I wrote about the unsexy ways in which AI has gotten into my workflow as a writer who works for myself. No interest in having it think much less write for me, but transcription/dictation advances are real and so are little workflow hacks to store ideas slate.com/technology/2026/05…

The biggest unlock for me with AI is still voice as the input mechanism. I’ve basically used voice as my primary work input for the last 9 months. I barely type anything now. Most days, that means Mac Whisper bound to a hotkey so I can dictate into any text box. Usually Claude Code or Codex. The other version is voice memos on iOS. When I have a project to think through, a feature to build, or a messy problem I need to understand, I’ll go on a run and just ramble into my phone. That sounds almost too simple, but it has changed how I work. Typing still has a lot of friction. It’s very easy to stare at a blank box and write nothing. Talking has much less friction. Once you start recording, it’s actually kind of hard to say nothing. And the models are forgiving. You don’t need to say it perfectly. You can talk in fragments, repeat yourself, circle around the point, and they can usually turn that into something useful. I heard someone describe typing as “Morse code for the brain,” and that feels right to me. Typing often compresses the signal before you even get the idea out. But the biggest thing is that voice forces me to get started. It solves the cold start problem in a very practical way. Try this: Pick the most important thing you need to work on this week. Record a voice memo where you talk through the whole thing. Go longer than feels necessary. Include the messy details. Then paste the transcript into Claude Code or Codex and ask it to ask you questions one at a time. Answer those questions with voice too. I think you’ll probably see the value pretty quickly.

Your meeting notes shouldn't be trapped in a mobile app. Monologue Notes has a CLI and an agent skill – list, search, and pull your transcripts directly inside Codex, Claude Code, or any terminal-capable agent. MacStories covered it: macstories.net/notes/monolog…

This is where we are right now. And i’m not gonna lie it feels pretty magical 🧚‍♀️ Qwen3.6 27B running inside of Pi coding agent via Llama.cpp on the MacBook Pro For non-trivial tasks on the @huggingface codebases, this feels very, very close to hitting the latest Opus in Claude Code, or whatever shiny monopolistic closed source API of the day is. In full airplane mode. Most people haven’t realized this yet. If you have, it means you have a huge headstart to what I call the second revolution of AI. Powerful local models for efficiency, security, privacy, sovereignty 🔥