Week 1: shipped one, scrapped one.

What shipped: Excacloud

The win of the week. I verified the full purchase flow end to end, tested the pro features, and split the project into proper dev and prod environments on Cloudflare. Pushing a change no longer means crossing my fingers over live users. I can validate everything in dev first.

It sounds small written down. It isn’t. “Production-ready” is the line where a side project stops being a toy, and Excacloud crossed it this week.

The day I threw away

I started a new browser extension, ThreadPulse, and reached for a tool that looked powerful on paper. It wasn’t. It asked for permission on nearly every step, rendered its UI wrong, and after a full day of clicking approve I had no working output to show for it.

So on Friday I deleted it all and restarted from scratch, this time reusing Excacloud’s project structure as a reference. By Saturday I had a working UI with profile analysis and export. Losing a day to reset was, weirdly, the best decision I made all week.

A tool that costs more time than it saves is a blocker. Cut it early; don’t wait for it to redeem itself.

The site has a name now

I decided to stop hiding my work behind a resume page. The plan: a real personal site at lptai.dev (projects, writing, contact) and one consistent handle, @lptai, everywhere.

The nudge came from a community post about the doubts of starting a project from zero: naming, icons, design, all of it. I realized my own early questions are someone else’s search query right now. That’s the whole reason this blog exists.

The workflow that tied it together

My notes used to live in two places, which means they lived in none. This week I found a setup that finally clicks: capture daily notes in Obsidian, structure them in Notion, and use Claude Code to draft posts and pull content ideas out of the pile. Three tools, one pipeline. This post is the first thing that came out of it.

What I learned

• Cut failing tools early: the sunk cost is never coming back. A day lost beats a week lost.
• Be your own first customer: ThreadPulse “felt like something was missing” because I hadn’t used it on myself yet. Dogfooding surfaces requirements specs never will.
• Good tools compound: Obsidian + Notion + Claude Code is worth more together than any one of them alone.
• Your old doubts are content: the questions you struggled with are the ones worth answering.

Next week

• Buy lptai.dev and wire up the redirect.
• Refine ThreadPulse by using it on myself first.
• Write up the Obsidian + Notion + Claude workflow in full.

That’s week one. More broke than I expected, and that turned out to be the interesting part. See you next Monday. ☕

— Tai