How many times have people posted here about how software is no longer optimized because CPUs keep evolving, making programmers lazy? It seems things have changed.
The question is: will LLMs manage to squeeze the most out of this hardware, better than compilers do?
In the few tests I’ve run, yes, a lot.
As a programmer and founder, I think the idea is incredible, I would just change the understanding of "Code", given that what we've been hearing most lately is that "a markdown file is all you need".
I think it's not too far-fetched to think about standards, cultures, guardrails, compliance, etc. being documented, versioned, but more importantly, verifiable and applicable. In natural language, no code needed.
Researcher Tatiana Coelho de Sampaio, PhD professor, who discovered polylaminin, a drug that has proven capable of reversing spinal cord injuries in humans, worked in silence for 25 years alongside a team of biologists to achieve this breakthrough
The text reminded me of one of Veritasium's latest videos [1] about power law, self-organized criticality, percolation, etc... and it also has a wildfire simulation
Node-based workflow for AI generation seems to be the right approach. Being able to chain different models (Flux for realism, Sora for video, etc.) and do actual editing in between steps is way more useful than single-shot prompting. The ComfyUI comparison is obvious but this looks a bit more polished. The branching/remixing workflow could be interesting for iteration. Also being able to create a entire workflow with just one prompt would be nice.
Congratulations on your work. I spent the day working with a mix of the Composer/Sonnet 4.5/Gemini 2.5 Pro models. In terms of quality, the Composer seems to perform well compared to the others. I have no complaints so far. I'm still using Claude for planning/starting a task, but the Composer performed very well in execution.
What I've really enjoyed is the speed. I had already tested other fast models, but with poor quality. Composer is the first one that combines speed and quality, and the experience has been very enjoyable to work with.
Good point. Many people (including me) switched to Apple Silicon with the hope (or promise?) of having just one computer for work and leisure, given the potential of the new architecture. That didn't happen, or only partially, which is the same.
In my case, for software development, I'd be happy with an entry-level MacBook Air (now with a minimum of 16GB) for $999.
OpenRouter should be responsible for this quality control, right? It seems to me to be the right player in the chain with the duties and scale to do so.
I see some pessimism in the comments here but honestly, this kind of product is something that would make me pay for ChatGPT again (I already pay for Claude, Gemini, Cursor, Perplexity, etc.).
At the risk of lock-in, a truly useful assistant is something I welcome, and I even find it strange that it didn't appear sooner.
I wish they had published what prompt was given to Claude to improve GPT-5-mini's performance, as well as a before and after comparison of a prompt that underwent this transformation.
I think the DBSC is on the right direction but while it generates separate keys per session to prevent cross-session tracking (Google's ultimate ad dream), the spec acknowledges a critical vulnerability: malicious sites can collaborate by attempting to guess public keys until they find matches, creating persistent cross-site user identifiers, essentially weaponizing the security feature into the ultimate tracking system that survives cookie deletion and VPN usage
Herbs JS author- https://herbsjs.org/
meet.hn/city/br-São Paulo
Socials: - github.com/dalssoft
Interests: AI/ML
---