Tech News

Show HN: Forge – Guardrails take an 8B model from 53% to 99% on agentic tasks

Hi HN, I'm Antoine Zambelli, AI Director at Texas Instruments.I built Forge, an open-source reliability layer for…

New accessibility features powered by Apple Intelligence

Apple has unveiled new accessibility features powered by Apple Intelligence that enhance tools like VoiceOver,…

Stilta raises $10.5M from a16z and YC to help companies rediscover the patents they forgot they had

Stilta announced Tuesday a $10 million seed round led by Andreessen Horowitz.

Show HN: Id-agent – Token efficient UUID alternative for AI agents

id-agent is a token-efficient alternative to UUIDs designed for AI agents, using human-readable word-based IDs that cost…

Gaussian Splat of a Strawberry

All the bugs they found

A developer discovered over 20 security vulnerabilities in Epsilon, a WASM runtime written in Go, using AI…

Elon Musk Loses OpenAI Trial, Vows Appeal After Jury Dismisses Claims Over Statute of Limitations

A federal jury has thrown out every claim in Elon Musk’s lawsuit against OpenAI and Sam Altman, finding the case to have…

Numexpr: Fast numerical array expression evaluator for Python, NumPy, Pandas

NumExpr is a Python library that accelerates numerical array expressions for NumPy by avoiding intermediate memory…

AI does not yet outperform engineers in managing IT outages

AI tools are rapidly advancing in monitoring IT systems.

CISA Admin Leaked AWS GovCloud Keys on GitHub

A CISA contractor publicly exposed highly sensitive AWS GovCloud credentials, plaintext passwords, and internal system…

My domain got abused on GitHub Pages

A developer discovered their domain was being abused on GitHub Pages when Google Search Console alerted them to a…

Theo Baker spent four years investigating Stanford. Before he leaves, here’s what he found.

"There's a common refrain among [young] people in this world that it's easier to raise money for a startup right now…

Mini Shai-Hulud Strikes Again: 314 npm Packages Compromised

A compromised npm account published 637 malicious versions across 317 packages on May 19, 2026, including popular…

PyTorch Landscape

Codex-maxxing

The author describes how Codex, an AI coding agent, has evolved from a tool primarily for coding tasks into a platform…

Show HN: Hsrs – Type-Safe Haskell Bindings Generator for Rust

Hey everyone!

Apple Silicon costs more than OpenRouter

LLMCap – A proxy that hard-stops LLM API calls when you hit a dollar cap

LLMCap is a proxy service that enforces hard dollar caps on LLM API calls across six major providers (Anthropic, OpenAI,…

Sieve – scans Cursor/Claude chat history for leaked API keys

Background: I was using Cursor to set up an OpenAI integration.The agent read my .env file, added the key to the config,…

The American Rebellion Against AI Is Gaining Steam

The last six months in LLMs in five minutes

Between November 2025 and May 2026, large language models underwent significant developments, particularly a "November…

An Apple (II) for Teacher

By 1980, the Apple II had become a major success, leading to Apple's IPO and making co-founders Steve Wozniak and Steve…

What political censorship looks like inside an LLM's weights (Qwen 3.5)

Researchers used mechanistic interpretability to analyze political censorship in Alibaba's Qwen 3.5 LLM, discovering it…

Anthropic co-founder to present AI encyclical alongside Pope Leo XIV

Pope Leo XIV will release his first encyclical, "Magnifica humanitas," on May 25, 2026, addressing the safeguarding of…

'FedEx says your parcel has drugs': The scam that trapped an Indian comedian

Ankita Shrivastav is among millions of Indians who fall prey to cybercrimes every year.

Mexican government breached by solo user with Claude, 150 GB exfiltrated

A solo attacker breached the Mexican government in late 2025 using only Claude and ChatGPT subscriptions, exploiting…

Lawyers Apologize After Fake Claude-Generated Quotes Appear in Trump Layoffs Case

Attorneys representing a former Homeland Security official admitted using Anthropic’s Claude Console to help draft a…

SandboxAQ brings its drug discovery models to Claude — no PhD in computing required

Other venture-backed companies like Chai Discovery and Isomorphic Labs have raced to build better models.

Alignment pretraining: AI discourse creates self-fulfilling (mis)alignment

Researchers conducted a controlled study showing that the discourse about AI systems in pretraining data significantly…