Fine-tuning a language model for your own purposes is easier than it has ever been. The technical side is a couple of hours of understanding and a few minutes of compute. While your friendly neighborhood agent handles most of the process, it needs some babysitting and course corrections.
Teams build shared understanding and intuition through product discovery. Discovery artifacts are an imperfect representation of what was learned, but help the team remember what they did together. Since agents don’t participate in the work, those imperfect artifacts need to get better to help them move in the right direction.
While working on my side project, Tilecraft I’ve been thinking a lot about the rendering of intentions, and how they influence the outcomes in vibe-coded software.
The best ideas and most creative solutions don’t just happen in the depth of work, but often in the between steps. By taking a step back from the work we often find new and better connections.
Last big change to my personal workflow was to switch to Joplin some time ago. I’ve been generally quite happy with using an open-source tool, as it has made me less concerned about my data becoming inaccessible due to company changes. But my notes were still in a database rather than files (and I’m slowly turning back to files to retain control, a good article on this topic: File over app). Plus,…