programming language 𒀯.

Why XML tags are so fundamental to Claude

We're verifying your browser Website owner? Click here to fix

Microgpt explained interactively

https://growingswe.com/blog/microgpt/opengraph-image?16ed498263341a16
Andrej Karpathy's 200-line Python script trains a GPT model from scratch, learning statistical patterns in human names and generating plausible new ones. The script's core loop involves tokenizing, embedding, attending, computing, and predicting the next token, with the model learning to minimize cross-entropy loss through backpropagation and RMSNorm normalization.

Chorba: A novel CRC32 implementation (2024)

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
A novel method for calculating CRC checksums efficiently without lookup tables or hardware support is proposed, achieving 100% throughput increase over current state of the art. This approach outperforms or matches hardware-accelerated solutions on x86_64 and ARMv8 processors.

Setting up phones is a nightmare

https://joelchrono.xyz/assets/img/banner.png
User helped parents set up new Android devices, minimizing data collection and telemetry by disabling tracking, uninstalling bloatware, and replacing default apps. They're seeking advice on phones with fewer privacy-invasive tactics and hardware that doesn't collect data excessively.

Flightradar24 for Ships

https://atlas.flexport.com/6.png
We use cookies and tracking technologies to collect and share browsing data for analytics and advertising. By using this product, you agree to our Terms of Use and applicable Terms and Conditions.

Everett shuts down Flock camera network after judge rules footage public record

Microgpt

http://karpathy.github.io/assets/microgpt.jpg
The user created a simplified 200-line Python script called microgpt that trains and infers a GPT model from scratch, showcasing the core algorithmic essence of large language models. The script demonstrates how the model learns statistical patterns from a dataset and generates new plausible-sounding documents, such as names, by predicting the next token in a sequence.

Allegations of insider trading over prediction-market bets tied to Iran conflict

https://www.morningstar.com/assets/img/marketwatch.dad5abb.png
Prediction markets like Kalshi and Polymarket faced controversy over insider trading and sensitive markets tied to Iran's supreme leader. Traders made large profits from advanced knowledge of US and Israeli strikes on Iran, prompting calls for legislation to ban such practices.

I built a demo of what AI chat will look like when it's “free” and ad-supported

https://99helpers.com/images/main.jpg
A satirical AI chat demo showcases every monetization pattern imaginable, including banners, sponsored responses, and freemium gates. The tool is an educational tool to help marketers and developers understand AI monetization, but it's not an actual ad network.

Interview with Øyvind Kolås, GIMP developer (2017)

https://www.gimp.org/news/2026/02/22/%C3%B8yvind-kol%C3%A5s-interview-ww2017/%C3%98yvind_Kol%C3%A5s.jpg
yvind Kolås is the maintainer of GEGL and babl, color engines of GIMP. He was involved in GIMP's project after fixing a patch for GIMP's transform tools.

Show HN: Logira – eBPF runtime auditing for AI agent runs

https://private-user-images.githubusercontent.com/41631269/556349675-d8b53356-e0ee-4a7e-b232-8548ba216610.gif?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.pm3ZV8x-ODNbkCD_WIsXCmuRJvD-RcR-hpIx13x8mQY
logira is an observe-only Linux CLI that records runtime exec, file, and net events via eBPF. It helps you see what actually happened during AI agent runs and other forms of automation, with per-run local storage for auditing, post-run review, search, and detection triage. Audit what an AI agent actually executed, changed, and connected to during a run (for example, codex --yolo or claude ...

You don't have to

https://www.scottsmitelli.com/articles/you-dont-have-to/conclusion.png
The author expresses frustration with the increasing reliance on AI-generated content and the devaluation of human connection and craftsmanship in the digital world. They argue that software engineering is an art and a science that requires care and attention to detail, and that the use of AI tools should not replace human expertise and judgment.