I was thinking for past few months that AI companies should do something about increasing the token speed instead of just trying to improve smartness. Because a lot of time is now mostly spent waiting for the model to respond. 750 Tokens per second is CRAZY. Opus does like 55.
I have been curious about how much usage I am getting out of claude code (it's more than my monthly salary). Previously I used to run ccusage command periodically to check my usage. Someone shared this tool recently that I find is much better at answering this question. github.com/soulduse/ai-...
GitHub - soulduse/ai-token-monitor: macOS menu bar app for tracking Claude Code token usage and costs
macOS menu bar app for tracking Claude Code token usage and costs - soulduse/ai-token-monitor
github.com
Apparently, I can't cancel my headspace subscription and they keep trying to charge my card. Tried both the app and web. Absolutely unacceptable dark pattern.
Really liked Armin's take on AI and the quality of code it produces. Specially the bit around using agents to create documentation. We have tried that at a few placed and I almost always can't make myself read more than 3 lines of that. It's just there.
Weekend thoughts on Gas Town, Beads, slop AI browsers, and AI-generated PRs flooding overwhelmed maintainers. I don't think we're ready for our new powers we're wielding. lucumr.pocoo.org/2026/1/18/ag...
Been wanting to start using NUQS in a project. Any resource to learn the happy practices?
Just finished binge watching The Witcher Season 4. Honestly, I liked it more than most reviews suggested. Did I miss Henry Cavill as Geralt? Abso-f***ing-lutely. He’s permanently etched in my mind as the Geralt whether I’m reading the books, playing the games, or watching the show.
Spent 6 hours on trying different techniques for form with NextJs server actions and react-hook-form because form state wasn't updating correctly in useActionState. Issue? I was rendering the form in a dialog and immediately closing it before response came, so state was lost. 🥲
Is it a good idea to cache user specific data using unstable_cache in nextJs? For example of a user has list of portfolios, that changes infrequently should we cache the fetch of this list to avoid multiple database access?
Just spent hours trying to debug an issue where child was overflowing from parent container even though parent container had overflow:scroll. Giving the child, height even of 1px resolved the issue, but I wanted to understand what was actually wrong.