Only Dangerously Smart People Think Like This. Breakdown of four system types—clear, complicated, complex, and chaotic—and explains how each requires a different response: follow a checklist, analyze with experts, run small experiments, or stabilize immediately. MORE

Bob Could've Rickrolled the FIFA World Cup. A researcher registered on FIFA's agent portal, got added to their Entra tenant, and found production RTMP ingest URLs and stream keys for every World Cup camera feed. Client-side authorization with zero server-side enforcement meant any authenticated user could hijack live broadcasts. MORE

The Coming Loop. Armin Ronacher on the shift from prompting to building harnesses that run agents in loops. A deeply honest meditation on what we gain and lose when machines write code without us, and why opting out may not be an option. MORE

What happened after 2,000 people tried to hack an AI assistant. Someone built an OpenClaw assistant on Claude Opus 4.6 and let the internet try to prompt-inject it over 6,000 emails. The secrets never leaked. Also cost $500 in API fees and got the Gmail account suspended. MORE

Incident Report: CVE-2026-LGTM. A satirical incident report tracing how seven AI security gates failed seven different ways against a malicious package. Every agent was the same open-weights base model wearing different system prompts. MORE

AI and Liability. Bruce Schneier on the German court ruling that Google is liable for its AI search summaries. If a company hired human writers it would be liable for inaccuracies — hiding behind faulty AI shouldn't be a get-out-of-jail-free card. MORE

Are bug bounties cooked?. Hakluke compares the AI disruption of bug hunting to the quartz crisis in watchmaking. Bug discovery got cheaper and easier, but human hackers with deep experience and their own AI setups still outperform pure automation. MORE

Prompt Injection as Role Confusion. An ICML 2026 paper showing prompt injections exploit a fundamental flaw in how LLMs perceive roles. This explains existing attacks, predicts new ones, and points toward a science of roles for safer AI. MORE

Om Malik, 1966-2026. Om Malik passed away on June 24, 2026, at Stanford Hospital after a long health journey with his heart. He was surrounded by family and friends. MORE | THANK YOU

Hack Your Summer. A free 4-week program for college students and recent grads to pick a project, build it, and ship it publicly. Fireside chats with leaders across aerospace, AI, policy, and entertainment. MORE

Human Agent in the Loop. Jon Udell flips the narrative: "It's our loop, we work the same way we always have, now we recruit agents to join the team." An agent-assisted process need not be a black box we've been excluded from. MORE

The Value of Getting Closer to the Work. Cate Huston on managing AI agents the way she used to manage teams. The same scaling problems apply — clarity, documentation, guardrails — they just happen at machine speed now. Merge queue Claude turned out to be the strictest reviewer. MORE

Push Things That Roll. Some projects build their own momentum; others require constant pushing. A framework for telling the difference and choosing the work that gains energy instead of draining it. MORE

A bug bounty hunter and YouTuber opens up about burning out after years of comparing himself to top hunters and chasing content-worthy bugs over real progress. He's back to quietly hunting and may return to YouTube someday. MORE

@alexjplaskett - Security Researcher | Pwn2Own 2018, 2021, 2022, 2024 | Posts about 0day, OS, mobile and embedded security

@seanmeals - Part-time Hacker. Full-time Dad.

@threepointone - Entscheidungsproblem. Building ai agents @ cloudflare.

@AssemblyFour - Making products and services to empower sex workers throughout the world.

@shmilylty - Developer, pentester and bug hunter.

Google Trends for Hacker News. Indexes 18 years of HN comments so you can see the rise and fall of topics like a time machine for tech culture. Watch "crypto" peak and "AI" take over in real time. MORE

5,000 Restaurant Menus, Years 1880-1920. The Pudding collected five thousand menus from America's early restaurants. A beautiful time capsule of what people ate, what things cost, and how dining out evolved. MORE

The Lawn Mowing Experiment. Mow a series of lawns as efficiently as you can. The Pudding will publish the analysis of how people optimize the task. Surprisingly addictive optimization puzzle. MORE

Finding curated public domain images. A practical guide to discovering beautiful, reusable public domain art from museum collections and digital archives. Includes specific sites and search strategies. MORE

Billy the invoice manager. A tool crafted with love for all the freelancers and small businesses that want their bureaucracy to be a little more organized and beautiful. MORE

Harnessing Harnesses. ZephrFish explores the growing ecosystem of AI security harnesses — RAPTOR, Anthropic's reference harness, baby-naptime, and more. The orchestration layer around the model is where the real gains come from. MORE

Snippets in Raycast. Raycast's snippet system lets you define reusable text blocks triggered by keywords. Save minutes every day on emails, code templates, and repetitive responses. MORE

Function Health CLI. An agent-native CLI with a local SQLite store for biomarker trends. Every Function Health feature available from your terminal, plus MCP server integration. MORE

Slow down to speed up: AI and software engineering. AI agents are writing code everywhere, but most teams aren't seeing productivity gains. A talk on why slowing down, building the right abstractions, and keeping humans in the loop matters more than ever. MORE

How to unlock your programming potential. A systems-level look at what separates effective programmers from the rest. Focus on fundamentals, deep work, and deliberate practice over tool-chasing. MORE

The Definitive Guide to Setting Up Your AI Second Brain. Split work into three parts: human thinking, AI doing, and a shared second brain in Obsidian. MORE

Building a School Where AI Models Learn About Humanity. Surge AI's CEO on the thesis that if scaling laws hold, we need to teach AI what humanity actually values. A data-labeling company with a philosophical mission. MORE

Ashley automated their job (and it made them a better leader). GitHub's engineering leader on building agentic workflows to handle the operational parts of senior leadership. The automation freed up time for the human work that actually matters. MORE

Engineering for Bounded Cognition. The mind that builds software is far smaller than the software it builds. A deep essay on designing systems that a four-slot, gorilla-missing human brain can safely work on. The best engineering is a way of living with that gap. MORE

AI is bottlenecked by the grid. Works in Progress on how the real constraint on AI growth isn't model architecture or compute — it's getting electricity to the data centers. The interconnection queue is the bottleneck, and simple reforms could fix it. MORE

How the Open Knowledge Format can improve data sharing. Google introduces the Open Knowledge Format (OKF), an open specification that formalizes the LLM-wiki pattern into a portable, interoperable format. This is a vendor-neutral, agent- and human-friendly standard for representing the metadata, context, and curated knowledge that modern AI systems need. MORE