Alex Cabrera retweeted
a few observations on value creation + neoCROs (aka bio data factories, aka lab-in-the-loop)
> there is an intense thirst for continuous data generation from frontier labs and neobio labs
> but severe dislocation between what pharma wants and what frontier labs want
>> pharma wants to augment existing data streams, with wet lab validation and benchmarking
>> frontier labs want net new data sources that scale fast, with lab-in-the-loop validation
>>> frontier labs are a shortcut to revenue and real-time experimentation, and what they're really buying is know-how and data
>>> pharma is slower, but it's the bigger and longer budget, given how messy and guarded its data is and how bespoke its needs are - many GTM motions will be FDE-like
>>>> there is transitory risk in leaning on frontier labs is that your proprietary knowledge ends up in their weights and you become the harness for someone else's model, which is a place to sit
>>>> those that bridge both worlds (harnessing, evals and model dev with frontier + neobio labs AND making the data and infra legible to large pharma) will be the long-term price setters and market makers
>>>>> but the ultimate moat is the pipeline itself: royalties and milestones on assets (who will be the first AI model co to do this?) - either co-developed by frontier labs + pharma, or internal (the neopharmas emerging).
the future is everything saved on s3 with
- pooled writes through a server for high write volume applications
- warm cached data on server for serving live applications
246
We've found Argon is impressively good at scientific tasks, especially dealing with raw, messy experimental data like cell painting and transcriptomics. Google is back at the frontier!
5
359
Alex Cabrera retweeted
This one vague tweet started one of the greatest virality stories for tech Twitter.
A startup no one had heard of was in a span of merely 13 days, worth $2.5B.
We investigated 5k+ posts and 56k+ relationship ties to find what actually made it "one of the greatest".
thread 🧵
Alex Cabrera retweeted
As frontier models become open and affordable, durable value moves to the scaffolding around them – the data, the orchestration, the workflow.
Instead of being locked into a vertical platform, developers can now bring intelligence to the harness of their choice.
This changes everything: imagine thousands of Claude Code moments purpose-built for every domain – legal, biotech, manufacturing, finance, insurance.
This is the inflection point of Kimi-K3. The model was never the moat, and now intelligence is being unleashed to every developer in the world. Hard to overstate how powerful this will be for the open model ecosystem.
Kimi K3 may be an important inflection point for AI. Potentially negative for Anthropic and OpenAI while being net positive for essentially every other company in the world. I mean that very literally. Although the real “Sputnik moment” would be an open-source frontier model that was also token efficient unlike Kimi K3 which is 50-70% more expensive to run than GPT 5.6 per Artificial Analysis.
Rationale:
A world where there are only 2-3 dominant frontier labs with 90% inference margins is net negative for every other layer while being awesome for those 2-3 labs. Those labs would become monopsonies for power, data centers, semiconductors and hyperscalers and would obviously vertically integrate over time into all those layers while also completely subsuming the application/software layers.
Anything that lowers margins and increases competition at the model layer is good for every other AI layer: power, semiconductors, hyperscalers, neoclouds and yes even software.
This is why Jensen is so supportive of open-source. An open-source model requires the *exact* same amount of compute to run as a closed frontier model of similar size and architecture. Kimi K3 is roughly the same price as GPT 5.6 Terra on a per token basis, which actually suggests that it is less computationally efficient as I am sure that GPT 5.6 is priced to a higher margin than K3. And given that K3 is a token wastrel, i.e. token inefficient, it is significantly more expensive per task than GPT 5.6 and Grok 4.5, which are much more token efficient. Cost per token and token efficiency (i.e. intelligence density per token) are the drivers of intelligence per unit of cost. The winning AI companies will be those that offer the most intelligence per $ over time.
Lower margin % at the model layer = more margin $ at every part of the infrastructure layer and is a godsend for software. This can happen either through open-source models like K3 at the frontier *or* having a vertically integrated model company like Meta, SpaceX or Google at the frontier. Both outcomes result in a lower margin % at the model layer as vertically integrated model companies don’t really care where the margin $ come from. This is why it was so painful for OpenAI and Anthropic when Google was right there with them from a model competitiveness perspective and why Grok 4.5 and Muse 1.1 were just as important as Kimi K3.
The reason Kimi K3 is only *potentially* negative for Anthropic and OpenAI is 1) the @ericvishria point that the Claude and ChatGPT products and harnesses may be more important than their models today and 2) the hypothesis that they have much more advanced model checkpoints internally that are already being used for RSI. In the latter scenario, reaching RSI even a few months ahead of other labs might be enough to cement a permanent lead.
Time will tell on both points. And likely fairly quickly.
Caveat would be that since Kimi K3 is not token efficient and thereby actually more expensive than ChatGPT 5.6, we may need to see a more token efficient open-source model at the frontier or see Grok 5/Composer 4/Muse 2 at multiple points on the Pareto frontier for this potential risk to Anthropic and OpenAI to play out. And I am sure they will both vertically integrate as quickly as possible while continuing the product/harness strength they have shown over the last 8 months.
love the information theory framing on something we've been seeing empirically for small molecule ML
New blog post on estimating the Shannon entropy per atom of various therapeutic modalities, and why this makes me pessimistic about ML models predicting interactions with arbitrary small molecules. There are just too many molecules out there!
corinwagen.github.io/public/…
1
15
3,114
Egregious security theatre. We can't ask a single question on our code base without getting blocked.
Replying to @claudeai
Fable 5’s safeguards detect requests related to cybersecurity, biology and chemistry, and distillation. Users are informed whenever a fallback occurs—on average in less than 5% of sessions.
We’ll keep refining the safeguards to reduce false positives.
It's been quite exciting using gpt-5.5 in our experimental data -> human outcomes agent
Introducing GPT-5.5
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex.
3
251
The next evolution of our autonomous research agent is here. Today, we’re introducing Deep Research and Deep Research Max via the Gemini API.
Powered by Gemini 3.1 Pro, you can now trigger comprehensive research workflows with unprecedented control and transparency, featuring:
🔌 Arbitrary MCP support
📊 Native infographic & chart generation
🌐 Fully cited reports grounded in the open web + your own files and data
All from a single API call. Meet the new agent 🧵↓
This is a fun idea. I hate the GH email and UI updates for comments, review requests, etc.
Took 30 min to get a nice toolbar app w/ your PRs and reviews, read status, etc.!
I've been vibe coding SwiftUI menu bar apps for my new Mac, turns out Claude Opus 4.6 and GPT-5.4 are both competent at Swift programming, no need to even open Xcode! simonwillison.net/2026/Mar/2…
1
4
543
the big labs should release alexa/google home competitors. make plugins/mcps for home devices. it can remind you if you forgot to program the coffee, update the lights dynamically based on your calendar, etc.
1
176
Alex Cabrera retweeted
1/ Most AI features die in demo-land.
The teams that actually ship have a system:
(1) evals
(2) build vs buy (on repeat)
(3) org design
amplifypartners.com/blog-pos…
Alex Cabrera retweeted
I wanted to share something I built over the last few weeks: isometric.nyc is a massive isometric pixel art map of NYC, built with nano banana and coding agents.
I didn't write a single line of code.
Alex Cabrera retweeted
For a decade, techbio optimized the drug discovery "paper mill": more molecules, faster. But the hard part - knowing which ones will work safely in humans - hardly moved at all.
Late last year, I spent time with Axiom Bio, a company built around the idea that the real bottleneck is predicting clinical activity in molecules, *not* merely discovering them.
Alex Cabrera retweeted
Announcing Worktrunk! A git worktree manager, designed for running AI agents in parallel.
A few points on why I'm so excited about the project, and why I hope it becomes broadly adopted 🧵






























