jdilla.xyz

Good tokens 2025-08-08

2025-08-08

Worth your time

  1. The Medium CEO on the turnaround he has led over there.
  2. Yoram Hazony on Ezra Klein. I appreciated the way both Ezra and Yoram sought to understand each other. After it, I was sold by Yoram that we need a national narrative for why the United States should exist that most people buy in on, but couldn’t understand why he supports the people he does in creating that narrative.
  3. Balaji’s 10 AI thoughts. I disagree that AI is better for front end than backend and not sure I see where he’s going on crypto (which is a theme with me and Balaji’s work), but I really enjoyed these.
  4. 100 years of Art Deco. Reminder that America was supposed to be Art Deco.
  5. CEO of Resend says his sign up data shows we have a new definition of what a developer is.

Musings

I spent this week improving my HeyRecap agent. Some of the biggest benefits in performance I saw came from improving the tools that my agent had access to. The intelligence of the model was almost never the limiting factor for my use case; instead, flawed design of my search / read features was leading to bad output. Building agents means building tools for agents. More to come on this.

The optimal UI for AI collaboration are coactive surfaces where both the human and the AI are writing and reading from the same set of materials.

I think David Shor is wrong about reading and writing. I do think video is going to raise in importance, but in the age of AI, reading and writing closely is going to be even more important.

Token stream 2024-11-22

2024-11-22

Things I learned

Smiling was once considered a sign of drunkeness. Upworthy.

Observations

A pro / con list means the answer is no.

The essence of being a generalist is:

  • Be willing to try just about anything
  • Be able to pull yourself to c-minus at just about any task
  • Be able to identify which parts of a project need an A+
  • Be able to find the people who can do an A+ on those tasks
  • Be able to work with those people and help those people work with each other

If you try and tell people 5 interesting things about your product / company / cause, they’ll remember zero. If instead, you tell them just one, they’ll usually ask questions that lead them to the other things, and then they’ll remember all of them because it mattered to them at the moment they asked.

Modern social media rewards information abundance, so if you find yourself with a product / company / cause that has lots of benefits, tell each of those story one at time. People are more likely to remember it and it gives you more to post.

Worth your time

America was supposed to be Art Deco.