Tim Dettmers argues the unit of research is now the ecosystem, not the paper

Tim Dettmers·medium signalXBlueskyLinkedInCopy link

Dettmers opened his September 21 post with a classroom poll where about 120 of 150 students raised their hands to say they fear not getting a job after graduating, and uses it to argue academia is about to have a renaissance precisely because of its limited GPUs. His concrete claim is that agents made individual research projects cheap enough that piecemeal papers stopped being good research: it is no longer hard to publish a paper, it is hard to publish a coherent ecosystem, which is why his lab is shipping a week of interlocking open-source components instead of prose. The post accompanies dlab's Open Source Week, framed as running frontier AI on your own hardware.