2026-04-29

The stack reveal — my favorite content format

Show the whole system on one screen. Don't worry about it being impressive enough. It is.

#content#signature-moves#youtube#format

The format

Pull up a terminal. Pull up the agent fleet config. Pull up the deploy state. Pull up the API spend dashboard. Cut to camera. Talk for 90 seconds about what's on the screen.

That's the stack reveal. It's my favorite content format because it's the only one that's impossible to fake.

Why it works

Most AI content is the host saying things. Stack reveal is the host SHOWING things — agent specs, terminal output, prompt logs, dollar figures. The viewer doesn't have to trust me. They can read the screen.

The other thing it does: it scales the perceived size of the operation. When someone sees 15 agent specs in ~/.claude/agents/, the impression isn't "this guy has 15 agents." It's "this guy is operating at a scale I haven't seen before."

The truth: most of those agents are 100 lines of markdown. They're not magical. But the impression they create is real, because what's on screen is real.

What goes in a stack reveal

For me, the standard checklist:

  • Active agent list (ls ~/.claude/agents/)
  • Skill count (ls ~/.claude/skills/ | wc -l)
  • API spend dashboard (or a screenshot of one billing page)
  • Recent commits across repos (git log --all --since='7 days ago' --pretty=oneline | wc -l)
  • One running process (Discord bot, briefing cron, sync job)
  • One result that landed today

Six things, on screen, narrated. 90 seconds. That's the format.

What it costs

Time to record: about 15 minutes. Time to edit: 0 — these don't get cut, they get one-take. Time to upload: 5 minutes.

About 20 minutes a week, gets me one short and the raw material for one long-form B-roll segment.

What you can steal

If you have a system you're proud of, the stack reveal is the highest-ROI content format I know.

The bar is low: a screen recording, a phone propped at desk height, narration. The thing that matters is showing the artifact, not talking about the artifact.

If your system isn't worth showing on screen, that's a useful signal. Build a system worth showing first. Then record.