Episode 12: Production Systems

Put two boards next to each other.

Board one is beautiful: color-coded Notion calendar, thirty green publish dates, content pillars, AI prompt templates, everything neatly dragged into “scheduled.”

Board two is ugly: a sticky note that says “claim has no source,” a transcript with 42:13 circled, a clip marked “funny but misleading,” a thumbnail sketch crossed out, and one idea card moved to a pile called “kill.”

The second board is the real system.

The first one only looks productive. Every idea has a date, but the work is not getting sharper. It is just moving across the board.

The real production system is uglier and more useful. It is a chain of gates.

Capture. Select. Produce. Translate. Protect. Learn.

At every gate, the idea either gets better, gets changed, or gets killed.

1. Capture: write down the real signal before it turns vague

The guru advice is: keep an idea bank.

Fine. But most idea banks are junk drawers. Screenshots, saved tweets, random hooks, a folder called “content inspo,” and zero explanation of why anything is there.

A real capture gate looks more like a messy detective wall.

One card says: “Three comments asked why this workflow broke.” Another says: “Guest said the clearest line at 42:13.” Another says: “TikTok search phrase: how to clip a podcast without looking lazy.” Another says: “Newsletter reply used the exact words our audience uses.”

MKBHD is a clean example. In the Cortex episode on his workflow, Marques talks about reviewing cars and phones and jotting down thoughts while he is actually using them. The note is not polished. It is just the live reaction before he forgets it. Enough of those notes become the review story.

That is capture.

Not “I need ten video ideas.”

“I found a real audience signal, and I know where it came from.”

2. Select: kill the idea before the camera is on

The guru advice is: generate more ideas.

The better system asks: which ideas deserve production?

This is where Paddy Galloway’s idea funnel is useful. The idea does not go straight from “interesting” to “recording.” It has to survive creator fit, feasibility, excitement, packaging, and one-page development.

Imagine the production board again.

One idea card has a topic, but no title. Dead.

One has a title, but no reason the audience should care. Dead.

One works for a competitor, but the creator clearly hates making it. Dead.

One has a strong promise, a rough thumbnail, a reason it fits the channel, and enough structure to explain the first minute. That one moves forward.

This is the step productivity gurus usually soften because killing ideas feels negative. But selection is the whole game now.

AI can help you generate fifty decent-looking concepts in a minute. That means generation got cheaper. Taste got more expensive.

The scarce skill is no longer “can I make something?” It is “can I tell what is worth making?”

3. Produce: automate the boring cut

The guru advice is: batch content and use AI to save time.

That is useful until it starts automating the wrong layer.

Batch the setup. Batch captions. Batch b-roll pulls. Batch thumbnail exports. Batch the stuff that wastes energy without improving the idea.

Do not batch the part where you decide what the video is actually saying.

4. Produce: keep the story human

Colin and Samir are a good 2026 example because they do not describe AI like magic dust. They talk about using it for concepting, scripting, and research, and using AutoPod to cut multicam footage faster. But the key part is what happens after that. If AutoPod handles the mechanical angle cuts, the editor can spend more time on story and b-roll.

That is the better AI rule:

Use AI to remove mechanical drag. Keep humans on the judgment.

The wrong version is a pipeline where AI writes the idea, AI writes the hook, AI finds the clip, AI writes the caption, and the creator just approves a vibe that sounds like everyone else.

Creation is getting easier. That is exactly why taste, source quality, and selection matter more.

5. Produce with people: meetings should sharpen ideas, not spread confusion

The guru advice is: hire an editor, hire a producer, build a team.

Sometimes that is right. But if the idea is unclear, a team only makes the confusion more expensive.

Good Mythical Morning is a better picture of how team production can work. Rhett and Link have talked about monthly meetings with writers and producers where they review, shape, and develop ideas. The host taste is still in the room. The idea load is shared. The meeting goes past “what are we posting next week?” It is a room where weak premises get worked on before anyone shoots.

That is the scene you want.

A writer pitches the game. A producer asks if it can actually be filmed. The host says what feels off. Someone sketches the thumbnail logic. Someone asks if the first minute has a real payoff.

Now the team becomes a filter instead of only a labor pool.

Hiring works when every person knows what decision they own. Who owns the angle? Who owns the edit? Who checks facts? Who decides whether a clip is misleading? Who says this is ready to publish?

Without that, hiring just gives your weak system more hands.

6. Translate: stop resizing the rectangle

The guru advice is: repurpose everything.

One podcast becomes ten clips, a carousel, a newsletter, five LinkedIn posts, and a suspiciously inspirational quote graphic.

This is where creator systems turn into content mills.

A real translation gate asks what job the asset has on each platform.

A short has to create tension fast. A carousel has to make one idea saveable. A newsletter can hold sources, caveats, and context. A LinkedIn post should carry one sharp operator lesson. A podcast clip should make someone want the full conversation instead of only consuming a random orphaned moment.

7. Translate: mark the clip while the source is still alive

Spotify’s clip examples make this concrete. The School of Greatness marks clip-worthy moments during recording. That means the clip system starts while the energy of the conversation is still visible.

Heart Starts Pounding is even more useful because it is audio-first. The host makes purpose-built visual clips instead of pretending every audio segment naturally contains a perfect TikTok.

That is the correction.

Repurpose the idea, not necessarily the footage.

The visual here is a transcript with marks in the margin: “standalone hook,” “needs context,” “good trailer,” “do not clip, misleading without setup.” That is a system. Dragging a random thirty seconds into a vertical template is just resizing.

8. Protect: trust is a workflow, not a vibe

The guru advice is usually quiet here because preflight checks sound boring.

But this is the gate that separates a creator business from a content blender.

Kurzgesagt is the extreme example. Their process includes source sheets, expert review, many script drafts, simplification, illustration, animation, music, and sound. Most creators do not need that whole machine. But the principle matters: trust is built before publishing.

9. Protect: run the preflight before upload

The visual is simple.

A source sheet open next to the script. A preflight checklist next to the upload screen.

Is the claim sourced?

Did we separate what the source says from what we think it means?

Does the title match the actual payoff?

Does this clip distort the full conversation?

Do we have rights to the music, image, guest clip, and AI asset?

Does AI disclosure apply?

Does this still sound like us?

This matters more in 2026 because the platforms are pushing in the same direction. YouTube has disclosure rules for realistic AI-generated or meaningfully altered content. TikTok requires labels for realistic AI-generated media. Meta has been moving against unoriginal reposting.

The more content gets easier to make, the more trust becomes the expensive part.

10. Learn: make the postmortem hurt a little

The guru advice is: study your analytics.

That can turn into dashboard theater too.

The learning gate is more brutal. Open the last five things you published and reconstruct the crime scene.

For each one, ask:

Where did the idea come from?

What signal made us believe it mattered?

What got killed before recording?

What did AI help with, and what did a human approve?

Was the clip made for the platform, or just resized?

Was the claim checked?

Did the comments show confusion?

Did the short send anyone to the full thing, or did it die as a floating moment?

Did the next idea get better because of what we learned?

If you cannot answer those questions, you do not have a production system yet.

You have a publishing habit with better stationery.

And that is the honest test. A real system leaves a trail. The production board shows where the signal came from. The clip log shows why the moment was chosen. The transcript shows what was cut. The source sheet shows what was checked. The preflight checklist shows what almost went wrong. The postmortem shows what changes next time.

If all you have is a calendar full of green checkmarks, the machine is not improving the work.

It is just helping weak ideas escape faster.