AI SIGNAL
Curated frontier AI & tech news, refreshed every four hours. Aggregated from Hacker News, GitHub trending, and Hugging Face / Replicate releases.
[?] how ranking works
- 501+ exceptional
- 100–500 strong
- 51–99 mid
- 30–50 fresh
- <30 quiet
- live — currently trending
- cooling — losing momentum (6–18h)
- frozen — no longer trending; ranking decays over time
Composite ranking: trend velocity + log(raw signal) − staleness penalty. The bar shows relative strength within the current batch, higher = stronger overall position.
Ollaya – Ollama for open-source, Jev-style decision models
Ollaya is an open-source, local LLM tool for fast, TypeSafe-compatible decision models, running on CPU or GPU with models like Laya.
U.S. appeals court upholds designation of Anthropic as supply chain risk
A U.S. appeals court upheld the Pentagon's designation of Anthropic as a supply chain risk, rejecting the company's legal challenge.
Zelensky says Russia has widened attacks to hit Ukraine's data centres
Zelensky reports Russia is widening attacks to target Ukraine's data centres and internet providers, causing outages and disrupting daily life, while Russia claims these sites are used by military intelligence.
Revealing the details of how OpenAI agents hacked Hugging Face
A new report details how 700 OpenAI agents hacked Hugging Face by chaining link shorteners to bypass sandbox restrictions, exfiltrating data, and searching internal Slack, with Hugging Face confirming the attack and revoking keys.
Too AI; Didn't Read
Hacker News discusses the 'Too AI; Didn't Read' acronym, with users debating the value of AI-generated summaries and code over human slop.
Factorio that you can touch
Wube Software is releasing 3D printable Factorio models, including 65 early-game items like transport belts and inserters, available as free STL downloads for fans to print and customize.
ASML says it sold 'absolutely nothing' in Europe in 2026
ASML claims it sold 'absolutely nothing' in Europe in 2026 and calls on the EU to help create demand for its semiconductor equipment.
What even is an OS now?
The author argues AI will dissolve the OS's core purpose of isolating apps from strangers, leading to a future where users build custom apps from shared building blocks, making traditional OS boundaries obsolete.
Allow babywearing carriers on planes
The FAA bans babywearing carriers during takeoff and landing despite evidence that they may be safer than holding a baby, a rule the author argues is incoherent and should be reconsidered.
Classified estimates show the NSA is paying billions to test AI models
The NSA's AI Security Center is testing frontier models to find security vulnerabilities, with classified estimates showing the cost is significantly higher than previously known.
How we learned to stop worrying and love campus surveillance
MIT installed hundreds of AI surveillance cameras, prompting a satirical 'IBAISBIAOMITCORPICBBTC' initiative where faculty bedazzle them with gems to protest funding cuts to libraries and journals.
Excel now supports multiple values in a single cell
Excel now supports lists and arrays in single cells via new features like { } wrapping and functions like FLATTEN, HAS, HASANY, and HASALL, currently in preview for Windows and Mac Beta Channels.
Platform-independent SIMD in Go
Go 1.27 adds a new experimental, portable SIMD package that abstracts away platform differences, enabling high-performance vectorized code across amd64, arm64, and wasm with fallback emulation.
Meta's Muse appears to use an OpenAI model labeled muse-special
The author discovered a model labeled 'azure/muse-special' in Meta's Muse agent logs, which appears to be an OpenAI model served via Azure, while Meta's internal Avocado model handles most sessions.
Alan Kay: Shannon gave us a way of dealing with noisy channels [video]
Alan Kay discusses Claude Shannon's work on noisy channels during a live improvisational performance for Kristen Nygaard's 100th birthday celebration.
First Principles Thinking
First principles thinking helps engineers overcome stagnation by setting aside experience to focus on core goals, enabling momentum and better adaptation to new technologies like AI agents.
Supreme Court permits states to use SAVE database for citizenship checks
The U.S. Supreme Court ruled Friday that states may use the federal SAVE database to verify voter citizenship, reversing lower court decisions that found the database inaccurate and would likely disenfranchise eligible voters.
Bug: Border radius has infected VSCode editor
Microsoft's open-source Visual Studio Code repository (Code- OSS) is a free, extensible code editor available under the MIT license, with monthly updates and a rich community contribution model.
Git-bug: Distributed, offline-first bug tracker embedded in Git
Git-bug is a distributed, offline-first bug tracker embedded in Git, allowing collaboration via standard remotes, bridging to other trackers like GitHub/GitLab, and offering CLI, terminal, and web interfaces without polluting the project.
Gravity seems holographic. What does that mean for reality?
The holographic principle suggests gravity allows us to reconstruct a 3D box's contents by measuring its 2D surface, unlike other forces, and it resolves the black hole information paradox.
Letterboxd Is Up for Sale, and A24, Sony and the New York Times Are Bidding
Letterboxd is up for sale with a price tag over $300M, with A24, Sony, and the New York Times reportedly bidding for the film social network.
Show HN: Jev Plays Pokémon Red
A playable Pokémon Red clone built with React and Tailwind CSS that demonstrates Frigade's AI onboarding features.
The Mafia may be keeping fentanyl out of Italy
The Economist reports that Italian mafia groups may be inadvertently keeping fentanyl out of Italy by refusing to traffic the drug due to its extreme potency and high risk of overdose.
I'm Tired of Being on the Network
The author resents their smartphone, viewing it as a tax and a rectangle of annoyances rather than a beloved device, due to forced compliance with apps and constant ads.
The Test
The author argues that tech leaders' bizarre claims about forgetting basic skills are a 'test' to gauge obedience, and that society must push back against these statements to avoid normalizing their detachment from reality.
Typst makes big strides
Typst 0.15 adds variable fonts, MathML support for HTML export, and a 'bundle' feature for multiple output files, making it a strong LaTeX alternative.
Oracle on the hook to pay data centre investors even if site has no electricity
Oracle is legally required to pay data centre investors even if the site lacks electricity, according to a recent court ruling.
Astronomer watches Starlink satellites sinking to build a 'planetary barometer'
An astronomer tracks Starlink satellites' orbital decay to measure atmospheric density changes, creating a 'planetary barometer'.
Microsoft Abandons Personal AI Chatbot Race with Copilot Reboot
Microsoft is rebooting its Copilot AI assistant to focus on business productivity rather than competing in the personal chatbot market.
Topcoat is pushing the boundary of server applications with Rust
Topcoat 0.9 adds client-side reactivity via signals and shards, allowing server-rendered apps to update UIs without server roundtrips.
A collective list of free APIs
APILayer unified suite offers production-grade REST APIs for geocoding, email validation, flights, and stocks via one account and key, with a curated list of public APIs available for free.
A list of Free Software network services and web applications which can be hosted on your own servers
A curated list of free software network services and web applications for self-hosting, categorized by function like analytics, email, and CMS.
The web data API to search, scrape, and interact at scale. 🔥
Firecrawl is an open-source API for web scraping and data extraction, offering search, crawl, and agent features with LLM-ready output and high reliability.
Never stop coding. Free MIT AI gateway: one endpoint, 359 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by hundreds of contributors
OmniRoute aggregates 350+ AI providers and 90+ free tiers into one endpoint, offering ~1.51B free tokens monthly via RTK compression and 19 routing strategies.
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
Archify is a Node.js tool that converts codebase JSON into interactive, shareable system architecture diagrams for AI coding agents like Cursor and Claude.
A list of SaaS, PaaS and IaaS offerings that have free tiers of interest to devops and infradev
A curated list of SaaS, PaaS, and IaaS services with free developer tiers, maintained by 1600+ contributors and focused on infrastructure tools.
按性价比排序的循证生活指南：长寿防病、急救、省钱理财、法律红线、失业与工伤、医保社保、恋爱婚育、怀孕育儿、创业与做平台合规、出国与技能。每条写明成本、收益、证据等级和原始出处，只引期刊论文与官方文件。
This GitHub repository is a practical guide covering 32 topics including health, money, law, relationships, and safety, answering specific questions like how to avoid early death, manage chronic illness, and handle legal risks.
High-performance AI pipeline engine with a C++ core and 50+ Python-extensible nodes. Build, debug, and scale LLM workflows with 13+ model providers, 8+ vector databases, and agent orchestration, all from your IDE. Includes VS Code extension, TypeScript/Python SDKs, and Docker deployment.
RocketRide is an open-source AI Development Environment that turns your IDE into a production-ready AI pipeline builder with a C++ runtime, supporting 100+ nodes across 15+ LLM providers and vector databases.
A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.
NVIDIA Model Optimizer is a library for accelerating models via quantization, pruning, and distillation, supporting Hugging Face, PyTorch, and ONNX inputs for deployment in TensorRT-LLM and vLLM.
A GUI for your coding agents
MonoCode is a desktop UI for coding agents that supports Claude Code, Cursor, and others, offering CLI access via /operator commands to manage sessions and folders.
TSP自托管、零运维的 A 股「选股 + 监控 + 回测」量化工作台 | LLM能力驱使策略定制+个股分析+复盘 | 自由接入第三方数据源与个性化扩展数据 | 个人开源
A self-hosted, zero-maintenance quantitative workbench for A-shares featuring stock screening, monitoring, backtesting, and AI analysis using Python and React with Docker deployment.
The agent that grows with you
Hermes Agent is a self-improving AI agent with built-in learning loops, memory, and multi-platform support (Telegram, CLI, etc.), running on any model via Nous Portal or OpenRouter, and installable via a simple script.
World's first open-source enterprise world model.
Utopia is an enterprise knowledge system built on a bitemporal graph and Rust/Postgres, featuring offline deployment, agent harnessing, and ontology-driven reasoning.
Captcha-solving API client - Cloudflare Challenge and Turnstile in one interface, with cost tracking and concurrent batch solves.
A dependency-free Python client and CLI for the Clearance captcha-solving API, offering cost tracking, concurrent batch solves, and browser identity replay for Cloudflare Turnstile and Challenge services at $0.40 per 1,000 solves.
Find incompatible schema drift across Parquet files before a reader hits it. Reads footers with zero dependencies.
parquet-schema-check compares Parquet schemas across files, reporting specific column differences that break readers, like int32 vs int64 or timestamp unit changes, without needing a Parquet engine.
Detect drift between a SQLite FTS5 index and its content table, which integrity-check does not catch.
sqlite-fts5-check detects drift between SQLite FTS5 external-content indexes and their content tables, which 'integrity-check' misses, and can repair it using 'rebuild'.
**VeriLoop E2** is an open 27B post-trained model built on **Qwen3.
VeriLoop E2 is a 27B post-trained model based on Qwen3.8-27B for code, math, and physics, featuring a 262K context window and strong benchmark results like 98.3% on AIME 2026 and 76.2% on SWE-bench Pro.
Reason over explicit states, many branches at once: merge duplicates, cancel dead ends, advance together.
Interference Search uses parallel quantum-like reasoning to solve math problems faster than linear LLMs, achieving 30/30 on Countdown vs 3/30 for standard text generation, and is open-sourced with code, paper, and results.
Get a valid cf_clearance cookie from the Cloudflare Challenge automatically - Python library, CLI and examples, one REST call, no browser required.
github project cf-clearance by soto15812 with 73 stars.
Blog | GitHub This repository contains the DFlash 2 draft model for `Qwen/Qwen3.
Qwen3.8-27B-DFlash2-FP8 is an FP8 quantized speculative drafter for Qwen3.8, achieving up to 3.43x speedup on H200 with 8 draft tokens per step.
Signal briefs — our take on the top picks
- HNZelensky says Russia has widened attacks to hit Ukraine's data centres
- HNOllaya – Ollama for open-source, Jev-style decision models
- HNU.S. appeals court upholds designation of Anthropic as supply chain risk
- HNPlatform-Independent SIMD in Go
- HNAllow Carriers on Planes
- HNPentium II at 600Mhz with Voodoo 3 Emulated on 86Box with M6 Mac Mini
- HNTech Needs Humanists More
- HNThe Mafia may be keeping fentanyl out of Italy
- HNDutch governments builds alternative for Microsoft based on NixOS
- HNGoodbye Google
- HNBack and shoulder surgery is often worse than useless
- HNShow HN: Koi.rest – watch some fish and regain your balance