Marc Köhlbrugge
@marckohlbrugge
🚧 https://t.co/aMZ4K3a24s 🏠 https://t.co/1KH3Ex4dcT 🤑 https://t.co/z2G662TXrN 🐴 https://t.co/80DsPhFw6c 🚀 https://t.co/sxmNUGTr0X 💎 https://t.co/Mz8Z7MkY3H 📣 https://t.co/oPODwYDwSS 💻 https://t.co/7vViQvjS0J 📸 https://t.co/NJnFBXCkML #️⃣ https://t.co/6qUnpK7nRx
I thought this was common knowledge, but this article is about to cross one million views. Lesson: what’s obvious to you, might not be to others. Share.
kill your free plan https://x.com/marckohlbrugge/status/2034292939262705804/photo/1
If anyone is thinking of writing a book, here's the workflow I recommend: 1) Go for a walk, and record voice memos of everything you want to say. Don't worry about structure. Do a complete braindump. 2) Use MacWhisper to transcribe all these voice memos. Dump it into Claude Code and brainstorm with it to find a proper structure. 3) Do more brain dumps. Again, I find doing these walking is most productive. 4) Transcribe again. Put into LLM. Ask it to organize into markdown files and an outline. 5) Work on the outline together with Claude Code. 6) Now ask it to identify missing gaps and interview you about it. Use dictation tools like Wispr Flow or open-source alternatives to answer them and expand the markdown files. 7) Read through the results. Do a lot of tweaking. Try to be very mindful of whenever something feels off. Rewrite it. At some point in this process (rather sooner than later) you want to get external feedback too. Perhaps when you get the first outline. Share it with early readers. Get their feedback. Feed it to Claude Code and again, ask it to interview you where useful. The goal is not to have the LLM write for you. The goal is to have the LLM help you structure your thoughts, identify gaps, and extract your knowledge. For my book, I did a lot of manual typing as well because I think when I started dictation tools weren't that good yet. And I think typing activates another part of the brain which is more painful to use, but probably good to do as well. So experiment with that too.
I had writer's block for my domain name book for the longest time. Then earlier this week, I asked Claude Code to review what I had written so far and interview me to fill in the missing parts. I dictated my answers and Claude Code adapted these transcripts into the manuscript. This has proven to be a very productive way for me to write. It's important to note that these are still my insights based on my own experience. LLMs don't have this type of knowledge, because it's simply not in their training data. It's also written in my personal writing style because I dictated the answers and it could see my existing writing. Claude Code simply improved the transcripts for better flow and fixing grammar. I also used it to "refactor" some parts of the book to incorporate some lessons from @robfitz' Write Useful Books like moving some of the most valuable parts of the book to the front and cutting fluff. I tried this workflow one or two years ago, but the LLMs weren’t ready yet. The result was generic-sounding slop I couldn’t put my name on. But now, LLMs are advanced enough to act more like an actual editor that helps you improve your writing while keeping your personality intact.
stop fucking around. remove your free plan. increase your prices. do some marketing. send some emails
Gave this link to Claude Code and it did this magic ✨ https://x.com/marckohlbrugge/status/2022890287841054999/photo/1
A bunch of background jobs for Startup Jobs started failing… Turns out it was caused by the collapse of the Iranian Rial which lead to the exchange rate exceeding 1,000,000 and overflowing a database column 😅 https://x.com/marckohlbrugge/status/2028236228970099180/photo/1
I just tried http://ui.sh and unfortunately I cannot recommend it. Here’s why… It’s too good I now want to redesign all my sites 🔥🔥
Took over 30 minutes and 5+ immigration officers closely inspecting every millimeter of my passport before I was free to leave mainland China. The reason? My passport looked too new 😂
what's your favorite, personal project, that was not a success?