This week: an open AI model for materials that has more downloads than DeepSeek but still can’t touch a bioink, a Singapore clearance for a 3D-printed wound-care implant, three fresh bioprinting papers, and cartilage printing aboard the space station.
The big thing
AI Model for materials beat DeepSeek’s downloads. It still can’t help fill the biology gap.
At our Biomaterials Frontier event, a Johns Hopkins and NIST materials scientist Prof. Kamal Choudhary mentioned, almost in passing, that his open AI model (DiffractGPT) for materials had been downloaded more times on Hugging Face than DeepSeek. But that is hard not to notice.
Kamal Choudhary has built the tools that make AI-driven materials discovery real: an open database with 200,000 users, models that predict a material’s properties in seconds, and a “ChatGPT for materials” that plugs into Claude. Hand one an X-ray pattern and it will propose the atomic structure behind it, including for compounds that do not exist yet.
And almost none of it can help develop biomaterials. At least not yet.
That is the uncomfortable part, and Choudhary said it himself when the moderator Craig Rosenblum asked. The AI is fluent in the physics of hard, crystalline materials. It knows little about biology: how it degrades, whether cells like it, whether it heals. Good data barely exists, and the experiments that would create it are the slow, expensive bottleneck.
Which is exactly why it is an opportunity.
The moat in (bio)materials science is not just the model. It will be the biological data few have, and the company patient and strategic enough to generate it.
Read our more in-depth review and analysis based on Dr. Choudhary’s work at Hopkins here, and see where the money goes next, in the full piece: More downloads than DeepSeek: AI Model for Materials (Expert Corner) →
Regulatory watch
- UltiMaker and Bioactivx: a 3D-printed wound-care implant clears Singapore’s HSA — Bioactivx’s synthetic, animal-free Bioactiv Matrix, printed on UltiMaker systems in an ISO 13485 cleanroom, won Singapore HSA approval, with US, EU, Australia and ASEAN filings pending. A proof point that additive manufacturing can meet certified end-use device standards, not just prototyping. (This is a Singapore HSA clearance, not an FDA action; no new FDA 510(k) AM clearances surfaced this week.)
Clinical & research
- In-situ bioprinting builds a blood supply inside new bone (Acta Biomaterialia · via PubMed) — Targets bioprinting’s core bottleneck, keeping thick tissue alive by printing its vasculature in place. Preclinical (mouse).
- A bioprinted dentin-pulp model that mineralizes and grows vessel-like channels (Biofabrication · via PubMed) — Stiffer zones drive mineralization while softer zones form vessels, a tunable test bed for regenerative dentistry. In vitro.
- A 3D-printed scaffold plus engineered stem cells regrows injured nerves (Bioactive Materials · via PubMed) — Scaffold-guided peripheral-nerve repair restored function in a preclinical model. Proof of concept (rat).
Also this week
- An AI database pairs drugs with excipients for on-demand pharmacy printing (VoxelMatters) — University of Mississippi with Colorcon; in development, not yet deployed.
- Cartilage bioprinting tops the research schedule aboard the ISS (NASA) — Reported microgravity crew activity, not a published result.
- ‘Subvoxel’ multimaterial bioprinting co-extrudes several bio-inks at once (Technology.org) — Demonstrated in cultivated meat; the multimaterial technique may transfer to tissue engineering.
From 3DHEALS
- Biomaterials & 3D Printing in Healthcare: event recap — The write-up by our intern Peter Hsu from the Biomaterials Frontier panel behind this week’s lead.
- Episode 127: Lean compliance for medical 3D printing, with Erik Boelen — The latest Lattice Podcast, on making quality systems workable at the point of care.
- On-demand course: AI Updates for 3D Printing and Bioprinting — Goes deeper on AI across the print-to-clinic workflow.
Comments