ai professional building Tim - a personal trainer ai that manages clients on imessage @whatrrdotcom
Los Angeles, CA
Joined May 2016
- Tweets3.2K
- Following2.6K
- Followers5.2K
- Likes4.6K
Pinned Tweet
DAY 2 OF 62: I built Tim, an AI texting buddy for personal trainers.
Trainers are stuck on their phones all day. Mid-set it's "can we move Thursday?", "running 10 late", "what time tomorrow?"
So I made Tim. He texts your clients for you. (Yes, blue texts.)
DAY 8 OF 62: An AI kill switch.
Hey! That's me at a hackathon in Irvine, running it live. It watches an AI write code, and stops it the second it goes wrong.
That night: 100 coding problems, 30 with a planted mistake. It stopped 24 of them, about 70 tokens in.
Every story and every number is checked against its source. If anything fails, the episode is held and I get an alert.
If it all passes, it records the show in my cloned voice and publishes it to Spotify and Apple.
DAY 6 OF 62: AI for commercial real estate.
A broker finds clients three ways: events, staying in touch, and cold calls. I built one tool that plans all three.
Staying in touch: current clients hear from him every month, future every two, past every three.
When someone's due, it writes the note with this quarter's market numbers for their area.
DAY 5 OF 62: AI for job hunting.
I applied to 1,000 jobs in one day. Then I took 10 interviews, just for fun.
I didn't fill out the forms. My bot did. It's called JOAB.
How it works:
1. Tell it the job you want, and where
2. It searches the job boards and finds the matches
3. It fills out every form and applies
4. It asks you first, or runs on auto
DAY 4 OF 62: AI for nonprofits, reading forms. I built Fetch.
A hospital faxes over a stack of forms about a homeless patient. The nonprofit has to read it all and decide if they can accept this person.
How I built it:
1. Find: reads every faxed page, sideways ones turned upright
2. Trace: each rule gets the page and the exact quote
3. Check twice: plain code matches every quote, then a second look at the page
4. Hand off: the nonprofit reads the results and decides
DAY 3 OF 62: I built an AI employee for a nonprofit in healthcare.
They take in homeless patients after a hospital stay. Once they're back on their feet, they're discharged. Then where do they go?
That's Sal's job. (Yes, it prints the plan. On paper.)
How I built it:
1. Every answer hits a safety check: plain code, then a small model
2. A fast model asks the questions
3. Another searches the list and suggests what to ask next
4. A bigger model writes the plan with placeholders, never phone numbers