🤔 The two most viral posts on my channel are how I got a cat and how I fixed my MacBook with Claude. The architecture posts about 13 agents, MCP and N×M integrations get several times less reach. For a long time I thought it was the market being ironic. Then I realized it's not.
The problem Simon Willison solves with a dedicated tool:
Showing an AI project to people isn't about being good at explaining. In February 2026 Simon Willison (creator of Datasette, one of the leading practitioners in AI development) released Showboat + Rodney - two tools with a single goal: to let coding agents visually SHOW what they built. The agent creates demo.md with live screenshots, Rodney automates the browser - clicks, captures the screen, embeds the proof right into the document. Because MD files, Telegram groups and repositories are text. And text requires imagination that a busy reader doesn't have. One of the best AI practitioners in the world admitted that showing "what you built" is a separate unsolved problem. Not an everyday complaint, but a systemic development problem.
Why personal pain beats technical architecture:
📌 The neurobiology of stories - a personal story activates the same neural patterns as direct personal experience. "Fixed my MacBook in 3 minutes" = hero + pain + solution. A post about N×M integrations = useful data with no emotional arc
📌 Retention: video vs text = 95% vs 10% - when you show how something works, people remember. When you explain how it's built inside - they barely do
📌 Recognizable pain - a broken trackpad makes sense to anyone who owns a laptop. "Why you need MCP" - only to those who are already building agents and have hit that exact wall
What structurally separates the "cat" from the "13 agents":
💡 A story with an arc - a problem, then an attempt, then a solution, then a concrete result. An architecture post is data without an arc, without a moment of resolution
💡 A promise with concrete value - "in 3 minutes I fixed what cost $400 at a repair shop" works. "Cut complexity from N×M to N+M" - only for those who know Big O notation
💡 Built-in proof - a photo of the cat or a terminal screenshot = proof that doesn't require faith. An agent architecture diagram = a conceptual drawing for those who already get the context
The market isn't unready for complexity. The audience is ready for any complexity if it's wrapped in a familiar pain or live proof that the thing works. The cat went viral not because a cat is simpler than 13 agents - but because the post has a recognizable hero, a real problem and a clear resolution. The right question isn't "why is technical content weaker" - it's "how do I package architecture through a pain the reader knows". MD files and repositories stay for those who are searching - but only a live story attracts those who don't yet know what they're looking for
⭐️ Simon Willison, creator of Datasette:
"The key challenge of working with coding agents is both letting them test what they've built and showing it to you. I don't trust any feature until I see it working with my own eyes"
🤷 Three days ago I migrated 13 agents and wrote about MCP. Now I'm writing about a cat I got a year ago. And I already know which one will get more. I came up with a test for myself: if I can't explain what specific pain this removes for a person right now - the post won't take off, no matter how good the architecture inside is