Okay, migadu.com it is. Cheap, cheerful and useful. I like their transparent and matter-of-fact approach.

email is such basic currency, and still such an unsolved problem. Every provider prices it like a 2010 collaboration suite, not a 2025 protocol service. I get that spam filtering makes consumption pricing messy but still... smells like unmet demand.

Oh yes, of course. This explains so much.

I’m sure there are a million CLI domain lookup tools. Now there are a million and one. I wanted something simple, not whois, just a way to see what domain variations might be available. So Claude and I built this: github.com/rossrader/reno

I'm at the "hmm... I should check out a new email client" part of my Christmas vacation. Came earlier than usual this year. Alpine feels like a solid choice...

I just posted something new to my blog. Today’s models fail loudly and visibly. That makes their limits easy to study and their mistakes easy to catch. Future systems will not be so forgiving., read more about it at https://rossrader.ca/posts/aisucks.html

I totally agree with this analysis. There isn’t an AI bubble, there’s an OpenAI bubble which is a very different thing.

also very cool (from May?).. stitch.withgoogle.com from Google Labs lets you describe a screen in plain language or upload a sketch and get real UI layouts back. can export front-end code or push straight into Figma and keep iterating. loving it as a way to polish a swift app...

can’t wait to play with this over the holidays.. an add-on for OpenCode that turns one AI coding assistant into a small team. helpers focus on understanding the codebase, reasoning, and writing or cleaning up code. can hook into language servers so it also gets structure.
github.com/code-yeongyu...

Sometimes the horror is overwhelming. Hate-driven shootings, continued bombings, victims including children. May all their memories be blessings. I think about my own son and how searing the pain must be. We need to be in active opposition to all this violence and suffering. We have to care and act.

Hat's off to the thoughtful UX designer (team?) at Google that came up with this marvelous way of making long tables more readable in a Google Doc by using a sticky table header. Even though it breaks convention around "something we might print", it works beautifully! Chef's kiss!

Love this.

Out of curiosity I decided to try and run the numbers on how much Netflix you can watch for the energy cost of a ChatGPT prompt

As far as I can tell it's between 5.1 and 10.2 seconds, depending on which end of the 2019 IEA Netflix energy usage estimate you use

simonwillison.net/2025/Nov/29/...

I just posted something new to my blog. I tried to see if I could get better results from combining the effort of three AI assistants, but the results weren't all that satisfying. But I learned a few things., read more about it at https://rossrader.ca/posts/multi-model-deep-research.html

The clash between 1.2.1 and 1.4 shows why a model may act in ways you didn't ask for. Its told to follow its own step order ahead of your preferences, allowing it to quietly override you. I wonder how much of what we call hallucination comes from system instructions rather than the model itself?

I'm starting to create many Claude Code commands recently. One of those things of, why didn't I do this before?

I just posted something new to my blog. A personal site feels right again, so I rebuilt one with a simple GitHub and Netlify workflow and a bit of nostalgia for how the web used to work. If you’re curious how it came together or..., read more about it at https://rossrader.ca/posts/hellowhirled.html