Timothy Kassis

Co-Founder & CTO @k_dense_ai | Ph.D. in Bioengineering. Previously @biostateai, @matterworks_bio, @MIT, @GeorgiaTech.

[Timothy Kassis @TimothyKassis]

Pre-registration is the TDD of science. Lock your hypotheses and decision rules before you look at the outcomes, and p-hacking has nowhere to hide. We built Science Superpowers to give research agents that discipline by default. Inspired by Superpowers, reimplemented for doing science with data. https://t.co/2bI1u4ykDE

[Timothy Kassis @TimothyKassis]

20 years later and I'm still waiting for a Kindle with a color E-Ink display. Is it ever coming?

[Timothy Kassis @TimothyKassis]

https://t.co/xJp4I5sO6A

[Timothy Kassis @TimothyKassis]

AI co-scientists are getting powerful fast. The real question for scientists is no longer “can this generate an answer?” It’s: can I verify the path from question to conclusion? We wrote about why the next bottleneck for AI in science is verification, not capability.

[Timothy Kassis @TimothyKassis]

@dr_alphalyrae Definitely true

[Timothy Kassis @TimothyKassis]

@DynamicWebPaige @GoogleDeepMind Nice set of skills! We're doing some benchmarking against our own Scientific Agent Skills. Will share soon. But do take a look at our scientific community validated skills with over 27k GitHub stars! https://t.co/cFLvrfjK0J

[Timothy Kassis @TimothyKassis]

Full blog post and code: https://t.co/TAjNPu3l0q

[Timothy Kassis @TimothyKassis]

I kept watching frontier models write beautiful code that any senior scientist would catch as wrong in ten seconds. Not because the model lacks knowledge, but because it reasons like a generalist when science needs a specialist. So we at @k_dense_ai built Scientific Agents: 503 open-source profiles, one per field, that teach any AI agent to think like an actual practitioner. The controls they'd never skip, the standards they hold to, the failure modes they rule out before trusting a result. It's a single AGENTS.md file that drops into Cursor, Codex, Claude Code, or Copilot, and it's free. I'd love for you to try it on your own work and tell me where it falls short.

[K-Dense @k_dense_ai]

We just released scientific-agents. 503 expert-thinking profiles that turn AI coding agents into domain-aware scientific collaborators.

Not “be rigorous” prompts. Actual AGENTS.md files for biologists, physicists, engineers, clinicians, chemists, geoscientists, and more. 🧵

[K-Dense @k_dense_ai]

We gave our research agent 78 public scientific & economic databases (PubChem, ChEMBL, FRED, Ensembl, the FDA...) through ONE Agent Skill, not 78.

Everyone else ships one skill per database. Here's why we didn't, with token + routing benchmarks. 🧵

[Timothy Kassis @TimothyKassis]

@aasiri Also check out @k_dense_ai

[Timothy Kassis @TimothyKassis]

@k_dense_ai @antigravity 💯

[Timothy Kassis @TimothyKassis]

@antigravity Also try Antigravity with Scientific Agent Skills! https://t.co/cFLvrfkhQh

[Timothy Kassis @TimothyKassis]

Pre-registration is the TDD of science. Lock your hypotheses and decision rules before you look at the outcomes, and p-hacking has nowhere to hide. We built Science Superpowers to give research agents that discipline by default. Inspired by Superpowers, reimplemented for doing science with data. https://t.co/2bI1u4ykDE

[Timothy Kassis @TimothyKassis]

Try it here: https://t.co/msjjdKxor2

[Timothy Kassis @TimothyKassis]

We just integrated Paperclip from (@gxl_ai) into K-Dense BYOK. You can now work directly with scientific papers, preprints, clinical trials, FDA docs, and regulatory sources, making it much more useful as a local BYOK research assistant. Paperclip is substantially more token efficient than our paper-lookup scientific agent skill. It's all free. AI co-scientist on your desktop. Would love to hear how people are using it and how we can improve it for everyone.

[Timothy Kassis @TimothyKassis]

@NVIDIARobotics @UnitreeRobotics @SharpaRobotics We'd love to put this in a wet lab someday!

[Timothy Kassis @TimothyKassis]

@botirkhaltaevv @modal Incredible company!

[Timothy Kassis @TimothyKassis]

@JeremyNguyenPhD Nice, also suggest taking a look at https://t.co/uUtudH4weM

[Timothy Kassis @TimothyKassis]

Anthropic is going public!

[Anthropic @AnthropicAI]

Anthropic has confidentially submitted a draft S-1 registration statement to the Securities and Exchange Commission.

Pending completion of SEC review, this gives us the option to pursue an initial public offering.

Read more: https://t.co/onGZAhRLvD

[Timothy Kassis @TimothyKassis]

How well models work with agent skills is directly related to how good they are at instruction following.