Let’s be clear: AI-assisted cyber attacks are NOT an excuse for pure negligence. OpenCode’s security posture has always been terrible: let us not forget the baffling OpenCode cross-origin RCE vuln from last year that let anyone execute whatever code they wanted on a local OpenCode HTTP server! I have not recommended OpenCode since then and have watched in horror as the team has ignored open source requests to harden the service and make it more transparent. Security is very rarely a technical problem: in my experience, it’s almost entirely cultural. You CAN make your services ironclad: you just have to believe in putting the time and energy into actually doing it. Fool me once, shame on you. Fool me twice, shame on me.

Amazon did collapse: the stock in the dotcom bubble crashed down to 6$ from 100$. Amazon only survived because Bezos somehow by a miracle found 600m$ in the bond market weeks before things froze up on 2000. Jassy then saved the company with AWS from 2000-2005 by turning it from a low-margin internet company to an actual hyperscaler that could "spin the flywheel". Let's not do a revisionist history on Amazon: it is a miracle it survived after such a crash.

You Can See Everything is a masterpiece: a re-reckoning of truth, Silicon Valley’s worst “fail fast” and “fake it till you make it” cultures. It examines the often perverse incentive-through-lines in business, entertainment, media, politics, and everything in between. The result? You leave questioning everything: questioning if you did in fact see everything.

AI needs new infra

a few of my sketches from working on the Pele logo

This is Pele, my new rendering library for diagrams and charts, compatible with the latest Mermaid syntax. - 100-500× faster than mermaid.js - 6× smaller bundle - No dependencies - Easy to theme - Responsive to device width In pursuit of making Obsidian insanely fast and cohesively designed, I always found the Mermaid library to be too big, too slow, and too hard to theme. Pele will be the new renderer in the next version of Obsidian, but you can start using it now. Pele [ˈpɛlɛ] is named after the goddess of volcanoes and fire. It's open source under the MIT license. Try the playground!

If we don't invest, the talent pipeline will completely dry up one day.

cracking open your recently shipped ai slopped test suites

In the 80s and 90s, nuclear scientists faced a difficult question: how do we safely and humanly store nuclear waste that will be dangerous for thousands and thousands of years? How do we communicate something is dangerous and potentially catastrophic to a civilization 10,000 years in the future that may not have writing, language, or communication as we know it today? Some of my favorite ideas that emerged: * Create a religion that can last through time that understands the nuclear symbols are evil and dangerous. * Craft landscapes and structures (pictured) that are inherently hostile and a person would not want to enter. * Breed a special species of cats that "glow" around nuclear waste as a supernatural symbol to future societies that those areas should be avoided. This tailscale snippet is a great example of this but in code. I've seen other "supernatural" ways to encourage code authors/readers to avoid certain areas: obscure ASCII symbols, obfuscated, minified, and difficult to read code, large comment blocks with something like "DO NOT MODIFY" or "WE DO NOT UNDERSTAND THIS BLOCK-DO NOT CHANGE" But in the age of AI, how do we discourage AI agents from modifying or mutating especially dangerous or sensitive code?

I’m super excited to be sharing The Beginners Guide to Training Your Own AI Coding Model on Your Own Data at @AllThingsOpen this year 🎉

I'M GLAD I'M NOT THE ONLY ONE

On episode 44 of @opensourceready, Brian (@bdougieYO) and John (@johncodes) speak with Mikyo King (@mikeldking) of @arizeai about building AI workflows that improve through observation and evaluation. They explore why evals extend beyond pass-fail tests, how open source helps establish shared practices, and when custom agent tooling earns its complexity. Tune in! hubs.ly/Q04zmFWh0

This is basically what a career in Open Source is like

This sounds like an absolute nightmare: 2 people, 1 team, one person has unfettered permission to slop as much as they want “to find value”. The other has to clean up the mess: guess where traditional “engineer” and “product” roles fall in such a hierarchy! On principle, I reject the notion that in the age of AI, you have to drop the technical bar to the ground in order to find value and engineers only function is to make the slop-system actually work. Entire industries have emerged from value which was driven by elegant and nuanced technical engineering work!!! Hyperscalers, containers, Linux, ML/AI, etc etc! The two-slice team also rejects the idea that design and delighting users is important at all: taste, user experience, agent experience, product flows, knowing the ICP, etc etc! Critical to a sustainable product! Instead, I propose the 3-station ship: - a crows nest - a helm - a star chart Each station equally important: someone must sit the crows nest to see what’s on the horizon (talk to customers, understand user needs, prioritize where to go), someone must man the helm (steer the ship, direct the agents, fire the slop cannons when needed), and someone must ponder the star charts to know where the ship currently is (taste, delightful user experiences, intuitive agent experiences) The rest of the ship? Manned and operated by agents doing the grunt work.

the east-to-west journey along 57th street in manhattan is fertile ground for creative reflection. it has it all. aspirational commercialism at its sugary center dissolving at either end into the void. euphoria, nihilism, nordstrom. you can live a thousand lives between the fdr and the wsh

Good morning New York!