open source just beat claude on legal tasks at 11x less cost.
but who’s paying the people who maintain the models that made this possible?
harvey tested open-source models against frontier ones on real legal work:
• GLM 5.1 Opus advisor: higher quality AND 60% cheaper
• post-trained Kimi K2.6: matched quality at $84 vs $954. 11x cheaper.
>
@ClementDelangue is right - the “frontier model wins everywhere” narrative is marketing.
YET no one is talking about the following:
> enterprises are structurally dependent on open source AI
> these are production workloads running through models maintained by teams that could disappear tomorrow
> there is no product today that guarantees open source infrastructure stays maintained
i believe the narrative for financialization of open source is coming AND the gap between how much value open source creates how much the maintainers capture is the biggest unpriced externality in tech
@paoloanzn is building toward this at
@freecodexyz - financing open source engineers for verified contributions and working on a model for enterprises to guarantee maintenance of their critical dependencies.
Imo this is going to be one of the defining narratives in crypto x open source.
if you’re an OSS project or developer - whether you’re building for the internet economy, agentic commerce, or onchain (solana, base, or any other ecosystem)- reach out to me directly.
we’re actively looking for projects to work with. DMs are open.