A bit of meta: Claude Design -> Claude Code or Codex.
IMO Claude Code almost feels like a hack for getting web-ish products off the ground. It does such a great job of making the UI mockups interactive and providing a great "base" that the agents can use. Once you get that "base" the way you want in Claude Design, export it, then literally just tell your agent "Hey, the exported design project is /at/this/path - read it and implement it". No it's seriously been that simple - either agent can handle it. From there, you can A/B test to see which agent you prefer. The better agent is going to depend on your project and your prompting.
My final answer on this is that we just can't say anything affirmative because all of our projects/codebases are completely different. I've gone back and forth on the "codex vs claude" being better, and while I'm currently of the believe that Claude is superior, I understand that might be the case for _my_ particular set of projects and _my_ personal way of interacting with the model.
Meh, seems he's using arbitrary metrics to make arbitrary claims (which is fine). But to just state that "Life is so much better in 2026 than in 1926 for Americans" is obviously a pretty nebulous statement. It's like saying "Beaches in 2026 are so much better than beaches in 1926". Sure you could cherry-pick some metrics to make the case, and someone else could cherry-pick metrics to make the opposite case. Sort've a "talking just to hear yourself talk" kind've thing.
I think Elon owns a lot of real estate in these guys’ heads. When you have $200B+ in the bank, little else matters others than being seen as a titan of industry IMO. I think the fact that Musk has Tesla and Spacex (even with all his extra curricular activities) makes a guy like Zuckerberg seethe that he’s just “the internet websites guy”.
TLDR: They're promoting a product they're working on with Cloudfare under the guise of it being an "open standard" [1]. Of course, in the docs, Step 1 is "Sign in with your Cloudfare account". Comes across a bit land-grabby.
As someone who's explored Open Design recently, I think that's about as close as we're going to get for now unfortunately. I agree it't not really on the same level as Claude Design.
Doesn't CLI use MCP (if need-be) ? I think you might attempting to say "Neat, discrete buckets of work?" (i.e., MCP - similar to HTTP in that work is broken into discrete canonical buckets), or "Virtually unbounded work made available via various CLI utilities?".
I would pick the second option. CLI.
However, for CRUD B2B SaaS I think MCP works fine (if not better than CLI).
Things are time bound by instruction creation - at some point you still need a human to dictate the instructions that the orchestrated agents use. From there I've found that -- (1) derive a goal from the instructions (2) break that goal into tasks (3) order those tasks into a DAG (5) spawn the agents to work via the DAG -- seems to be doing everything I want it to do.
Working the same, the nature of the work has changed. Less time spent on the minutia of syntax and project scaffolding. More time spent on how the minutia compose into a larger system.
True, but I think the point I'm trying to make is that when it comes to deploying (what are more often than not) web services, getting to the point with systemd where it "just works" requires more pain than I'd like - especially with regard to production deployments (reading logs, checking service status, wondering why my env vars aren't being read, etc).
If at the time when I was cutting my teeth on systemd, I had access to something more lightweight and "do one thing well", I think I would've gotten a lot more sleep :)
I'm personally tired of getting stuck in config/deployment hell every time I want to deploy a long-lived web service. Sure I eventually learned how to use systemd, but systemd has SO many things baked into that I simply don't need. systemg is a lightweight process supervisor that features everything you'd typically want when running/managing production web services in the wild.