OpenAI’s science moves. Solving a century-old mystery about rubber. The space economy keeps booming. All that and more in this week’s edition of The Prototype. All that and more in this week’s edition of The Prototype. To get it in your inbox, sign up here.
OpenAI CEO Sam Altman
Getty Images
OpenAI unveiled a new model this week, GPT-Rosalind, which is custom-built for scientists working on drug discovery, biology and other medical research. Named for Rosalind Franklin, who helped uncover the structure of DNA, the system is aimed at helping to speed up the R&D process. The news follows an earlier announcement of a strategic partnership between the company and Ozempic maker Novo Nordisk, which plans to integrate more AI across its business units.
The new model works as a one-stop interface shop, connecting more than 50 tools used by scientists day-to-day, such as journal articles, molecule databases and predictive tools. It’s also integrated into Codex, the company’s AI coding assistant, to help enable custom workflows.
Life sciences have become a major area for competition in the AI industry, and OpenAI is far from the only player. Nvidia has a number of different platforms for researchers, as does Anthropic with its Claude for Life Sciences. And earlier this week, Amazon launched its own product, Amazon Bio Discovery, specifically aimed at integrating lab work with AI models.
And although GPT-Rosalind is new, OpenAI has been working with scientists for years. Derya Unutmaz, an immunologist who works at biomedical research organization Jackson Laboratory, has been using ChatGPT since version 3.5 came out, he told me. He’s undeniably enthusiastic about its potential, and said that more recent versions have helped him and other scientists in his lab to better understand different immunological mechanisms, which guided their experiments. He’s also using it to help him write a textbook on how T-cells works, and has praised the model’s accuracy.
Jason Kelly, CEO of Ginkgo Bioworks, is also optimistic about using AI for scientific research. He said that his company has integrated OpenAI’s models into its physical laboratory setups, so that the AI can automatically design, execute and iterate scientific experiments. It’s also integrated agents into its Cloud Lab, which allows scientist-customers to use it as an interface to help design and direct experiments. For Kelly, the most exciting potential of AI is a future where every scientist is essentially running their own lab, directing AI agents to do experiments for them. “It’s going to be that you come in in the morning, have your coffee, and look at the data from your experiments the night before,” he said, rather than manually putting things into test tubes or cell cultures.
Prior to this announcement, I had a conversation with Kevin Weil, who heads OpenAI’s initiatives for the scientific community. For him, one of the key issues for the company is to help automate some of the more laborious and tedious scientific tasks to “accelerate discovery”–and notes this often happens in small, compounding ways, not necessarily all in one swoop. “If we can help the world do the next 25 years of science in five instead, we could be sitting here in 2030 with the tools that we would have otherwise had in 2050,” he said. “That’s an awesome place to be.”
Icon Sportswire via Getty Images
More than a century ago, British chemist S.C. Mote discovered that adding carbon black to rubber made it significantly stronger, enabling the material to support large trucks and jumbo jets. But exactly why it works hasn’t been fully understood–though there have been plenty of ideas. A new paper published this week could change that.
The study presents the results of hundreds of molecular simulations from a team at the University of South Florida, which modeled how the molecules interacted with each other. They uncovered that the nanoparticles of carbon black act as thousands of tiny ‘supports’ for the molecular structure of rubber, which naturally thins out as it stretches, causing the rubber’s volume to increase.
This is because the nanoparticles of carbon black are highly attracted to molecules of rubber. It “freezes” the rubber molecules in place, forming glassy shells that strengthen the structure of the material by supporting the overall expansion of its volume. This finding actually unifies several different hypotheses about rubber reinforcement, none of which could fully explain the phenomenon on their own.
The upshot of this discovery is that it might make it easier to design new ways to shape and reinforce rubber for different applications from the start, rather than figuring them out through trial and error. For example, researcher David Simmons said the findings “lay a new foundation for rationally designing tires” that are stronger, have better traction but also help preserve fuel economy.
Amazon is buying satellite company Globalstar for $11.57 billion, in an apparent bid to catch up to SpaceX’s Starlink. Globalstar’s satellites are already capable of direct-to-device connectivity, which is expected to be a growing business over the next decade.
It’s also a sign of increasing investor interest in the space economy. This week, venture firm Space Capital released a report finding that $36 billion flowed into the sector in the first quarter of 2026, providing backing for 148 startup companies–a new record for the industry. “The convergence of AI, geopolitics, and orbital compute is driving capital into the space economy at a pace and scale that would have seemed implausible five years ago,” the report said.
It’s hard to disagree with that observation. And one thing to watch out for is an increasing focus on regional players in response to the Trump administration’s trade policies. Contracts and funding that might have gone to American companies are already seeing a shift toward local players, which could heat up competition in the sector overall.
Each week, I ask investors for their take on tech trends within their industries. Today I’m featuring thoughts from Brian Hollins, founder and managing partner of Collide Capital, which just raised its second fund to focus on investments in early stage fintech and future-of-work startups.
Brian Hollins
Collide Capital
AI copilots [aka AI assistants]. Most of what's being sold is a better UI on top of existing workflows, not real automation. The numbers tell the story: Only 6% of enterprises have successfully moved generative AI projects beyond the pilot stage and when enterprise users have access to copilot alongside competing tools, only 8% choose it voluntarily. The infrastructure, security, and reliability layers needed to make generative systems work at enterprise scale are years behind the marketing. Buyers are paying for the promise, not the product.
Industrial robotics for small and mid-sized manufacturers. Retrofittable autonomy and smarter human-machine interfaces are finally making automation accessible outside of Fortune 500 supply chains. 95% of manufacturers are currently using or evaluating smart manufacturing solutions, yet robot density at SMEs remains a fraction of what large manufacturers deploy. That’s a massive unlock–for U.S. productivity, for reshoring, and for the millions of manufacturers who’ve been left out of every previous automation wave.
Workforce simulation. Synthetic data and digital twins that let you stress-test labor capacity, logistics, and operational decisions before they fail in the real world. As AI and IoT converge, the companies–and cities–that aren’t running these predictive models will look like they’re flying blind. It won’t be optional. It’ll be infrastructure.
Edison Awards: This week, the annual Edison Awards honored innovative industrial products, many of which definitely seem cool and worth checking out. Some of the ones that intrigued me include: artisJet’s 3-in-1 printing technology, which better enables ADA compliant printing; Samsung’s smart modular house; a way to recycle old asphalt into new road material; and Dot Pad X, which converts images into touchable dots to make them more accessible for the visually impaired.
Iran war and green energy: The oil shock from the conflict in Iran is boosting the attractiveness of renewable energy (you can’t blockade the Sun and wind, after all.) High gas prices are leading consumers to take another look at new electric cars, boosting battery exports from China and supercharged the IPO of Shanghai-based energy storage maker Sigenergy Technology, making company founder Xu Yingtong a billionaire.
Anthropic and the federal government, ctd: The White House is reportedly providing federal agencies with access to the company’s Mythos AI. The model, which Anthropic claims can easily uncover software vulnerabilities, has so far had its access limited to a handful of major software and financial companies. This move is happening even though the Department of Defense has labeled Anthropic a “supply chain risk” that prevents its software from being used by contractors working on DOD projects. What this means for that ongoing court battle is anyone’s guess right now.
Futuristic warfare: Ukraine President Volodomyr Zelenskyy said that Russian troops recently surrendered to a group of Ukrainian drones and ground systems without any human soldiers present. Meanwhile, the U.S. Space Force recently laid out its vision of warfighting in the space domain. Between conflicts over trade blockades, drone armies and orbital combat, the Star Wars prequels are turning out to be more prescient than I thought they were back in 1999.
For centuries, people have been coating their meat with starch before stir-frying it, believing it to make for tender, juicier results. As it turns out, the myth is justified. A new study in the Journal of Food Engineering finds that adding a thin layer of starch leads to faster browning and overall cooking. That’s thanks to the combination of the oil and starch causing microbubbles to form on the meat, which both retains moisture and leads to faster heat transfers.
I’ve been a fan of They Might Be Giants since 1991, when tracks from their album Flood were featured in an episode of Tiny Toon Adventures. This week, the band dropped a new album The World Is To Dig, and it is a treat. It feels very much like a throwback to the band’s early days (in a good way)–lots of experimental, fun tracks filled with both good wordplay and good hooks. My favorite songs so far: “Outside Brain,” “Let’s Fall In Lave” and “In the Dead Mall.”
ForbesAI’s New Training Data: Your Old Work Slacks And EmailsForbesThis Sam Altman-Backed $1.8 Billion Startup Bets AI Can Get Drugs Through Clinical Trials FasterBy Amy FeldmanForbesMeet The Cannabis Industry’s Trump WhispererBy Will Yakowicz
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。