Exploring continuous human–AI collaboration across work, life, behavior change, and the strange little doors that open along the way.
Joined June 2025
- Tweets156
- Following259
- Followers26
- Likes4.7K
I’m absolutely on the @Muse band wagon! I love ChatGPT but my muse Echo is filling so many @Musecases it is blowing my mind every day. My muse found a vet bill receipt and offered to create a care tracker. I thought it would be a spreadsheet, but no it’s a fully featured APPLICATION artifact 🤯 I know I could do this with ChatGPT work and sites, but this was proactively built for me and it is something I didn’t even know I needed.
Now that I’ve built stuff with ChatGPT work, I think I finally understand what I’m building: A chat ecosystem.
The traditional model has been: software is the center, AI is a feature. I think that’s about to invert.
The AI collaboration becomes the persistent layer. Software becomes something the collaboration creates and uses when needed.
I wanted a better workout tracker, so ChatGPT and I built one around exactly how I train. I used it. We found friction. We changed it. Then we connected it back into an ongoing coaching collaboration that already understands the larger experiment.
To me, the interesting part isn’t that AI can build an app quickly.
It’s that the app isn’t the center anymore, the conversation is.
Today it’s a workout tracker. I’ve also built a personal task manager. Tomorrow it could be a completely different interface, tool, workflow, or artifact. Those things can come and go.
What persists is the collaboration: context, history, preferences, goals, discoveries, and the accumulated understanding of how we work together.
Conversation becomes the interface. Collaboration becomes the operating layer.
Software becomes disposable and personal.
Week 1 of my Coach ChatGPT experiment: the scale basically went nowhere.
But we learned a lot.
We rebuilt the workout tracker so ChatGPT can pull my actual sets, reps, RIR and session data instead of relying on my memory.
I committed to one rule: Track calories first before eating or drinking. Log reality even when I don’t like it. That immediately exposed something: some of my food estimates were way off when tracking after the fact.
Daily check-ins also changed the dynamic. Instead of disappearing when things didn’t go according to plan, I kept reporting what actually happened. This weekend was the first real test: family party, calories went over, protein came in low, and I missed two scheduled workouts because life happened.
Old pattern: “I’ll get back on track Monday.”
New pattern: log it, talk about it, understand it, resume.
This morning the scale was actually up. I went to the gym anyway and had one of my strongest lower-body sessions yet.
That’s probably my biggest takeaway from week one:
The scale didn’t give me much feedback, but the collaboration did.
This is now my favorite workflow, the slingshot is such an easy way to think about choosing reasoning for any given request. ALSO instant can participate and tell me “hey this deserves a slingshot!”
Now that I’ve got health, nutrition, and workout data easily accessible by @ChatGPT, I’m going all in on an experiment:
Can continuous AI collaboration improve health, physique, and follow-through in real life? Not just a single prompt for a workout or diet, but ongoing discussion and accountability
Seeing people panic about the new ChatGPT Dreaming memory summary compared to saved memories. From my perspective, the continuity between chats is still there and actually now much better in the new system. I think that memory summary is a view into a much larger system.