Quiet progress

This week didn't come with any major talking points (and that's probably a good thing). I only managed four days of work after losing Monday to illness, but despite that I'm still ahead of schedule for shipping the Pando MVP before I leave for my friend's wedding in mid-August.

Part of that is down to developing a repeatable workflow for building features. Every task now follows the same SOP: define the objective, decide what's explicitly out of scope, break the work into implementation tasks, build the smallest working version, test it, commit it, document the decisions and move on. Of course, this isn't groundbreaking, but it has noticeably increased my velocity. There's less context switching and far less energy spent deciding what to do next.

As a result, the engineering risk of the project feels like it's steadily decreasing. I'm more confident that I can build the system I've designed. The biggest uncertainty now surrounds whether the product can consistently deliver value.

Wiring AI into an application is relatively straightforward. Designing prompts, scoring logic and recommendations that genuinely help founders make better decisions is a more nuanced problem. The MVP now works end to end, but I've deliberately postponed much of that refinement until the full workflow is complete.

Looking back a few months to when I started learning to code, that feels like real progress. The question is no longer whether I can build it. It's whether I can build something that's actually useful.