# Content Is Code

Matt Palmer, Conductor | AI Engineer World's Fair 2026 | 10:53

Source: https://www.youtube.com/watch?v=yv6xovSsB1U
Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks.
Page: https://aietalks.com/talks/content-is-code
Published: 2026-07-18
Tags: code-generation, data-pipelines, developer-experience, structured-outputs

## TL;DR
- Code is becoming a fast way to create technical communication, including documentation, websites, videos, slides, and product tours.
- When code is cheap, structure, documentation, design tokens, and careful processes become the main limits on quality.
- Teams that keep software structured can shift content production into code and build repeatable content pipelines with AI.

## Summary
Matt Palmer argues that code is becoming the main medium for producing technical content. This includes documentation, change logs, product updates, websites, videos, slides, motion graphics, and product tours. He describes three eras: handcrafted content, expensive code that required professional engineers, and cheap code generation through AI. The change moves the bottleneck away from the ability to write code. Palmer says the expensive part is structure: design tokens, brand rules, clean pull requests, accurate documentation, and clear separation between parts of a codebase. These provide the source of truth that AI systems need to create consistent assets. He calls the required care conscientiousness and says AI rewards organizational excellence as much as technical skill. Palmer expects content engineering to follow creative technologists, with declarative pipelines producing walkthroughs, screenshots, documentation, and product updates from code.

## Key ideas
### Technical communication now includes every way a product explains itself
[00:00](https://www.youtube.com/watch?v=yv6xovSsB1U&t=0s)
Palmer uses content to mean technical communication about the problem a product solves. That includes documentation, change logs, product marketing, emails, videos, home page tours, and other DevRel work. A change log depends on an accurate product diff, while product emails depend on understanding what changed over time. He also includes interactive assets such as a product tour instead of a static screenshot. The common thread is communicating what a product does or how it has changed, whether the output is written content, a video, or a coded asset.

### Product surfaces and videos can increasingly be built as code
[02:09](https://www.youtube.com/watch?v=yv6xovSsB1U&t=129s)
Palmer shows a Conductor product tour built entirely with React and Remotion. He recreated the product surface and used a Remotion scene to walk through a sample user flow. He is candid that the example is buggy and that its flows are incomplete, but he expects this kind of coded product walkthrough to become normal. His broader prediction is that complete product surfaces and videos will be generated through code rather than assembled as separate manual assets.

### AI has made code cheap enough to become a general content medium
[03:10](https://www.youtube.com/watch?v=yv6xovSsB1U&t=190s)
Palmer divides content production into three eras. The first is handcrafted, where time, money, and skill limit what can be made. The second has mature code and cloud tools, but professional engineers remain the bottleneck. The third is the current AI era, where people pay Anthropic, OpenAI, or similar providers instead of always paying a human engineer. AI can help create videos, documentation, websites, and motion graphics, although the results are not automatically good. An engineering mindset and an understanding of the underlying systems still matter.

### TypeScript and React have become Palmer's preferred medium for assets
[05:37](https://www.youtube.com/watch?v=yv6xovSsB1U&t=337s)
Palmer says the fastest way to build an asset today is through code or code generation, not only when building software. He uses React or TypeScript for nearly every ancillary asset in the presentation, including prototypes and other creative work. This surprised him because his background is in data engineering and Python. To make high-fidelity assets, he had to learn TypeScript, React, CSS, and HTML. He presents this shift as evidence that code has become a practical medium for communication.

### Structure is more expensive than code
[06:36](https://www.youtube.com/watch?v=yv6xovSsB1U&t=396s)
Palmer rejects the common answer that taste is the scarce resource. He says structure is expensive. That means maintaining a structured codebase, keeping brand guidelines consistent, using design tokens, separating front-end and back-end code from those tokens, and maintaining clean pull requests. It also includes tagging pull requests, writing descriptions, distinguishing features from bug fixes, tracking reverts, and keeping internal documentation accurate. Without this information, even capable AI agents cannot reliably solve organizational problems.

### Conscientiousness separates polished output from AI slop
[07:35](https://www.youtube.com/watch?v=yv6xovSsB1U&t=455s)
Palmer defines conscientiousness as being meticulous, careful, and guided by professional duty. As AI handles more of the work of creating assets, the human responsibility shifts toward checking whether the result matches expectations. He says the important difference is increasingly less about being the most technically skilled person in the room. AI rewards conscientiousness, organizational excellence, and structure. The assets he showed were produced from design tokens, structured code, and structured assets, and become much harder to reproduce when those foundations are missing.

### Content production needs a structured source of truth
[09:09](https://www.youtube.com/watch?v=yv6xovSsB1U&t=549s)
If code is the source of truth, Palmer says that source must itself be structured. Content can move left into the codebase only when the code contains the organization, design rules, and documentation needed to generate consistent communication. He connects this to organizational excellence rather than to an isolated prompting trick. The quality of generated content depends on how carefully the team creates and maintains the software from which that content is derived.

### Content engineering will turn manual assets into repeatable pipelines
[09:49](https://www.youtube.com/watch?v=yv6xovSsB1U&t=589s)
Palmer predicts that the strongest communicating teams will add discipline to software development and shift content production toward code. He calls 2026 the year of the creative technologist in DevRel and predicts that 2027 will be the year of the content engineer. He expects declarative content pipelines to produce walkthroughs, documentation, screenshots, and product updates through React and AI. These pipelines would replace many manual production steps with outputs derived from the codebase.

## Notable quotes
- "Code is becoming the way that we communicate." (00:15)
- "Structure is expensive." (06:56)
- "AI rewards conscientiousness. AI rewards organizational excellence. AI rewards structure." (08:38)
- "In order to communicate, we need structure and conscientiousness around the way that we create code." (09:24)

## Tools & references mentioned
- Conductor
- React
- Remotion
- TypeScript
- CSS
- HTML
- Anthropic
- OpenAI
- Replit

## Who should watch
- You produce documentation, product updates, demos, or DevRel assets and want to understand what can move into code.
- Your team uses AI to generate interfaces or content, but the results vary because design rules, documentation, or code organization are incomplete.
- You are building internal tooling or content pipelines and need a practical argument for treating structure as part of the product.