.png)
Why we signed the Open Weights and American AI Leadership letter, and what it means for the people building products.
This week, Agno joined Microsoft, Meta, NVIDIA, Hugging Face, Vercel, and others in signing the Open Weights and American AI Leadership letter.
Agno is unapologetically pro-builder. As a team of builders building a platform for builders, we have carried that in our DNA since the first commit. And that conviction gives us a simple test for issues like this.
We ask one question: is this better for the people building products?
Open weights pass that test. They give developers more choice, more control, and more room to create. That's why we signed.
The debate around open weights usually centers on policy, regulation, and competition between model providers. Founders and engineers care about something simpler: the freedom to choose the right model, deploy it where they want, avoid lock-in, and evolve their systems as AI changes. Open weights put every one of those decisions back where it belongs, with the person building the thing.
Every major software movement expanded developer freedom
The technologies that shaped modern software share one trait: they gave developers more control. Linux let organizations own their infrastructure. Python made software easier to build. PostgreSQL gave teams a world-class database without proprietary licensing. Kubernetes made applications portable across clouds. PyTorch accelerated AI research by putting cutting-edge techniques in everyone's hands.
None of them won because they were open for openness' sake. They won because they removed barriers and gave builders more choice.
Open-weight models follow the same pattern. Instead of reaching intelligence only through proprietary APIs, organizations can run models on their own infrastructure, evaluate several options side by side, fine-tune them for their workloads, and decide for themselves how to deploy. That flexibility drives innovation.
AI moves too fast for lock-in
The landscape shifts every few months. New frontier models arrive, open-weight models close the gap, reasoning improves, costs fall, and inference gets faster. The model that wins today may not win six months from now.
Building your product around a single provider assumes today's decisions will still make sense tomorrow, and history suggests they won't. The companies that succeed won't be the ones that guessed correctly once. They'll be the ones that adapt quickly as the landscape moves.
Open weights make that adaptation possible. You can evaluate new models as they emerge, compare them against your existing stack, and switch when a better option appears. It's not just about saving money. It's about keeping the ability to make technical decisions based on what your product needs rather than what one vendor happens to offer.
More choice creates better AI
Competition accelerates innovation instead of slowing it. When builders can reach multiple capable models, providers have to compete on performance, quality, latency, efficiency, and developer experience. Infrastructure providers compete to make deployment easier. Tooling sharpens, benchmarks get more rigorous, and applications improve.
Choice also changes how you spend. With one model, everything routes through it and you pay frontier prices for work that doesn't need frontier capability. With several, you match the model to the job: a small specialized model for the high-volume work, a frontier model for the problems that actually earn it. That discipline is what keeps AI economically sustainable as usage scales into millions of everyday tasks.
Everyone moves faster because no single company controls access to the underlying technology. Open-source software followed exactly this path, so it's fair to expect open-weight AI to follow it too.
Enterprises need control
For many organizations, the question isn't which model performs best. It's where that model runs. Healthcare providers, financial institutions, governments, and large enterprises carry hard requirements around data residency, privacy, security, and compliance, so running models inside their own environment becomes a requirement rather than a preference.
Open weights give them that option. They can choose infrastructure that meets their needs instead of reshaping their needs around someone else's platform. As AI works its way into critical business systems, that flexibility matters more every quarter.
The future belongs to builders
Open weights matter because they give builders agency, not because they cost nothing. The future of AI won't belong to one model, one framework, or one company. It will take shape as an ecosystem of competing models, fast-moving capabilities, and developers picking the right tool for every job.
That future is healthier for the industry, better for innovation, and better for businesses. Most of all, it's better for the people building products. We support open weights not because everyone should build the same way, but because everyone should have the freedom to build the way that's right for them.