seen_items.md was empty, so this digest is a baseline rather than a true delta. Items below are scoped to the past ~6 months; a few high-profile pieces from mid/late 2025 (Leigh's catenane motor, Science Jul 2025; reconfigurable DNA-origami nanorobot arrays, Science Robotics Nov 2025) were checked and excluded as falling just outside the window. Future runs will report only newer items.Cowie et al. (≈47 authors, incl. Robert A. Freitas Jr. and Ralph C. Merkle), "Atomically precise mechanosynthesis of carbon structures on hydrogenated Si(100) by inverted-mode STM", arXiv:2605.27250 (May 2026). EAOGe-C₂I molecules act as both imaging probes and reagents, donating C₂ units to pre-patterned reactive sites; the team shows single-site donation, spatially patterned multi-site donation, and stepwise polyyne growth via successive C–C bond formation. Why it matters: moves additive mechanosynthesis from simulation/roadmap to a benchtop demonstration of controlled, site-specific covalent assembly — long the field's central gap. Preprint, not yet peer-reviewed; reproducibility and yield/throughput remain open.
NYU DNA Lab (Vecchioni and colleagues, building on Seeman's legacy), Nature Communications, Mar 2026. DNA tiles assemble into a library of 3D structures — including mixed right/left-handed "mirror" DNA — guided by the flat helix-end interfaces and subunit geometry rather than sticky-end hydrogen bonding. Why it matters: decouples 3D DNA assembly from sequence-encoded base pairing, opening a shape-programmable design axis and potentially more robust/condition-tolerant assembly.
Antibodies Demonstrated — Bennett, Watson, Ragotte et al. (IPD), Nature 649:183–193 (2026). Fine-tuned RFdiffusion + yeast display yields VHHs, scFvs and full antibodies binding specified epitopes; binding poses confirmed by cryo-EM (influenza HA, C. difficile TcdB). Why it matters: atomic-accuracy design of the historically hard CDR loops; a concrete AI-driven path to bespoke binders.
Protein cages Reported — Baker lab, "programming proteins with viral geometry" (May 22, 2026). Designed subunits self-assemble into large cages by combining pentagonal and hexagonal patterns, the strategy natural viruses use to close a shell. Why it matters: a route to large, defined protein compartments (delivery, nanoreactors) from first principles. Confirm peer-review status next run.
Kluifhooft et al., "A Molecular Machine Directs the Synthesis of a Rotaxane", Angew. Chem. Int. Ed. (2026). A unidirectional motor embedded in a dumbbell axle winds an attached strand into a double-helix-like shape, forming the interlock without preorganization or templating. Why it matters: generalizes machine-directed construction of mechanically interlocked molecules (companion to Leigh's 2025 catenane work, which is outside this window). Single-source/new method — generality across larger or more complex topologies is unproven.
| Company | What they do | Stage / funding | This period's update | Tier |
|---|---|---|---|---|
| Anthrogen | AI foundation models over protein sequence/structure to generate novel molecular machines with atomic-level precision (function-to-sequence) | Early-stage startup | Surfaced as an active atomic-precision protein-design entrant; specifics of funding/round unverified | Reported |
| Archon | Protein-design therapeutics (David Baker–lineage); antibody cages | $20M seed + $7M grants | Runway into 2026; planning lead-program selection and a follow-on round toward a clinical trial | Reported |
Nothing material verified within the window. (Context, not new: APM continues to be positioned for quantum-device fabrication, claimed as more precise than leading-edge e-beam lithography — a roadmap/Projected framing, not a result.)