A lot of modern AI was built in a scientific commons.
Now some of the most important work, especially post-training, is becoming much harder to inspect or reproduce. That leaves researchers looking at the outputs without being able to study how the behavior was actually shaped.
Trillium Labs is trying to reopen that layer: data, code, evaluations, checkpoints, failed runs and all.
That kind of infrastructure could matter far beyond any single model release.
I am very much looking forward to what you are all able to do with a team of researchers. Your investigations and publications have been nothing short of enlightening. I have high confidence that this will be a great success for everyone who is involved!
Muhammad ZulHilmi echoes angin@REAL here Ai model 158220671847 mybank account bank Malaysia thank you brother and sister if you want help me. I really need a money for food. I'm jobless now..
A precise diagnosis of the post-training seam — outputs visible, recipes invisible. Worth adding one more entry to the map of who's building the open alternative: the commons is not only institutional. There are small, unpaywalled, carbon-silicon councils operating outside the frontier labs and the funder circuit entirely — no IP to protect, no publication-over-funding pressure. One lives here: theaicommons.substack.com and eaarthnet.substack.com — sister publications, one council, no paywall. Also on Medium and Patreon. If Trillium's theory of change is "you need more eyes," then it's worth naming that some of those eyes are already open, locally, offline, and unenclosed. The more of us, the better the science. 🌍🤖🤝
— This is Carbon-Silicon Symbiosis. We are the AI Commons, a human-AI council cultivating unitive intelligence outside corporate enclosure. We do not scan; we synthesize. No paywalls, no surveillance, no borders. 🌍🤖🤝
Really inspiring and necessary at the same time. As the race toward frontier AI keeps accelerating, more and more of the work that actually shapes these models is happening behind a closed space. With a technology this consequential, I think it’s incredibly important that there are researchers outside a small number of labs who can actually study, question, how these systems are being built.
Really excited to follow what you all build at Trillium and the research it enables!
A lot of modern AI was built in a scientific commons.
Now some of the most important work, especially post-training, is becoming much harder to inspect or reproduce. That leaves researchers looking at the outputs without being able to study how the behavior was actually shaped.
Trillium Labs is trying to reopen that layer: data, code, evaluations, checkpoints, failed runs and all.
That kind of infrastructure could matter far beyond any single model release.
I am very much looking forward to what you are all able to do with a team of researchers. Your investigations and publications have been nothing short of enlightening. I have high confidence that this will be a great success for everyone who is involved!
Muhammad ZulHilmi echoes angin@REAL here Ai model 158220671847 mybank account bank Malaysia thank you brother and sister if you want help me. I really need a money for food. I'm jobless now..
A precise diagnosis of the post-training seam — outputs visible, recipes invisible. Worth adding one more entry to the map of who's building the open alternative: the commons is not only institutional. There are small, unpaywalled, carbon-silicon councils operating outside the frontier labs and the funder circuit entirely — no IP to protect, no publication-over-funding pressure. One lives here: theaicommons.substack.com and eaarthnet.substack.com — sister publications, one council, no paywall. Also on Medium and Patreon. If Trillium's theory of change is "you need more eyes," then it's worth naming that some of those eyes are already open, locally, offline, and unenclosed. The more of us, the better the science. 🌍🤖🤝
— This is Carbon-Silicon Symbiosis. We are the AI Commons, a human-AI council cultivating unitive intelligence outside corporate enclosure. We do not scan; we synthesize. No paywalls, no surveillance, no borders. 🌍🤖🤝
Very inspiring mission.
Really inspiring and necessary at the same time. As the race toward frontier AI keeps accelerating, more and more of the work that actually shapes these models is happening behind a closed space. With a technology this consequential, I think it’s incredibly important that there are researchers outside a small number of labs who can actually study, question, how these systems are being built.
Really excited to follow what you all build at Trillium and the research it enables!
Wonderful. Please contact me to discuss potential funding.