AI FROM MY POVAI, explained by someone who actually ships it.
Practical write-ups from building real AI products — what worked, what broke, and what I'd do again.
Latest
All articlesSet Up an Obsidian Vault with Claude — the Easy Way
You don't need to become an Obsidian power user. Download the app, point Claude at the folder, paste one setup prompt, and let your AI build the whole vault and keep it organized. Here's the easy way, end to end.
- obsidian
- claude
- setup-guide
- ai-workflow
Turn Your Vault Into a Knowledge Graph Claude Can Traverse
A vault of linked notes is the raw material. The real leap is a graph your AI can walk — asking how ideas connect, not just searching for keywords. Here's how to build it with Claude and plain files, no special tool required.
- obsidian
- claude
- knowledge-graph
- ai-workflow
Build a Memory Pipeline So Your AI Stops Forgetting You
Every new session, your AI starts blank — it doesn't remember what you fixed, decided, or learned last time. Fix it with a two-step loop: save lessons to small files after each session, load them back in at the start of the next. Here's how to build it with nothing but Claude and plain files.
- memory
- claude
- ai-workflow
- setup-guide
Write a Rules File So Your AI Stops Making the Same Mistake Twice
Tired of repeating the same instructions every session? A rules file (CLAUDE.md) is where you write your preferences, standards, and hard-won lessons once — and your AI follows them automatically, every time, without you asking.
- claude-code
- claude-md
- ai-workflow
- rules
Stop Re-Explaining Yourself — Package It as a Skill
Every time you catch yourself typing the same multi-step instructions into your AI, that's a skill waiting to be written. Do it once, name it, and your AI runs the whole thing correctly on its own from then on.
- claude-code
- skills
- ai-workflow
- productivity
Stop Making One AI Do Everything — Use Subagents Instead
One AI doing every step runs out of context and works slow, one thing at a time. The fix: hand off scoped tasks to subagents — each with a clean context — and send the right model to each job. Here's how to actually do it.
- agents
- subagents
- claude-code
- ai-workflow