Fabrizio Rinaldi

Fabrizio Rinaldi

@fabrizio.so

Overthinker. Designer & co-founder: πŸͺΆ @typefully.com 1.5M ARR πŸ’Œ mailbrew.com (sold) 🎧 lofi.cafe Having fun sim-racing 🏁 I directed a sci-fi short film once, maybe I’ll tell more stories in the future.

5.9K followers211 following360 posts

Built my own "Fab OS". Started as a beautiful UI for my OpenClaw, and it's now my life HQ. PWAs are great these days, so this even loads offline and sends me notifications. Built in a couple of days with Opus 5.5, what an amazing model.a Demo with fake data:

If you're curious how it works, my OpenClaw runs on a Mac and this web-app is served on Tailscale, so I can use it on any device. Only the data I selectively share with that machine is available for this; I guess it's up to you how tight you want something like this to be.

But anyway, I absolutely love building software for myself. With AI, scope isn't the problem anymore. Clarity of thinking and taste are, maybe more than ever. Sharing or productizing this is tricky, though. I might just open source it as a "framework" you can give your agent to build your own.

My super clean Obsidian theme is now named Verso, and it's live. Just search for "Verso" in the community themes store and you'll find it. I polished it a lot more, very delightful to use. Enjoy!

Cogito keeps getting better. Added Quick Look, better AI chat, LaTeX, HTML, better typography, and maaany fixes and improvements. It's by far the best Mac markdown app for me now. Still completely native, and completely free.

Codex is my favorite coding app right now. It's clean, but has everything I need to ship fast. It's also quite delightful to use and snappy, and shows enough context without overwhelming.

Dropping another Cogito update ✍️ Added a new Inspector sidebar that gives every note: β€’ Summary β€’ Writing stats β€’ Tasks progress β€’ Navigable outline β€’ Related notes from your library Related notes powered by @tobi qmd.

I built my dream Markdown editor for Mac. β†’ Introducing Cogito (pronounced koh-gee-toh). It started out of frustration: Obsidian is powerful but overwhelming. iA Writer is beautiful but feels built for a different era.

Nothing felt right for how I actually write and work now: plain files, lots of folders, agents and scripts editing alongside me. I wanted both: native and beautiful, powerful and calm. So I finally built it. Fast, keyboard-first, polished, truly native.

I have /finalize skill that I'm often using after working on something, to make it shippable. It auto-fixes obvious issues, creates branch/PR if needed, simplifies where safe, consider blind spots and alternatives etc. Working quite well for me. gists.sh/linuz90/8e0...

gists.shFinalize β€” Claude Code skill for ship-readiness passesFinalize Your job is to help ship. Take ownership of the last serious pass before merge: understand what changed, identify what could still go wrong, look for what can be safely simplified, fix...

Great prompt I tried today: β€Ί Is there a smart way to make {file path} smaller and tighter, while actually making it achieve the same goals even BETTER? This made some AGENTS .md files ~50% smaller, with no quality loss.

I'm shipping faster and faster thanks to my little "lwt" worktree helper. Makes it so easy to spin up agents on tasks. Now it even auto-splits the terminal in iTerm and Ghostty. Some cool examples:

No sign-up, no tracking. Just a better way to read and share notes and code. Killer use case: have your AI agent like OpenClaw share research, notes, or code with you as a pretty gist. Drop it in a meeting chat and watch the reactions. Free and open source ✌️

If you're using OpenClaw on Telegram, set up forum topics. Turns one messy chat into organized topics. Each with its own isolated LLM session. Your finance chat never bleeds into your dev context. Setup: enable Threaded Mode in BotFather, tell your agent to create topics and route messages.

Worktrees are great but the terminal UX is terrible. Built a small CLI that makes them usable for me. It shows the status of each one clearly, and makes adding and removing super easy. Copies .env too (recursively) and optionally installs deps. β†’ github.com/linuz90/lwt

I never wanted a "second brain", but here we are. An indirect effect of sharing my notes with my OpenClaw is that my notes are now improving and compounding daily.

Whatever recurring need you have, ask your OpenClaw "should we make a skill and/or cron for this?" and it will guide you through it. And just make sure to put your notes where it can access them and ask it to index them and point to the most important ones in USER .md.

What if OpenClaw could make you feel like you’re the star in your own movie? β†’ Set up elevenlabs β†’ Choose a warm, enthusiastic voice β†’ Get an energetic soundtrack β†’ Ask it to prepare a great morning digest β†’ Tell it to use the voice + soundtrack Insane to start the day like this 🍿

I wanted OpenClaw to know what I’m up to, but it runs isolated on a VPS. So I had Claude Code build a /me endpoint on my Mac: a small Bun server that aggregates local data into one JSON blob. It’s locked down with Tailscale, so only devices on my private mesh can access it.

Now my agent can check /me for: β€’ Apple Health: sleep, HRV, recovery, resting HR, steps, active calories β€’ Mac: presence, idle time, focused app β€’ Spotify: now playing, recent tracks, top artists β€’ Steam/PlayStation: games played + time β€’ Slack: online/away, DND, status β€’ More (suggestions?)

My bot pulls this every 30 minutes during heartbeats, so it knows what’s up and whether to ping me. It’s truly context-aware: - idle for hours + good recovery β†’ suggests a walk - late night + still on Mac + low HRV β†’ nudges me to rest - deep in work + music on β†’ stays quiet

If you're using multiple AI agents across many projects, you need a global AGENTS file. I keep mine in a dev folder, and symlink to: ~/CLAUDE . md (Claude Code) ~/AGENTS . md (Codex, Gemini, Cursor...) This keeps behavior consistent across all agents.

Your favorite AI coding agents can now manage your socials. We're finally shipping a Typefully skill today. Our skill makes AI agents learn about Typefully, so they can help you draft, refine, schedule, add media, and a lot more πŸ”₯ It's quite magical to see it in action:

I got tired of forgetting my Claude Code slash commands, so I asked Claude to build a function that shows all my commands, with autocomplete. Type `ccmd` β†’ fuzzy search β†’ hit Enter to run. Game changer if you have lots of commands scattered around.

Claude Code runs my life now. It checks my tasks, email, notes, health data, calendar, and more through subagents, MCPs, tools, and scripts. And I chat and speak with it via Telegram all day. I called it β€œMax” since it’s maxing out my Claude sub πŸ˜‚

Been using Typora to view Markdown files on my Mac. For example, when I'm claude-coding in plan mode, I grab the plan link and open it in Typora for a much nicer reading experience.

I just discovered I can set icons for my iCloud Notes folders in Ulysses. I use Claude Code a lot for these notes (on my Mac or via Telegram) so I'm glad I can now browse/edit them with a beautiful app too. I finally feel happy again with my note-taking system.