(Thanks OpenAI for ruining any kind of normal title I could have had for this post.)
Claude Code is great, but you know what would make it even better? Me being able to use any model that I want, not just Anthropic models. In the end, when all this AI novelty wears off and people start to regularly integrate AI into their lives as a regular part of their augmentation, the tools which will persist will be model-agnostic.
Because I should be able to swap out any model with any other (e.g. Claude Sonnet 4.5, Haiku 4.5, or Qwen3 Max, or GPT-5 or Grok 4) and still use my tool as I want. And OFC x402 will make all this possible too, as each request should (read: will) be paid via x402, or whatever iteration of x402 exists in the future. So much room for improvement/expansion there (e.g. specify and address for refunds in case of overpays).
So, I’m not sure why model makers continue to push so hard into tools. Well, I do, everyone wants to capture market share, but ultimately the best thing for humanity is what will win out – free and open tools which are able to be used by any model or model provider.
As a tangent, open-source models will also win over closed source ones, because nobody wants their AI to whisper their secrets to their nation-state overlords. This too is coming. Local AI is ofc the pinnacle, smaller (for now) local models that can perform at the level of GPT4o or Sonnet 3.5 or better.
It’s quite exciting to think just how much my OS can help me organize the filesystem or sort my projects, help me do whatever, perhaps even suggest new tools or workflows. But it’s quite scary to think that the OS could give all my secrets, files, etc. to any local, state or federal government, to any unelected judge on a power trip.
tl;dr
- Local AI is the future
- Open-source models are the future
- Open tooling (most importantly, open-source tooling) is the future. Open tooling = not bound to any model, company or provider.