Co-founder & the Dude @ FinalDose (YC P26) PhD @ Oxford Oncology Ending all cancers
San Francisco, CA
Joined February 2016
- Tweets413
- Following382
- Followers406
- Likes541
Pinned Tweet
FinalDose: We’re going to end cancer in the open
We plan to build FinalDose as public as we possibly can and write up as we go, including the parts that don’t survive! Please help spread the word:
@FinalDoseLFG
substack.com/@finaldose
Entirely true. Those thinking you gotta hire in the US because they work harder is crazy. Rate of progress is undoubtedly fastest in China by an insane margin.
For an AI-driven biotech founder, the bottleneck is not generating designs. It is verifying them and stacking wet-lab outcome data you can retrain on.
I saw this on China CRO site visits: design throughput is already ahead of assay queues. In my census of 262 funded US/EU biotechs, AI protein and antibody shops skew hybrid (~44%) for that reason. China built the service layer. My estimate from filings: 35–45k CRO in-vitro discovery staff vs 2.5–3.5k in the US. Operators staff 20 chemists in weeks; multiple shifts on one instrument set (my observation). @WuXi_AppTec: US customers = 72% of continuing revenue. @pharmaron: 3,300+ customers; top-20 only 20.1%.
Rule I would run:
• Keep design and interpretation in-house.
• Buy verification by the sample on that capacity until one assay line sits above ~40% utilisation (illustrative US workcell break-even).
• Prefer machine-orderable assays: fixed input, numeric output, public price — @adaptyvbio binding from ~$99–169/design; @Ginkgo ADME-One $199/panel (company figures).
Cheap design makes the wet lab busier, not idle. The scarce seat connects the model, the assay and the next decision. Can't agree more with that @LifeSciVC Bruce Booth’s Oct 30 essay "The Fruits of AI Drug Abundance: Indigestion In The Near Term"
Great value
Heart attacks are the #1 killer & almost totally avoidable! Early detection of plaque is key. Over 35 - get your CAC scan for $150 & 15 mins. Working closely w @American_Heart to make CACs more standard & accessible! Big changes coming that will save 1000s of lives. 🇺🇸 ❤️🤍
Jeff Liu retweeted
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
95% of our engineers already run self-improving loops. The next move is graphs"
In only 10 minutes, she builds her whole Claude Code setup live, right from an empty terminal
Empty Terminal → Agents → Loops → Graphs → Self-Improving System
Prompting was the old method. Engineering the system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the full Graph Engineering guide below ↓
lnkd.in/p/e-CUijf5
Go like comment repost support Canos!
Oxford startup YC F26 bringing human grade cancer care to pets. Lets go!!
Jeff Liu retweeted
This is f*cking insane. Anthropic engineers just gave a 2-hour Stanford lecture on how AI agents really work
The clearest explanation I've seen, whether you've never touched AI or you build with it every day
Chats → Prompts → Agents → Loops → Systems That Improve Themselves
Getting good at AI isn't about writing a better prompt every time
It's about building a system that learns from every run and gets sharper on its own
95% of people will spend 2 hours on Netflix tonight, the 5% who spend it on this will build what everyone else ends up using
Worth more than any $2,000 AI engineering course
Bookmark it and watch tonight
Then read the guide below with ready-to-copy prompts to build a system that improves itself ↓
Literally 100% global market share for programmable, sequence-activated cancer therapies in one pic. Historical photography
Started with us bantering over a post I made over Jingkun Zeng 's paper with Jennifer Doudna, linked up in Berkeley, and became friends ever since. We got the same wild ambition which is obviously why he flew into London solely for us. He didn't see any of his 50 other friends.
Wild @FinalDoseLFG repping half of the field
Jeff Liu retweeted
This is f*cking insane. Anthropic engineer:
"You're not supposed to prompt Opus 5.5.
You're supposed to build a system that prompts itself. At Anthropic, 90% of our engineers have already built these systems."
In just 30 minutes, she explains how Anthropic builds AI agents that remember context, learn from mistakes, fix their own errors, and improve with every run.
This is the shift from simply using AI to engineering systems that work autonomously.
Honestly, it's better than most $1,000–$2,000 agent courses I've seen.
Watch the video, then save it for later
SF has the highest density of grind, distraction, larp, random shit, coolest shit, literal human shit postcode ever
You can always create constraints even in a resource abundant environment. In fact, you must find them to accelerate progress
Jeff Liu retweeted
It’s been a while! Like every other biotech nerd, I’ve been feeling quite context-bloated as of late, and writing is the only fix I know. So here's a long one on AI and clin dev - if you run trials for a living, I’d appreciate any criticism and feedback :)
The only way to tell if you're in SF is if your uber driver also knows about agents and posttraining
Heading to SF from Nov 6 - 15 for conferences, love to meet with founders in the sci-fi space! Especially interested in people who've done data sale to frontier labs, we're producing quite an interesting set of data with some labs approaching us
No clue how this works so love to learn! @FinalDoseLFG
How have the west gotten to this place today? The f ed up thing is all I can see from all my social circles point to an easy win for anyone that stands with progress, less tax, and tighter border control but election results across board is the opposite EVERYWHERE.
I need another account to see what the other side is chanting
