Vamsi Varanasi

Today, we’re launching @TuvaAI to build a human-centric future for agentic science. Humans collaborating with AI can learn more than either working alone: humans guiding the research, equipped with agents that can test ideas quickly and autonomously, will deliver the next wave of scientific discovery. But science is all about the details. Neither Einstein nor AGI can drive your project forward without intimately understanding your growing, interrelated web of hypotheses, intermediate results, working threads, failed experiments, datasets, notes, and water cooler conversations. Our first product, Rao, was born out of my frustration having to reteach LLMs these constantly changing details for every single task in my own research. Rao maintains a persistent, dynamic research memory that learns and remembers project context, history, and threads. This research memory then links into our scientific agents as well as other AI software you may already be using, such as Claude or Codex. With Rao, every scientist turns into a PI managing a proactive, agentic research team that automatically stays on the same page. Rao is live in VS Code and by CLI, and can help with any scientific task that can be done in a computer. Our early users include computational biologists, theoretical neuroscientists, chip designers, and polymer physicists who use Rao daily to ideate, do math, analyze data, and write papers and grants. Now, we’re excited to announce Rao in closed beta. If your science mostly happens in a computer—theory, computational modeling, experimental data analysis, scientific writing—Rao can help you discover more, faster. Sign up for our waitlist, and if you’re a fit, we’ll work with you to tailor new, contextual agents that augment the way you do science. Link in the comments. More personally: I’ve been doing scientific research since I was a kid. Scientists do so much for the world, choosing a painstaking career for the joy of discovery, developing technologies we use every day along the way. And yet, the tools scientists use are often cumbersome and woefully out of date. For me, Tuva is much about empowering the people of science with products they love as it is about the discoveries we’ll help them make. And I’m lucky to work with some wonderful humans along the way—@SamsaraDurvasu1 @anuvellore @KimchiOfer among many others. If our mission resonates, please reach out.

@billhehehe @TuvaAI we'll get you set up with tuva!

@feynmanpoint @TuvaAI Remembering the details of your project is super important, I wish this type of context /workflow/SOP management is built in

Today, we’re launching @TuvaAI to build a human-centric future for agentic science. Humans collaborating with AI can learn more than either working alone: humans guiding the research, equipped with agents that can test ideas quickly and autonomously, will deliver the next wave of scientific discovery. But science is all about the details. Neither Einstein nor AGI can drive your project forward without intimately understanding your growing, interrelated web of hypotheses, intermediate results, working threads, failed experiments, datasets, notes, and water cooler conversations. Our first product, Rao, was born out of my frustration having to reteach LLMs these constantly changing details for every single task in my own research. Rao maintains a persistent, dynamic research memory that learns and remembers project context, history, and threads. This research memory then links into our scientific agents as well as other AI software you may already be using, such as Claude or Codex. With Rao, every scientist turns into a PI managing a proactive, agentic research team that automatically stays on the same page. Rao is live in VS Code and by CLI, and can help with any scientific task that can be done in a computer. Our early users include computational biologists, theoretical neuroscientists, chip designers, and polymer physicists who use Rao daily to ideate, do math, analyze data, and write papers and grants. Now, we’re excited to announce Rao in closed beta. If your science mostly happens in a computer—theory, computational modeling, experimental data analysis, scientific writing—Rao can help you discover more, faster. Sign up for our waitlist, and if you’re a fit, we’ll work with you to tailor new, contextual agents that augment the way you do science. Link in the comments. More personally: I’ve been doing scientific research since I was a kid. Scientists do so much for the world, choosing a painstaking career for the joy of discovery, developing technologies we use every day along the way. And yet, the tools scientists use are often cumbersome and woefully out of date. For me, Tuva is much about empowering the people of science with products they love as it is about the discoveries we’ll help them make. And I’m lucky to work with some wonderful humans along the way—@SamsaraDurvasu1 @anuvellore @KimchiOfer among many others. If our mission resonates, please reach out.

Before tech - I was a scientist myself (got a PhD in semiconductor physics) I can tell you first hand that research and research processes need tools like this to enable better experiment design and speed up learnings and iteration to enable new discoveries I am personally very excited about this work, and believe it can be extraordinarily high impact Let's go @feynmanpoint and team @TuvaAI !

@asmah2107 :)

Context management is actually a complex distributed systems problem.

@TuvaAI Learn more about how Rao works on our website: r.tuva.ai/Immz5un

@TuvaAI Join us to build the future of scientific research: r.tuva.ai/vTdHQiX

@TuvaAI We’re letting people off of our waitlist quickly: r.tuva.ai/qRLSXAz

@arshdilbagi Awesome!

Introducing Adaline 2.0 - The Agent Self-Improvement Layer Adaline turns Traces into Behaviors, Behaviors surface Issues, Issues become auto-generated Evals + Data, Adaline then generates new agent candidates and tests them. You review the winners and ship!

I have spent my entire life working on this and thinking about this for the past 4 years. I don't know what will happen in 20 years, but I can promise you that on the 5-10 year timescale, scientists are not out of their jobs. AI is going to massively accelerate the pace of science, increase productivity, let individual scientists make way more discoveries way faster, and is going to make science overall more fun. But the model is going to be collaboration between humans and AI, not replacement. The key difference here between science and e.g. software engineering is that science is not verifiable in any rapid/convenient way (unlike software), unlike programming. We still need humans for their scientific taste.

rough shape of progress based on current data '22 - chatgpt - consumer s/w '25 - claude/coding - enterprise s/w '26 - '29 - [redacted] - advanced mfg, iphone killer, materials 2.0 '29 - ?? enabled by cambrian explosion of r&d to clear progress bottlenecks

While I generally agree with the point that agents being good at code means the first application in bio will be agents writing code (to paraphrase), this misses the key point about how to enable this effectively: context.

@zan2434 @eddiejiao_obj @drewocarr This is sick man

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see. @eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

@sam_kececi I've been waiting for this!!

We raised $6.5m to build humanity’s platform for uploaded consciousness. Sentience creates one unique model for every person — a digital twin of your mind — to remember everything, recall what matters, and operate as you. I started this company because I am worried. For the first time, we have intelligence that will fully replace humans across a range of domains within a few years. For those outside of the tech echo chamber, this is a period wrought with confusion and uncertainty about what actually matters anymore. I’ve had numerous conversations with people who aren’t sure if they have any value as a unique human being, or whether their own thinking matters. While Silicon Valley goes full speed ahead towards AGI, many people are left wondering what place they have in this AI future. If I can communicate one thing to everyone out there with these doubts, it would be this: your unique knowledge, memories, and who you are still matter. In fact, these things matter now more than ever. The problem is that the course of AI is heading towards a world where all of us will outsource our thinking to the same one-size-fits-all AI models. This is not a hypothetical future. If 100,000 people ask a question to ChatGPT, every single person gets the same answer. You can see the cost of uniform AI models across writing, social media, and even academic papers. This uniformity is more than annoying – it’s dangerous. There’s a genuine chance that we lose the texture and vibrancy that makes us unique as a species. I founded The Sentience Company to arm real humans against this dystopian outcome. First, your Sentience lets you collect everything that holds context from your life, learning from what you do across every platform – starting on desktop and mobile. Never forget a detail again. Second, your Sentience becomes the best recall engine for everything in your life. Never copy/paste context or search across 50 chrome tabs again. Finally, your Sentience becomes the full simulation of you – an AI model that thinks and acts like you, to scale and share your unique ideas and interact with others. Your Sentience emulates more than your context. It understands your values, emotions, drive, and goals. We’re creating a world where you can leverage your own Sentience model alongside the models of your colleagues and friends to jam on ideas and access their knowledge 24/7. We’re not building Sentience to scale AGI and replace more human thinking. We’re building Sentience to scale you. We’re proud that many amazing humans are supporting our mission. Our round was led by @kevinzhang (@BainCapVC), with participation from @ditzikow, @adityaag, @evantana, @AgrawalArian, @gopalkraman, @JPBrebner (@southpkcommons), @rex_woodbury, @tmrohan, @soleio, @anniecase1, and many more.

@arjunrajlab @OdedRechavi The mindset I've found helpful: in the age of AI, every junior scientist becomes a PI of their research team of bots, and every PI a department chair

Transitioning to being a PI in the age of AI Computational biology is in a period of upheaval that is both exhilarating and terrifying. Rapidly, we are approaching a moment of “analytic abundance”, where basically idea you can think of (and several you didn’t) magically appear within minutes of you thinking of them. Of course, the central proximal challenge is the evaluation of the sheer volume results—how do we know they are right when we don’t have the time to check over every line of code? I think it’s very telling that when I talk to AI-pilled faculty, they are exhilerated, but many trainees seem more cautious and far more ambivalent. I think that’s because faculty often have been removed from the details for a long time and probably haven’t checked over a line of code in years. They are used to managing (rather than doing) analysis. Over time, they usually develop a sense for whether things seem right or wrong. In this day and age, this is the skill that you, too, must develop. How do faculty do it? I am guessing every faculty member has their own list of internal sanity checks, but here are a few of mine: * Checksums. I look for things that should add up correctly (percentages add to 100, etc.). If it looks even a little bit off, I ask questions. * Never let go. If something doesn’t make sense, I don’t let go until it does make sense. Never relent! * Explain stray datapoints. Always dig into outliers in the data. How did they come to be? Often, they reveal some hidden assumption or something unexpected about the data. * Do not tolerate warnings. If code gives you a warning, resolve it. Do not continue, do not pass go, until you either understand or eliminate the warning. * Track the number of datapoints. Even a single missing row can be a sign of some fencepost bug. And I’m sure many more that I’m forgetting right now. Basically, it’s a transition from a maker to an interrogator. I also feel it worth reiterating that this is a highly unsetting period of time. I have been fortunate (?) to have 16 years of time to make a transition that people are now being asked to make in months. Again, exhilarating and terrifying, all at once!

when did we agree to live in a world where we fill in otp codes 10 times a day, every single day???

@arscontexta What is science if not adding nodes and shuffling edges on humanity's shared context graph?? Imagine bots remembering incremental results as they help scientists, and scientists sharing/merging learnings by comparing diffs between graphs (we are building this; dms open)

@feynmanpoint how does it change science

running a small tech startup on top of a company graph feels like an unfair edge right now

tis a time of great optimism (my orchids are blooming again)