Mars astronauts could live in houses made of yeast and jello, say scientists

theregister.com Sep 12, 2026

Mars astronauts could live in houses made of yeast and jello, say scientists

Curated by Shubham · TechCapsules · Sep 12, 2026

AI-summarised brief · reviewed before publication

Researchers from The Hong Kong University of Science and Technology and The Hong Kong Polytechnic University have unveiled a low‑energy 3D‑printing technique that could turn Martian regolith into habitable structures using bioengineered yeast and gelatin. The method mixes yeast engineered to secrete adhesive proteins with a gelatin hydrogel and simulated Martian soil, then extrudes the blend through a printer nozzle. In the planet’s cold, thin atmosphere the material freeze‑dries, leaving a lightweight, porous composite that hardens without the high‑temperature sintering traditionally required for in‑situ resource utilization. The team’s paper, published Thursday, argues that the approach could produce shelters for future astronauts while minimizing power consumption and equipment mass, addressing a key logistical hurdle for sustained human presence on Mars.

💡 Why It Matters

· A freeze‑drying bio‑composite sidesteps the massive energy demands of conventional regolith processing, making on‑site construction feasible for early Mars missions.

Read full article on theregister.com

Hear how AI can engineer nature’s comeback at TechCrunch Disrupt 2026

At TechCrunch Disrupt 2026, Ben Lamm, CEO of Colossal Biosciences, led a Real World AI Stage fireside chat on using artificial intelligence and synthetic biology to resurrect extinct species. The discussion examined the technical methods behind de‑extinction, the growing role of AI in modern biology, and the ethical debate over whether engineering nature serves as a conservation breakthrough or distracts from protecting existing ecosystems. The event highlighted how emerging tech is reshaping biodiversity science and policy.

💡 The conversation spotlights a pivotal shift where AI directly informs ecological restoration, challenging traditional conservation priorities and prompting urgent policy dialogue.

microsoft.com · Sep 14

How pharmaceutical leaders are operationalizing AI

Pharmaceutical leaders are moving beyond experimental AI pilots to embed artificial intelligence into core scientific, manufacturing, and operational workflows. This shift aims to accelerate drug development, which traditionally requires over a decade and billions of dollars, while addressing pressures for supply chain resilience and personalized care. Microsoft’s “AI for Better Health” initiative supports this transition by combining trusted data with human expertise. Key industry players are already operationalizing these tools to enhance decision-making speed and quality. Novo Nordisk reduced time-to-insight from weeks to minutes using a governed AI agent on Azure. Amgen deployed a Catalyst Copilot in six weeks to make institutional knowledge searchable. Almirall developed an assistant accessing over 400,000 documents, cutting retrieval times from days to seconds. These examples demonstrate that the focus has shifted from proving AI’s viability to scaling its enterprise-wide impact, enabling researchers to evaluate more hypotheses and improve patient outcomes through faster, data-driven innovation.

💡 The competitive divide in pharma is no longer defined by access to AI technology, but by the ability to integrate it into repeatable decision-making systems. Companies that successfully merge enterprise data with human expertise will drastically shorten discovery cycles, turning a decade-long process into a more agile, efficient operation.

psypost.org · Sep 14

How virtual reality is changing brain-training technology

A review in Applied Psychophysiology and Biofeedback evaluated 31 studies that combined EEG‑based neurofeedback with virtual‑reality (VR) environments. Researchers from the University of Graz found that VR feedback—such as making a flower bloom or a spaceship accelerate by altering brain waves—was generally preferred over traditional two‑dimensional displays, with participants reporting greater enjoyment, interest and perceived competence. The studies spanned healthy volunteers and clinical groups treating chronic pain, migraines, and post‑stroke symptoms, and employed varied tasks like navigating a ball through a forest or adjusting virtual lighting. Sample sizes ranged from single‑case reports to 100 participants, many under 50. While the review suggests VR can enhance engagement in brain‑training therapies, it also notes the need for more rigorous, larger‑scale trials to confirm specific benefits.

💡 Enhanced user engagement could accelerate the adoption of neurofeedback as a viable therapeutic tool for diverse neurological and pain conditions.

cyprus-mail.com · Sep 13

How AI could replace animal testing

Scientists across Europe are integrating artificial intelligence, human cell cultures and organ‑on‑a‑chip technologies to predict the toxicity of chemicals, aiming to replace the roughly 150,000 rats and mice used annually under the EU’s REACH regulation. The effort is coordinated by the €60 million ASPIS research cluster, which funds three initiatives, including Professor John Colbourne’s programme at the University of Birmingham and Professor Mathieu Vinken’s ONTOX project in Belgium. These projects target chronic toxicity testing, which is costly—up to €15 million per substance and several years—and often yields limited relevance to human health. By 2026 the teams hope to demonstrate AI‑driven, in‑vitro methods that meet safety requirements while adhering to the EU’s three‑Rs principle of replacement, reduction and refinement of animal use.

💡 Successful AI‑based testing could slash billions in compliance costs and accelerate market entry for new chemicals, reshaping the regulatory landscape for safety assessment.

northern-times.co.uk · Sep 13

Virtual reality lab training comes to Dornoch Academy

Pupils at Dornoch Academy participated in a week‑long workshop designed to teach National 5 Laboratory Science students the digital and practical competencies required for modern medicine manufacturing. Delivered by the Scotland RESILIENCE Hub—a specialist training centre at Heriot‑Watt University—and coordinated by teacher Dr Adrian Allan and Highland Council STEM Improvement Officer Katie Elder, the program combined hands‑on aseptic technique drills, precise measurement exercises, and assembly of bioreactor equipment with immersive extended‑reality (XR) simulations. The session aimed to support students’ Candidate Review coursework for the Careers Unit, which outlines real‑world science careers and employer expectations. Teachers praised the pupils’ teamwork, adaptability and technical precision as they moved between virtual labs and physical bench work, reflecting the skill set needed to address the UK’s shortage of qualified life‑science technicians.

💡 The training directly equips a pipeline of students with the exact competencies employers lack, accelerating the closure of the life‑science skills gap.

9to5mac.com · Sep 12

Apple researchers unveil SimpleDesign, a new AI model for protein design

Apple researchers have introduced SimpleDesign, a streamlined artificial intelligence model capable of jointly generating protein sequences and their corresponding three-dimensional structures. Building upon their previous SimpleFold study, which utilized flow-matching for structure prediction, this new model addresses the broader challenge of protein design. Unlike existing systems that rely on complex, multi-stage training processes involving autoencoders and latent space representations, SimpleDesign operates directly in the data space. It employs an end-to-end training approach using general-purpose Transformer blocks, eliminating the need for intermediate tokenization steps. The model was trained on over two million protein sequence-and-structure pairs from the AFESM dataset. During training, researchers corrupted both sequences and structures with varying degrees of noise, allowing the system to learn folding, inverse folding, and co-design tasks simultaneously. Benchmarks indicate that SimpleDesign delivers competitive performance in structure and sequence generation, producing plausible proteins with sequences that match or exceed those of most competing multimodal models, despite its significantly simpler architecture.

💡 By eliminating complex multi-stage training pipelines, Apple demonstrates that high-performance protein design does not require computationally expensive intermediate representations. This architectural simplification lowers the barrier to entry for biological research, allowing scientists to generate viable protein candidates with greater efficiency and less infrastructure overhead than previously possible.