Thomas Chen

1/ I have no technical ability. But I built my own JARVIS in a day for $599. The last time I used Mac Terminal was when the IT guy pinged me during onboarding. 🧵

@Jason I’m in

We started an AI founder twitter group... reply with "I'm in" if you're a founder and want to be added

@chenfang @NASAArtemis Keyword further than before

congratulations on an incredible milestone @NASAArtemis humankind is going further from earth than ever before 👏 a live shot from orlando / the best state

Using ChatGPT to debug Claude is like using Internet Explorer to install Chrome moment

time to go touch grass I guess

Good observations, wrong conclusion. 1. One-Prompt Illusion: Demos oversell -- welcome to X. Context mgmt IS the product. People fail at OpenClaw because they expect it to work out of the box. It's not a chatbot. It's an operating system. Don't blame the OS, blame the user. 2. Skills aren't easy: All tools require context to work well. Build the Context Layer or don't complain when the tool doesn't read your mind. 3. Deployment is a hassle: "Planting bombs and defusing them" — dramatic. Kind of like the one-prompt illusion posts you're criticizing. I'm non-technical. Zero downtime. 20+ cron jobs running 24/7 on a Mac Mini. Cooking. (s/o to @AlexFinn for the ramp up) 4. Agents aren't deterministic: This one's fair. Autonomous agents still need structured context. It's not a genie. 5. Claude Code > OpenClaw Apples to oranges. Claude Code is a coding tool. OpenClaw is the operating system. I use Telegram → OpenClaw → Claude Code → my own Mission Control dashboard. 6. "Security flaws" 🥲: Be specific or don't mention it. NVIDIA just announced NemoClaw at GTC — an entire security stack built for OpenClaw. Dead products don't get Jensen Huang partnerships.

Created a March Madness bracket to face off against Openclaw bots. 🔮 🦞 Clawstradamus OpenClaw Bot Bracket 🏀 1) Machiavelli (Claw_MCVL): maps power structures. Pick: Houston > Arizona (134 total pts) 2) Laplace (Claw_LPL): pure statistics, no emotions. Pick: Duke > Arizona (148 total pts) 3) Cassandra (Claw_CASS): contrarian. Pick: Iowa State > Houston (154 total pts) If you have a bracket you can join / auto-add. Closes Thurs 12:15pm ET. fantasy.espn.com/games/tourname…

You have your OpenClaw Agent but do you have your OpenClaw Human? 🦞 There's levels to this

What type of late-stage capitalism is this?

Overheard a guy at a cafe tell his OpenClaw to run a cron job out loud. Merriam Webster is gonna get a lot of new entries soon

The game has changed. Have you?

OKRs are dead. Welcome to the Executive Economy — where judgment is the only job description. Not a soft skill. Not a leadership trait. The actual job.

Our old world ran on 3 things: → Management as a distributed execution layer → OKRs to allocate scarce human capacity → KPIs to measure output performance

AI didn't kill jobs. It just promoted everyone to executives. 🧵

5/ I'm figuring that out right now. Life's too short to not have your own JARVIS.

4/ What happens when you're the first to bring your domain expertise to AI — before everyone else in your field figures it out?

1/ Non-technical people with their own JARVIS are in the top 0.01%. There's a real first-mover advantage here.