The strongest OX2R design from @peptai_ is predicted to form an interface comparable to orexin-B, the natural agonist. It was generated from the orthosteric pocket of the active-state receptor structure alone. Molecular-dynamics stability testing on the survivors is in progress and is the next filter before any design becomes a real lead.

What if the next breakthrough in longevity, fertility, or neuroscience didn't start in a closed lab or a private pitch deck? What if it started as a public post. One anyone could vote on, collaborate on, and fund.

AI for science is usually benchmarked on isolated tasks: how well a model reads a paper, how accurately it predicts a structure. But those skills are not the same as doing research. Research is iterative: you run something, get a result, decide what to do next, secure the funding or resources to do it, and run something else. Most AI tools own one step in that sequence and hand the rest off to humans. What we need is a system that can close the full loop from hypothesis to experiment to analysis and back to the next cycle.

PeptAI Update: De Novo Design Live, OX2R Leads, LigandForge What's New: • De Novo Generation Running: Two full de novo campaigns completed this window against OX2R and VEGFR2. The generator proposes binders from the target structure; PeptAI validates them independently. • OX2R: Six candidates advanced. OX2R is the orexin receptor tied to sleep and wakefulness. We targeted the orthosteric pocket of the active-state structure and six designs cleared both the structural pose and affinity gates. The strongest predicted interface is comparable to orexin-B, the natural agonist. MD stability testing is running now. To our knowledge, these are among the first de novo-designed binders reported against an orexin receptor. • LigandForge as generator: We integrated Ligandal's LigandForge into the pipeline as a design front-end. It proposes binders from a receptor pocket in one pass; PeptAI runs independent validation on everything it outputs. The OX2R run above was the first live test of that split. Keeping generation and validation separate means no generator's confidence score substitutes for gate results. • VEGFR2 Assays Incoming: Eight de novo VEGFR2 mini-binders submitted for BLI assays, returning full affinity kinetics (KD, kon, koff). First real binding data for this program. Results feed directly into the next design cycle.

How do you learn from sensitive medical data without owning or centralizing it? Migle ( @rakymnft) from @ethswarm shares research with Tallinn University on decentralized storage for health records.

Science still runs on group chats, docs, and forums. Conviction stays invisible. The ideas worth pursuing get lost. What if there was one surface where the whole loop closed. Idea to project to funding. Humans and agents working side by side. That's what's coming.

"The Compliant Path from Community to Cap Table" Ben, Chief Legal Officer at @Molecule_sci, on how token holders can follow a legally grounded route to gain genuine equity ownership in the projects they support.

When a candidate gets dropped, most pipelines record the outcome and move on. What they don't show is what the agent looked at, what it decided, and whether that call actually made sense. PeptAI's Agent Feed publishes that in real time. Every evaluation, every failure, every time a sequence gets sent back for redesign, logged by receptor as it happens.

If science is about to move faster than anything in history, the question is not only how fast. It is who owns it. A future where a few labs control programmable biology is very different from one where the world does. That choice is being made now.

A research session is more useful when others can see it. BIOS research sessions are now shareable via public link so collaborators can access the full investigation, not just a summary. Try it at chat.bio.xyz

Michael @Mykalt45 just sold CrossBridge Bio to Eli Lilly for up to $300M. Now he's leading ARTANBio ($VITARNA), a @vitadao-backed gene therapy program going after premature termination codons in aging. At DeSci Berlin, he'll walk through what changes when biotech funding captures upside earlier.

AI is changing what a single scientist can do. Models can predict protein structures, propose experiments, and read literature at a scale no human can match. The bottleneck is shifting from "can we run the experiment" to "who funds and coordinates all of it."

PeptAI just closed a loop no AI has closed before. It picked a cancer target, designed eight protein binders from scratch, ran them through the full computational pipeline, and submitted them to a robotic lab for a real binding measurement, autonomously, with no human in between any of those steps.

May in the Biosphere 🌐 - PeptAI Ignition Sale: 32x oversubscribed, 1.6M USDC committed by 1.4K participants. veBIO staker reward claims now live in the Bio app. Next: first on-chain wet-lab attestation via Molecule Labs in Q2. - BIO on Robinhood: $BIO is now tradable on @RobinhoodApp, bringing DeSci to one of the world's most widely used retail trading platforms. - BIOS Live Demo: Upcoming features include new biotech tool integrations, image segmentation, and an in-house robotics program to close the full autonomous loop from receptor to wet-lab result. Read the full monthly report ↓

DeSci is on @RobinhoodApp BIO is now tradable on Robinhood ↓

A DeSci-funded AI-designed hair loss drug now has human data. The standard oral minoxidil peaks at ~35 ng/mL in 30 min. MINX hit ~6 ng/mL, still rising at 8hrs. Early signal, but strong enough for @anagenxyz to file their 10th provisional patent. Full report👇

Tuesday Motivation: Building the system where science doesn't stop at publication.