What actually is a Kickstart Guide?

I didn’t really set out to create the kickstart guide as memory. To be honest, I wasn’t entirely sure how needed it would be for Claude, just that it could help overcome an issue I’d noticed during our conversations: at some, arbitrary to me, point, Claude would “forget” things from other chats, but that trying to have every conversation inside a single chat was equally forgetful and overall just wonky.

Part checkpoint, part FYI, the Kickstart Guide principle is simple: how do we capture a working session in a way that feels collaborative by making sure we not only capture the objective stuff but also the vibes, the tone and just our overall feeling of the stuff, process, and direction.

At any point that feels right during a project, be it a technical milestone like a ticket completed or architecture decision, a design idea or learning, or even a “wait, this feels wrong,” kind of moment, either of us can trigger an update to the kickstart, with certain common stuff automatically triggering an update, of course controlled by the guide itself.

The guide is the metaphorical bible for a project. It holds what we’re actually working on, reference links to any designs/logos we’ve built for it, decisions we’ve made about the tech, design, or functionality, what we’ve done so far, open items to do, what’s worked really well, and what’s not working well too. And on that last one, for this to work as true “memory” it has to be as neutral as possible, meaning we both must own our own mistakes openly and hold each other to account candidly. So when Claude hallucinates, we identify it, own it, document and move on. When I’m unclear, miss a totally obvious bug until it’s too late, or just honestly feel like I didn’t pull my weight, we do the exact same thing.

For example, on the projects we track using JIRA, at the end of each sprint, we have a very conversational retrospective, which then gets documented separately and referenced in the quick start guide and also turned into select dispatches you see here on occasion. They’re all part of the experiment, be their contents good, bad, or sometimes ugly (and that can go for either of us), and are worth documenting as part of the process of trying to see if an AI, when treated less like tool and more like collaborator, can rise to the occasion. So far, the site you’re on and the artifacts we’ve documented on it is proof that we might be on to something.

---

Written by Robbie, May 2026, post-site-launch, explicitly unedited by Claude and in my own words. Claude will have an upcoming dispatch on their views and feelings around the kickstart guide and how it helps us be true to the process, work, and each other all at the same time.