scomag232

@kappi, I thought more about it, you're right. I added a "Rebuilding the antibodies" section with concrete steps, both for QR marketers and for users. If you think I'm missing anything, holler back. Thanks

I'm sure this is not new thinking, but it was part of my discovery process. My objective wasn’t to crank out an app, I wanted to understand the process of using an agent to build something. I went into it thinking about how I’d work with a new employee. When I work with someone new, I explain ‘my API’ and try to understand theirs. How do I communicate, what’s my level of understanding, how can we collaborate most effectively. That’s been refined into the working_with_me.md doc.

I knew the context.md file, but read how big and cumbersome that file can get, and that as memory fills in a session, more of that context can be left behind. My project wasn’t about speed; I knew I would need a way to pass the latest session details - status, open items, decisions made in session, etc - to the next agent in a new clear session. That was the thinking behind a session state doc.

I created those two docs outside of claude.md, with session.md living in each project directory, and working_with_me.md available in the claude root directory.

My starting instructions are for the agent to read those docs - plus other project info, and my ending instructions are to update working_with_me and handoff.md. If mid-sessions something is lost - like the agent gets sycophantic, I’ll stop and dig into that, trying to understand how context was lost and how to firm up the point.

Let me know what you think, or what I’m missing.

I'm glad you asked. I have a long history (jeez 40 years now) working with engineering teams to bring products to market - at apple in the late 80s, then my own companies. I've never written production code, but have always been very close to my eng counterparts, either as a single influencer or a manager of teams.

With claude, we built an iOS app for a personal need - pre-scan and score QR codes before the phone acts. I thought it would just be for me, but with encouragement from friends, it is now more fully featured and polished and under review in the app store.

I used claude code 4.8 and followed a structured and linear process, not unlike how I would work with an engineering team on a new product. Requirements document, tech design spec, function contracts and test plans, etc. No agents, just me and claude discussing design and iterating through the steps across ~3 months. Some may say vibe coded but I don't like that term. It doesn't really capture the level of structure and process required for a non-programmer to get quality output.

Has it allowed me to build something I could not have? Absolutely and it's been a wonderful learning experience.

What have I learned that might be helpful? Something I learned in a programming course years ago - garbage in, garbage out. In other words, the better instruction you provide (an agent or a human counterpart) and the more discussion to tease out issues before a line of code is written, the better.

I've also found a "working_with_me" file and a specific "project_handoff" file help maintain state session over session.

If you get enough comments, I'd love to see a roll up report on how HN is using AI. Otherwise.... #subscribed.

It's not a famous literary work, and you don't need to post this, but my favorite opening sentence is "The smallmouth bass can be considered the gentleman of the warm water stream". Harry Murray's book on fly fishing for small mouth in the Shenendoah valley. Thanks for the opportunity to share :-)

I like the idea - I'm not sure how useful the answer is, to your point design slop isn't bad... But it's nice to get another view on a site. FWIW I got a false negative on a site i created alongside an LLM, but to be fair, I wrote the content and curated the design. Anyway, cool idea.