Longer reflections and practical guides from using AI tools.
Running two dozen agent sessions on a laptop is solved. Planning work so it can be handed off, and seeing what needs you across all of it, is not. Here's what I want, what exists, and where the gap is.
In one drop, OpenAI has released hundreds of new maths results from a single unreleased model, claimed solutions to long-standing open problems, including one a leading mathematician called "an instant Fields Medal". Most are not yet verified, but maths is being industrialised, and it's making mathematicians dizzy.
Why use a bookmarking app when AI plus a markdown repo does it better? Capture is simple, the gardening is now largely automatic, and bookmarks become papertrails and knowledge bases.
I want VS Code Live Share, but for a coding-agent session — both of us typing into the same repo, same terminal, same prompt, at once. Here's what I found instead.
I want one place to watch many AI agents at once, on the Claude and ChatGPT plans I already pay for, with Beads in the loop. Here's what exists and what doesn't.
TypeSafe's new System One model puts fast, typed decisions at the center of AI automation—and makes a compelling case for keeping code in charge.
OpenAI solved a massive unsolved maths problem — with a model that's only been training since August 28.
I don't want to wait on one Claude Code thread before starting the next — here's what's actually available today for running several at once on the same project.
A trending question, a 100-hours-vs-20-hours comparison, and my own recent experience that Codex has been unusually solid.
Why a Claude subscription can appear to deliver hundreds of dollars of API-token value, and why that does not mean the numbers are directly comparable.
A practical look at local tools for tracking live usage, tokens, sessions, and history across Claude Code and Codex.
I've got more work than one agent session can handle, and I want a manager layer that farms it out and tracks it cross models
Optio, Paperclip, Gas Town, or Claude's own Agent Teams — which one actually manages a fleet of agents instead of just watching them? A field survey with a verdict.
A simple checkpoint pattern for resuming AI work across sessions and repos.
How to schedule cloud and local Claude Code tasks for when the next subscription usage window begins.
A hands-on look at using Stripe Projects to provision Cloudflare services and credentials for coding agents.
We used to work in apps, each with its own screens. Now we work in the AI, which builds the views we need, and apps become storage that people and AI read and write.
Why reusable agent skills resemble software libraries, and how the emerging package ecosystem compares with npm.
Forty years ago, all you had was the terminal. Now it's back thanks to AI and voice.
From AI-powered tools to AI as the operating system.
