I built phone-use CLI. computer-use for phones.
I gave Claude Code a bug and two phones. It reproduced it on iPhone and Android in parallel, fixed my code, and showed me the fix.
It also installs your build, records the screen, and is self improving.
Open source. Link below
1) Today we're announcing Atomwell, our state-of-the-art foundation model for predicting experimental outcomes, alongside scaling laws for molecular discovery.
We trained models from 800M to 9B parameters on ~80M experimental measurements, and found clean scaling laws in both model size and data.
We built an 8000 sqft. high-throughput wet-lab in 30 days, to solve molecular discovery - starting with pesticides.
Most teams in drug-discovery don’t receive target-organism verification until clinical trials. Pesticides allow us to rapidly validate our model in the target organism directly. With the help of our high-throughput lab, we were able to generate compounds that killed a $13 billion pest, the Fall Armyworm.
We believe that the bitter lesson is coming for biology.
I've gotten to play with this new app and it's a really very cool, very interesting design tool. Okay, I'm biased, Character invested in Subframe. But I mean come on they have a screenshot of Mezzanine by Massive Attack in the video so you know they're legit
Meet the new Subframe: built for designers who code.
Most of your work is now in code. But when you need to explore or nudge the details, prompting isn’t enough. You need a canvas.
Now moving between code and design is as simple as copy / paste.
Meet the new Subframe: built for designers who code.
Most of your work is now in code. But when you need to explore or nudge the details, prompting isn’t enough. You need a canvas.
Now moving between code and design is as simple as copy / paste.
Today @inductive_bio is launching Beacon-2, which can predict human dose directly from chemical structure.
A dirty secret of AI for drug discovery is that most systems optimize the wrong reward, making them near useless in practice. Human dose is the right reward – it’s what every drug program is already optimizing. Making it computable will unlock the real-world impact of agentic drug discovery.
Beacon-2 builds on our Beacon-1 models that beat 750+ competitors to win all three OpenADMET blind challenges. Whereas Beacon-1 provided a suite of models for individual molecular properties, Beacon-2 is a single system that predicts human dose end-to-end.
It has three configurable modules: an ADMET-PK foundation model, a potency module that can use our fine-tunable co-folding affinity model or physics-based methods like FEP, and a PK/PD model with physiologically-informed mechanistic models of pharmacokinetics.
We share validation results from Beacon-2 in a blog we are releasing today. We show that across 20 real-world drug programs running through Inductive, dose predictions from Beacon-2 agree with those from experimental data with a Spearman’s ρ of 0.61. And in a similar study on 325 publicly available compounds from the ExpansionRx OpenADMET competition, we show a Spearman ρ of 0.79.
As a proof-of-concept, we gave Beacon-2 to Indy, our medicinal chemistry agent, which used it over 5 autonomous design cycles on a recently disclosed SARS-CoV-2 program to cut predicted human dose 17x, which can be the difference between a drug program that gets killed and that makes it to clinical trials.
We compared the compounds that Indy designed using Beacon-2 against compounds from an optimization loop that used a potency-only reward function. The Beacon-2 compounds were favorable across three independent measures: they were preferred by chemists in blinded comparisons 87% of the time; were more druglike according to QED; and had a lower projected dose when computed by alternate methods relying on in vivo rat models and allometric scaling. Based on these encouraging results, we are now running prospective studies where compounds optimized by Beacon-2 will be validated in the lab.
Read the full post here: https://t.co/nVgAinJ9nu
At Inductive, our long-term goal is to build superhuman chemical intelligence, and by completing the reward function for autonomous drug discovery, Beacon-2 is an exciting step toward that goal.
Reach out if you’d like to get involved.
@benbirnbaum@inductive_bio In the big picture what is this gonna do for patients? Like once this tech is widely used how will the world be different? (better i assume haha?)
We made a bunch of improvements to this during G6 and are doing another webinar to share what's new.
Tuesday 9am PDT
https://t.co/b7L1kG6dXy
Any questions, reply or DM me!!
Thx!
📣 Quick announcement: Applications are open for the next Character Foundations, our one-day founder masterclass in San Francisco! More info and application link at https://t.co/VWtwQpw7bm
This is a VERY special event, and one of our favorite days of the year — it's the only time we really pull back the curtain for founders OUTSIDE our portfolio.
• What is it? Hands-on workshop where we share our Foundation Sprint, Design Sprint, and other company building methods with founders.
• Who is it for? Early-stage founders! Our methods are based on our work with 300+ teams building products, and optimized for the uncertain early days.
• When is it? All day September 8.
• Where is it? San Francisco.
• How much does it cost / What's the catch? It's completely free and equity-free for founders. You just have to apply.
• Why should you apply? a) You'll learn something actionable, b) you'll meet some great founders, and c) you'll have fun! I promise :)
Please reach out via comment/DM if you have questions. Thanks for reading!
Over 1 billion PDFs are created every day, but your agents still can’t read them reliably.
Today we’re releasing Parse 2.0, the most accurate document parsing API in the world.
Extend already processes millions of pages daily for leading AI teams like Brex, Mercury, Opendoor, Flatiron Health, and hundreds of others. Now, its even better.
Parse 2.0 is SOTA quality on RealDoc-Bench, our open source benchmark that measures agent success rate on real world docs that agents actually encounter in production.
We trained Parse 2.0 on 1M+ pages of the hardest documents seen in production. Here’s how it stacks up:
- #1 in healthcare, real estate, logistics, and financial services
- 95.7% agent Q&A accuracy on 581 docs (next best: 92%)
- 0.847 F1 on layout (next best: 0.759)
Give it a try today and build production-ready document agents with Extend.
Today we’re launching Design Canvas for AI agents.
AI can build a feature, but every "design tweak" turns into another prompt.
We built a drag & drop editor for Claude Code, Cursor, Codex to explore and iterate on your UI, so you get the best design by default.
No endless prompting. Ship better designs in code with Subframe.
🚨 Back by popular demand: Character Foundations, our one-day masterclass for founders! Applications are now open for our next event on March 26 in SF.
Come spend the day with @jakek, @eliblee, and me! Learn our unique approach to product design, meet other founders, and participate in real conversations about building startups today.
Last year, we launched Character Foundations on a simple premise: in the age of AI, product, design, and marketing matter more than ever.
Today, that's 10x true.
For many teams working in AI, coding and shipping are no longer a significant bottleneck. They can build and launch faster than they ever imagined.
But so can every other team.
As a result, what matters most today is:
1. Product — Am I building something that meaningfully solves an important problem for customers?
2. Marketing — Can I cut through the noise and stand out in a market where everybody is building and shipping fast?
But there's something else important going on in AI...
The capability of software has expanded dramatically. Today you can do things with software that you simply could not five years ago: discover drugs, design chemicals, invent materials, analyze research, automate filings, close deals, etc etc.
For teams building here — what we call "frontiers of the software-addressable economy" — there is an even greater set of challenges, because there is simply no playbook for how to build a software company like this.
- How do you sell software products to a company that doesn't buy software?
- How do you build trust in your product, when it's doing a job that's done by experienced human experts?
- How do you go full-stack and sell outcomes, not just software?
To be clear: We don't have easy answers for these questions. Nobody does.
We don't believe in playbooks and cheatsheets that tell founders WHAT to do.
Instead — Character Foundations is all about HOW to answer these questions.
We've used our Design Sprint and Foundation Sprint to help 300+ teams navigate the intense unknowns of building new products and businesses today.
We're on the court with founders every day, using these methods to give our portfolio companies at @CharacterVC the greatest possible chance of success.
We want to share the Design Sprint and Foundation Sprint with more founders. So, for the first time ever, we created an opportunity for founders to sprint with us without raising money from Character.
Character Foundations is free and equity-free: There are no fees to attend, and we take 0% of your company. You just need to apply.
We love doing this work with founders. And Character Foundations has become one of the truly peak moments of our work together at Character.
You can read more and apply at https://t.co/VWtwQpw7bm — and if you have questions, please reply and I'll drop in to answer it.
Thanks!!
Our team’s work outperformed:
OpenAI’s GPT-OSS model by +49.32 pts
Gemini 2.5 Flash-lite model by +6.12 pts
Claude 4.5 Sonnet by +6.49 pts
We’ve all seen the news headlines of large consulting reports that hallucinated. They lead to lawsuits, fired employees, and bad PR.
Our policy engine dramatically improves accuracy and our results speak for themselves:
- 3.3× accuracy gains on TruthfulQA
- +49 pts truthfulness
- 96.5% hallucination prevention on HaluEval
A question we get often is
“How can you control models you don’t own?”
Answer:
You don’t fine-tune the model.
You govern the reasoning path.
the future is extremely promising, but it has some advanced dangers.
so let's def/acc -- accelerate the defenses to big risks, so we get the good future.
hackathon nov 22 at @vivariumsf. sponsored by @BlueDotImpact (others to be announced soon).
https://t.co/VPd0vyylhl