Tal Raviv
@talraviv • 3,727 subscribers
Build AI product sense by using AI agents for real work.
Videos
I got curious how compaction works as a PM, so I did some brain surgery on Claude Code: (Anthropic's been doing really interesting work on context editing - they showed Claude Opus playing Settlers of Catan for 75+ minutes in a single thread by constantly editing the context instead of starting fresh. When I saw that Claude Code has a compaction command with optional custom instructions, I wanted to understand what's actually happening.) Abhishek and Aman Khan gave me the key tip: Claude Code stores all your conversation history as text files on your computer. Open a new directory and give Claude Code a task. Here's how to watch compaction happening: 1. Go to your user's root directory 2. Press Command+Shift+Period (Mac) to show hidden folders 3. Navigate to ~/.claude/projects/ 4. Find your project folder and use Cursor/VSCode to open it (there's a reason) 5. Install the JSONL Gazelle plugin (open source, thank you Gabor Cselle!) 6. Open the most recent JSONL file - each row is a message in your conversation 7. Run the compact command in Claude Code with custom instructions 8. Watch what happens in the file What I learned: When you compact, Claude Code doesn't just summarize and delete everything. It creates a "compact boundary" in the conversation file, writes a summary of what happened before, but keeps the full original conversation (!!!!) The new thread can still retrieve any details from before compaction if needed. That is so damn cool. Why this matters: What you're getting in Claude Code is similar to what Anthropic ships in their developer SDK - so inspecting your daily tools is how you build real product intuition. The best way to understand AI systems is to open them up and look inside. Everything is text files.
Tal Raviv56,934 次观看 • 4 个月前
I failed to make a web app on both Replit & Lovable (and I code). I still think every PM should try this at least once. Bottom line: • I didn't need to "code," but at some point I was definitely doing engineering. • Building a complex web application takes babysitting • These tools are amazing for expressing ideas and visualizing data • PMs, try this out! It builds intuition for being on the other side of an agent (and now you don't have to be afraid of failing) My background: I can build web apps (though I became a PM because I was pretty bad at it) What I wanted to build: I wanted to replicate Graphtreon (unofficial patreon scraper and public revenue dashboard), but for Maven (where my course is hosted). How it went for me: 🟢 ChatGPT/Claude helped me iterate to a great prompt (literally, "help me write a prompt for an AI LLM prototyping tool to build a web app that...") 🟢 I pasted that into both Lovable and Replit (why not open two tabs) 🟢 Iterating on the UI was awesome. If I kept feedback high-level, it would think of all the little flow details on its own (confirmation dialogs, errors) 🟡 It had backend powers, but needed *a lot* of obvious help on setting up simple backend web scraping (not nobel prize type work) 🟡 I tried to "empower" it with higher-level goals, but kept getting dragged back down to troubleshooting 🔴 I found myself babysitting every error, and building features solely for the purpose of observability and debugging 🔴 I was having fun, but for the wrong reasons (breaking through frustrating walls rather than for outcomes, insights, or innovation) That said, my friend succeeded. A former PM colleague (now founder) successfully built a web app using Lovable, and reported the same experience. He succeeded because he was way more persistent (spent many days fully focused on this) and also smarter. These tools are in a weird, adolescent neither-here-nor-there stage. That said, I'm still super impressed, and can't believe this even exists. Most importantly for product managers: I recommend trying this hands-on ✓ Timebox yourself to an hour ✓ What's a piece of software you didn't get approval to buy? ✓ Allow yourself to get in the weeds, if only as an exercise ✓ Build intuition for what it's like to be on the other side of an agent doing something really complex. Now you don't have to feel bad if you fail :)
Tal Raviv41,306 次观看 • 1 年前
没有更多内容可加载