I used my org2html package to recreate the content of the @claudeai article introducing Opus 5.5 on my demo blog, showcasing the capabilities and what we can achieve with the package so far.
This is a real-time demo using org2html watch command to see the changes each time I save the .org file.
@DanielSmidstrup It's just for the critical things, you can use plan mode, ask for the gaps and questions then join in a loop and avoid those questions if those where already resolved.
I’m not a design expert, although I’m always learning and improving, and just for context, I made this logo about 8 years ago, and I’ve been using it ever since.
I’d love to ask for your help and hear what you think about it. Are there things you would change, or any advice from your experience that I could apply to the rebranding I’m planning?
I just want to have possible insights from people that have much experience on logo and brand design.
I’ll say it again, I’d like to see a future where you can use a model tailored to your needs, running locally, solving your specific use case without paying for or wasting compute on capabilities you’ll never use.
Frontier models are obviously incredibly capable, but in many cases, you’re using a model that can do millions of things you’ll never need. Why use all that compute when a smaller, specialized model could deliver the same or even better results for your specific problem?
It’s basically the AI version of "don’t use a bazooka to kill a bug".