Why Isn't AI Coding Showing Its Applicability Across Repos?

- 4 minutes read - 678 words

Back in April, word got out about an unreleased, highly capable version of Claude called Mythos that, when pointed at large codebases, could surface serious, previously undiscovered vulnerabilities — including in software that was decades old. It was good news copy and the press fanned flames for several news cycles.

But after having done a lot of AI-assisted engineering through Summer 2025 and doing a lot of AI-assisted process management at work in 2026, I have to ask:

If the current models and tools are Ragnarok-grade deadly, why isn’t it more visible doing basic engineering?

Or:

If the current models and tools are so sophisticated e.g. Claude Code for Design, why is Claude Code on the desktop so damned ugly both aesthetically and engineering-wise?

In short, there’s a lot of visible lack of dogfooding of the AI tools that makes their claims look hollow. I’m not saying the emperor has no clothes, but at this point he’s doing a hot monarch summer in lamé booty-shorts and a crop top. The bungling is so obvious it makes me wonder what level of dysfunction has beset Anthropic’s product management and marketing.1

Here’s a list of bungled market communication.

Establish Engineering Credibility

A product push that could move the conversation is this (and I use Claude Code because I’m the most familiar with it):

If the product claim is that achieving ambitious projects is merely a question of throwing tokens (i.e. computation tokens) at the problem, a raft of successes here will be better marketing to the people who matter than a Super Bowl ad.

Examples:

  • Why hasn’t Anthropic ported Claude Code to FreeBSD? If the retort is “small market share, other priorities,” then my counter is: “Exactly. But with your so-called brilliant tools, those arguments shouldn’t matter, right? Right?”
  • If Claude Code is so good at engineering, why is Claude Code written in Electron: a framework designed to help web development developers port their skills to the native OS application space. Why are you shipping a glorified Chrome browser session instead of a native framework-exploiting, elegant, market-leading solution? Why not wow us with a Claude Code on macOS experience that looks good enough to have from Cupertino?2
  • If Claude Code for Design is so good, launched Summer 2026, why is the Anthropic desktop tool such junk visually?

This is a complete failure at dogfooding: using your tool in a visible way that demonstrates that you are a booster of it and that it is giving you competitive advantage.

That Anthropic’s product team is bungling this is curious. Either Engineering runs the shop without thinking like a business, or top-level executive management isn’t aligning the department heads effectively.

Here’s what happens when your marketing and engineering teams don’t align and don’t dogfood: taste makers and experience engineers scoff. Here’s Dave Copeland scoffing.

Conclusion

In short, if the tools are as transformative as the marketing says, the companies with the most access to them, the most budget, and the most obvious incentive to show off should be the ones with the best-feeling apps. Instead they’re punting on the exact thing they’re telling the rest of us is now trivial.

Footnotes

  1. Dario, baby, hit me up. For a cool one-time $2mm cash payment, I will personally get out of bed in the morning and go fix your product execution.
  2. I know the real reason, because John Gruber already made it in response to a similar complaint from Nathan Bruenig. Some of the engineers who built the app literally helped create Electron in the first place. The marketing on this angle could be so much better: The guy who invented this has no fear of working in languages he doesn’t know because of our awesome tool — the same one we’re trying to sell to you! ——