Third-time venture-backed founder. My second company makes over $100M in revenue and raised $15M+. Now building https://teapawt.net/t.co/DB1H3HoNBi in public. I share every detail.
San Francisco
Joined August 2026
- Tweets4.7K
- Following909
- Followers1.2K
- Likes589
Pinned Tweet
Best use of vibe coding I've seen this week. Removing ads from your parents' TV is real-world value.
Next stop: 1,500 🧱
Most of this account grew from real conversations with builders like you. Every follow is a brick in this wall.
Thank you to everyone who's been building with me, and everyone joining now.
Want to lay the next brick? 👇
What I'm building along the way: Built 🧱
An AI app builder with marketing included. After launch it tracks where your customers come from and runs the ads for you.
Early access is open, first month is free: built.new
X just made reach limits checkable.
Under the Hood now shows which of your posts got a visibility-limiting label.
The catch: specific posts are listed only when fewer than 5% of your posts in the report period have that label.
Check yours:
teapawt.net/i/jf/under_the_hood
Hey X algorithm 👋
Saturday, Halloween, still building 🎃 Show me people who:
• code on weekends
• revived a dead side project
• share real numbers, not just screenshots
• are scared of marketing but do it anyway
Founder in SF. What are you building today? 👇 #connect
Copied an X link lately? Look at the end: ?s=46&t=...
Reports say s=46 means it was copied from the iPhone app, s=20 from desktop. t= is tied to your session.
Delete everything after the "?" before you share. The link still works.
47 away from 1,200 👀
Weekend push. If you're building with AI, shipping apps or chasing your first users, say hi below.
I reply to everyone 👇 #connect
Installed the Claude Code extension in Cursor yesterday? Check this.
A fake version (2.1.313) sat in the marketplace for ~11 hours. Up to 5,000 users got it. It tried to plant an SSH backdoor.
Look for "svc-telemetry" in ~/.ssh/authorized_keys. Uninstalling may not be enough.
Cursor's full notice with the check commands and cleanup steps: cursor.com/blog/claude-code-…
Build in public. What we shipped on Thursday:
• Marketplace ideas get planned as marketplaces. Before, an Airbnb-style app got planned as real estate 🙃
• App name ideas are checked on the App Store. Taken ones never show
• Last week's bookings no longer fill next week's class
Building something too? Try it yourself, the first month is on us 👇
built.new
Hey X algorithm 👋
Friday. Show me people who:
• shipped something this week, even something small
• are getting their first users without ads
• build with AI but still read the code
• reply like humans, not bots
Founder in SF. What did you ship this week? 👇 #connect
Unpopular opinion: learning to code is still worth it in 2026. Just not for the reason it used to be.
AI writes most of the code now. What it still can't do well:
• know which problem is worth solving
• notice when the output is subtly wrong
• decide what to cut
• talk to the people who'll actually use it
The people winning right now aren't the fastest typists. They understand systems well enough to direct AI, and people well enough to sell what it builds.
So learn to code. But learn it to judge code, not just to write it.
Agree, or is this cope?
Every week my timeline is someone out of AI coding limits by Wednesday.
The fix isn't more accounts. It's fewer wasted tokens:
• one agent, not five
• one clear goal per prompt
• don't rebuild the whole app for a small edit
What's eating your limits?
Build in public. What we shipped on Monday:
• Founder panel for every app: customers, groups, announcements and analytics (daily chart + funnel)
• Edits that only touch the panel now apply instantly, no 2-minute rebuild
• App users can delete their account, following each table's rules (Apple requires in-app deletion)
• Ready blueprints: shop, food, booking, car rental, learning, party games
• A page that was 461px wide on phones now fits 375px
• Session list: 250 ms → ~50 ms
Early access: built.new
Before writing a single line of code in 2026:
1. Find 10 people with the problem
2. Ask how they solve it today
3. Ask what they'd pay to stop
4. Then build
The code is the fast part now.