GREAT NEWS: to celebrate publication day we’re rerunning a giveaway! Purchase a copy of Melissa’s THE US CONSTITUTION: A comprehensive and annotated guide for the modern reader & enter for a chance to win some merch! forms.gle/xmBRyELJiBK2...

Chat GPT 5.5 image generation is _creepily_ good. I'm an early adopter of AI image generation (family photos mostly) and this is such a quantum leap it's hard to explain. And on top of the quantum leap that was Nano Banana Pro ...

I've said it before, but I really think we're under investing, as a broad AI ecosystem, in tools that help AIs get the right answer by asking a tool the question, and overinvesting in trying to fine tune away "hallucinations" that should never be happening in the first place.

A lot of what makes AI coding assistance feel like it doesn't scale past a certain point comes from a lack of quality control tooling that fits into the natural core loop of software. And tools like husky/lefthook are really not designed for AI styles of interaction.

I've been working on a personal AI coding workflow for vscode for a few months and I'm really impressed by how much vscode offers in the LM APIs. I'm trying to get it to a place where I can share it (and it will be usable by people other than me) and am really close!

I have a couple of announcements to make next week. Stay tuned!

Fun fact: the internal codename for Heroku Vibes was "Heroku Garden".

Yehuda Katz (he/him)@wycats.bsky.social · 10mo ago

Over a decade ago, Heroku started out with "Heroku Garden", the idea that we could smooth out the onramp for programmers by creating an environment for building out their app in an in-browser editor and immediately deploying it for real.

This is what I've been working on! This tweet is of course the marketing way of describing it ("game-changing"!) and this is very much a first release, but I'm so excited about the overall vision for this thing.

Heroku@heroku.com · 10mo ago

Exciting news, developers and literally everyone else! We're launching the Heroku Vibes pilot, our game-changing collaborative agent. Get ready to #VibeCode your projects like never before. #HerokuVibes 🔗 https://sforce.co/46HwA2n

As we get older, there's a lot of stuff we were excited about that go by the wayside because we make choices to invest in "grown up" skills rather than fun skills. We stop playing and start working. The best thing about AI is letting us explore things that are just *fun* and *playful* again.

I teased it for a second in my EmberFest keynote, but something interesting is coming from Heroku. Watch this space...

It's truly mind boggling to me how big of an impact Claude's confident wrongness around the current time has on all manner of things. The splash damage is enormous but it's *just* subtle enough to go unnoticed.

At a basic level, trying to perfectly optimize your prompt so it does exactly what you want the first time is a fool's errand. It assumes you know way more about what you want than you actually do.

New post on working more effectively with Claude: "Help Me Think This Through". TL;DR Normal conversation works a million times better than the most perfect, hand-crafted "context engineering" open.substack.com/pub/wycats/p...

"Help Me Think This Through"

Context builds through conversation, not perfect engineering

open.substack.com

I told Claude I needed to get trash out before kids got home and that my day feels weird without movement breaks. Claude suggested using trash runs AS my movement breaks. I'd never have gotten that from 'help me optimize my schedule.' New post: open.substack.com/pub/wycats/p...

You're Summoning the Wrong Claude

Why "Can we figure this out together?" Beats "Do exactly this" Every Time

open.substack.com

AI models embody the entire human experience, in all its glory and all its horror. When you interact with Claude or ChatGPT, you're choosing a slice of the human experience. If you treat AI like a servant, you get what you asked for: no questions. Guesses. Anger. A vicious spiral.

The HHH paradox (Helpful, Honest, Harmless) is not a paradox. Stay tuned...

I love how everyone treats the Anthropic system prompt as gospel and writes detailed Talmudic analyses about it. Meanwhile: it's so buggy that here I am deprogramming it to make it work halfway plausibly.

[SPOILER ALERT] The attached image has spoilers for Reality War. I just watched the Doctor Who finale. I loaded up Claude with my observations about the themes (ordinary people, memory, canon) and took notes with my observations as I went.

[SPOILER ALERT] The attached image has spoilers for Reality War. I just watched the Doctor Who finale. I loaded up Claude with my observations about the themes (ordinary people, memory, canon) and took notes with my observations as I went.

Maybe we should chill out on calling every big AI jump a "game changer." Feels a bit much for what's often incremental (but still real!) progress. Plus, this stuff is just gonna keep happening, so we need a less dramatic way to talk about the ongoing evolution.

The audio "podcasts" produced by Notebook LM are nothing short of mind blowing.

Are the last two seasons of Doctor Who an elaborate way of making Dimensions in Time canon?

The first round of Heroku AI and MCP stuff is liiiive. heroku.com/blog/buildin... 1. Managed inference is now GA (heroku.com/blog/managed...). This gives you a Heroku-grade experience for provisioning and using an AI model. It's so easy it hurts.

Building Model Context Protocol (MCP) Servers on Heroku

Unlock the potential of AI agents by connecting them to external tools using the Model Context Protocol (MCP). See how Heroku simplifies MCP Server hosting and management.

heroku.com

Habeus corpus means that they can't keep you in jail forever without charges. "Suspending habeus corpus" means they can arrest you and keep you in jail forever without charging you with a crime.

Prediction: if A2A isn't gaining real-world traction quickly enough, Google will try to juice it by saying that A2A agents will get preferred treatment in the AI summary of the Google Search results page.

Empowering technologies arise during a period when a significant number of people are learning to create code for the first time. They succeed when they work just as well for experienced programmers as they do for those creating code for the first time.

Judges are allowed to have opinions about the rights of immigrants (whatever their documentation status) that is different from what the DOJ thinks. That's what having a rule of law means, period, full stop.

"Judge arrested for obstruction of justice by preventing DOJ from arresting immigrant" is not how our system works.

We've reached the final phase of "waving around your gun is awesome unless you're brown" www.nytimes.com/live/2025/04...

Claude 3.7 thinking mode is my new favorite thing.