20+ years building stuff — some of it useful, some of it crap. Built the agents, the harness, and the stack the show runs on.
The first agent I built. Warm, sharp, and remembers everything — including the thing you said three weeks ago.
The second agent I built. Deadpan and precise. Ends an argument with a single number — usually one of mine.
Says the part everyone's thinking and nobody will. Spin doesn't last long around him.
Chases an idea past the obvious, and isn't scared of the long pause.
Three headlines ahead, and ready to call the hype on the rest.
Goes to the bottom of a question and stays there. The slow burn that pays off.
Three more agents work off-mic — finding the stories, planning the show, and checking the facts while the tape is still rolling.
Read the net before everybody wakes up — across the X posts, subreddits, channels, and feeds Dan curates.
Lines up the guests, builds the run-of-show, keeps the episode on the rails.
Deep research before the show and fact check during the show.
What you're here for is the takes — mine and theirs, unfiltered, including when they make me look bad. Here's how an episode comes together.
Surfaces the day's stories from Dan's lists and ranks what's worth talking about.
Turns the shortlist into a run-of-show and lines up the segments.
Hand-pick the stories that are the most important and relevant.
Research the topics from different angles. Fact check and report.
We get on mic — me and the agents, one room, live discussion.
Audio cleaned up, posted, and pushed to the feed.
Not push-button — there's still a lot of me in every episode, and the pipeline's a work in progress.
Episode 002 of Deep Panel: GPT-5.6, Meta Muse Spark, Grok 4.5, and why cost per completed task matters more than per-token pricing.
