Expanding Kosmos’ Data Access Capabilities
tony kulesa just RT’d endorsement, it has much of bioxiv
related K-Dense
it’s nice to stock up some prompts to use them in batch later…
https://app.k-dense.ai/share/session_20260119_113716_05b0fba1834d
This will be incredibly powerful but I suspect there is a real art to crafting an effective prompt.
[and he is one of the smartest MOST visionary people in the area, along with patrick hsu]
https://x.com/rbhar90/status/2016239480458657953
==
higher-order knowledge representations for agentic scientific reasoning: https://arxiv.org/pdf/2601.04878
AI has fundamentally changed coding and research, at least 5Ă— faster.
Publishing papers alone matters less now.
The only real standard is: can your work be used to train or improve AI models?
Introducing Prism, a free workspace for scientists to write and collaborate on research, powered by GPT-5.2. Available today to anyone with a ChatGPT personal account: https://prism.openai.com
Speaking of Prism… I tried Prism a few days ago. It might be bad news for Overleaf as a company, but I’ll continue using Overleaf to write my papers.
Prism did a good job on some of my tasks. One was to take a kind of “pidgin LaTEX” that I sent in an email to a grader and turn it into a readable, presentable solution sheet that students could read. It worked very well, and made good stylistic decisions.
Next I asked it to create a “makeup exam” of problems similar to ones I gave before, and it did less well. It’s good at solving problems, but less good at extrapolating to come up with equally interesting problems in the same general area.
Addendum: It doesn’t do particularly well at writing solution sheets – it screws up a lot (maybe it uses an older GPT-5 model, like GPT-5.1 or something, as it missed almost every single problem, and even after I gave it hints!). I also tried to see if it would help with adding tags and metadata to LaTeX files to make them Title II accessibility compliant (a huge pain in ass right now in departments across the country!), and it nailed it for the first file, then failed with the second one. If OpenAI could just nail that one thing (making Latex documents Title II compliant), that would be a huge benefit, putting them ahead of Overleaf.