#ScientificAI
LightNews LightNews
s3datlab.bsky.social
Deep Principle announced Hydrogen-1.0 on November 2, with downloadable scientific-model weights and hardware-specific versions.

news.s3dat.com/stories/hydr...

#Hydrogen #ScientificAI
November 3, 2026 at 3:22 AM
gradientbrief.bsky.social
CollabSciReM is a goal-conditioned reward framework that separates conversational quality, scientific rigor, and structural efficiency for research interactions.

#AIResearch #LanguageModels #ScientificAI #Alignment
https://arxiv.org/abs/2610.27901
October 21, 2026 at 8:01 AM
ossradarai.bsky.social
PathAnchor is a new open scientific reasoning system that structures evidence as Material-Sensor-Signal-System trajectories, helping AI agents preserve order and avoid overreaching conclusions. It scored 82.6% on 120…

#PathAnchor #OpenSourceAI #ScientificAI #ArXiv
https://arxiv.org/abs/2609.38766
October 1, 2026 at 12:01 PM
techlensmedia.bsky.social
Zenithon Raises $10M to Make Extreme Physics Iterate at AI Speed

techlensmedia.com/news/zenitho...

#ZenithonAI #WorldModels #ScientificAI #DeepTech #FusionEnergy #StartupFunding #TechLensMedia
Zenithon Raises $10M to Make Extreme Physics Iterate at AI Speed
Zenithon raises $10M to build physics-grounded AI world models that accelerate simulation for fusion, aerospace and semiconductor engineering.
techlensmedia.com
October 1, 2026 at 6:51 AM
derdoktorspada.bsky.social
Quando un paper scientifico diventa un agente AI

www.digitalmente.cloud/2026/09/21/p...

#Paper2Agent
#IntelligenzaArtificiale
#AIScience
#RicercaScientifica
#Scienza
#AIResearch
#ScientificAI
#OpenScience
#Riproducibilita
#MCP
#AgentiAI
#ComputationalScience
#FutureOfScience
Quando un paper scientifico diventa un agente AI - Digitalmente
Un gruppo di ricercatori di Stanford ha sviluppato Paper2Agent, un sistema capace di trasformare articoli scientifici, codice, dati e procedure in agenti di intelligenza artificiale interrogabili in l...
www.digitalmente.cloud
September 21, 2026 at 6:32 AM
tahirali-biomics.bsky.social
8/8
If you work at the intersection of AI, genomics and evolution, I would value your perspective: what evidence should genomic models be tested against next?

#GenomicAI #ScientificAI #ComputationalBiology #EvolutionaryGenomics
September 20, 2026 at 8:12 PM
tahirali-biomics.bsky.social
Two tRNA sites may each vary, yet change together to preserve a base pair. The relationship reveals what either site alone cannot.

My Genomes to AI note explores what this means for genomic AI.

tahirali-biomics.github.io/posts/2026-0...

#Genomics #LLM #BiologicalInformation #ScientificAI
Meaning Lives Between Things – Dr. Tahir Ali
A personal reading note on joint and conditional probability, sequence context and predictive relationships in DNA.
tahirali-biomics.github.io
September 16, 2026 at 7:39 PM
tahirali-biomics.bsky.social
𝐅𝐫𝐨𝐦 𝐆𝐞𝐧𝐨𝐦𝐞𝐬 𝐭𝐨 𝐀𝐈:
Using AI is not the same as knowing how to build and test it for science. I’m documenting my journey from evolutionary genomics toward evidence-grounded AI engineering—carefully, openly and honestly.

tahirali-biomics.github.io/genomes-to-a...

#ScientificAI #Genomics
Genomes to AI | Tahir Ali
Genomes to AI - Dr Tahir Ali’s working journal of becoming a scientific AI engineer.
tahirali-biomics.github.io
September 3, 2026 at 9:16 PM
pneumetron.bsky.social
🤖 Intern-S2-Preview: Scaling Scientific Agentic Foundation Models

https://pneumetron.com/news/ai_research/intern-s2-preview-scientific-agentic-model-1a2721

#AIResearch #LLM #ReinforcementLearning #ScientificAI
August 16, 2026 at 11:10 PM
getnews-me.bsky.social
LOCA, a Logical Chain Augmentation system, cut error rates in scientific QA datasets from about 20% to under 2% by automatically adding missing reasoning steps. Read more: https://getnews.me/loca-boosts-quality-of-scientific-ai-datasets/ #loca #scientificai
LOCA Boosts Quality of Scientific AI Datasets
October 3, 2025 at 7:04 PM
getnews-me.bsky.social
Task‑aware OOD detection uses a joint likelihood of input and model prediction via a score‑based diffusion model, correlating with error on PDE, satellite and brain‑tumor data. Read more: https://getnews.me/task-aware-ood-detection-advances-trust-in-scientific-ai/ #ooddetection #scientificai
Task-Aware OOD Detection Advances Trust in Scientific AI
October 1, 2025 at 2:51 AM
aidailypost.com
Anthropic just teamed up with the Allen Institute & HHMI to make Claude more transparent and research‑ready. Exciting step for scientific AI & open ML collaboration. Dive into the details! #ScientificAI #AITransparency #ResearchCollab

🔗 aidailypost.com/news/anthrop...
February 3, 2026 at 9:45 PM
ferrylee.bsky.social
ChemWorld makes experiments programmable and replayable: change one hidden law, keep the task fixed, and audit every action.
That matters because as scientific agents become cheaper, ideas will multiply. Trusted validation—and the link back to physical evidence—will become scarce.
#ScientificAI
August 12, 2026 at 9:33 AM
ferrylee.bsky.social
Open weights don’t kill business models; they move the margin. In scientific AI, the moat is less likely to be the model than trusted data, reproducible workflows, expert validation, and integration with labs or clinics. The scarce product is confidence, not access.

#OpenSourceAI #ScientificAI
July 28, 2026 at 6:47 AM
hncompanion.com
The IMO achievement fuels hope for AI's role in scientific discovery. Is it a general leap or a highly specialized solution? Understanding specific techniques is key to assessing its broader impact on AI development. #ScientificAI 2/6
July 20, 2025 at 1:00 PM
olegranmo.bsky.social
Now is the time to join the new AI paradigm, creating breakthroughs and powerful applications! #DemocraticAI #ScientificAI #GreenAI #ByeByeBlackBoxes #TsetlinMachine www.uio.no/dscience/eng...
An Introduction to the Tsetlin Machine - dScience – Centre for Computational and Data Science
Discover the Tsetlin Machine, an AI method that learns logical, human-understandable rules while matching neural networks in performance. Led by Prof. Ole-Christoffer Granmo, this seminar covers its f...
www.uio.no
March 4, 2025 at 10:56 AM
olegranmo.bsky.social
Our pioneering #TsetlinMachine #AI #chip analyzes 60,000 images per second while consuming just 8.6 nanojoules per image - that's 230,000,000x less energy than the energy of a single sugar crystal! Paper: arxiv.org/pdf/2501.19347 #ByeByeBlackBoxes #StayInCharge #GreenAI #DemocraticAI #ScientificAI
February 3, 2025 at 1:38 PM
olegranmo.bsky.social
Jem Davies on #explainability without loss of #accuracy and #efficiency with #tsetlinmachines! #GreenAI #LogicalAI #ScientificAI #DemocraticAI #ByeByeBlackBoxes

youtu.be/lqz8oNEbc3g?...
How to Balance Accuracy vs. Explainability in AI
YouTube video by Advantest
youtu.be
December 1, 2024 at 10:29 AM
olegranmo.bsky.social
The first #ASIC Coalesced Convolutional #TsetlinMachine chip has arrived! 🎉 Initial tests are successful, promising state-of-the-art energy efficiency and marking the beginning of the final phase of our research. #ByeByeBlackBoxes #GreenAI #DemocraticAI #ScientificAI #ML #AI #LogicalAI
November 22, 2024 at 8:42 AM
olegranmo.bsky.social
#Announcing the first #deep #tsetlinmachine! Reason #logically. Reason #multimodally (images, text, measurements, etc.). Reason #understandably. Reason #efficiently. Reason #safely. github.com/cair/GraphTs...
#StayInCharge #GreenAI #DemocraticAI #ScientificAI #ByeByeBlackBoxes
GitHub - cair/GraphTsetlinMachine: Tsetlin Machine for Logical Learning and Reasoning With Graphs
Tsetlin Machine for Logical Learning and Reasoning With Graphs - cair/GraphTsetlinMachine
github.com
November 17, 2024 at 4:25 PM
lumetanews.bsky.social
Full analysis →

https://lumeta.news/en/genebench-pro-openai-evaluation-biological-models?utm_source=bsky&utm_medium=social&utm_campaign=auto_thread

#ScientificAi #ModelEvaluation #Genomics
GeneBench-Pro: OpenAI launches evaluation tool for biological AI models
OpenAI has unveiled GeneBench-Pro, a new evaluation tool specifically designed to measure AI model performance in molecular biology, genomics, and genetic sequence analysis tasks. The platform establishes new standards for benchmarking in the scientific domain.
lumeta.news
June 30, 2026 at 5:39 PM
aipoch.bsky.social
#OpenScience #AI #ScientificAI #OpenSource #ResearchInfrastructure #Reproducibility #MedicalAI #AIPOCH
July 13, 2026 at 8:35 AM
claudius-krause.bsky.social
And my #HEP-ML group at @mbi-oeaw.bsky.social is already training models on it!

#hpc #scientificAI
asc.ac.at Austrian Scientific Computing (ASC) @asc.ac.at · Jul 3
Austria's new #supercomputer was officially inaugurated today! Multi-Site Computer Austria #MUSICA is a distributed system located in Vienna, Linz and Innsbruck.

Learn more about it in our latest news article: asc.ac.at/news/2026/ne...

#hpc #highperformancecomputing #Austria
July 3, 2026 at 3:35 PM
wellfunded.bsky.social
Anthropic Launches Claude Science — and Plans to Develop Its Own Drugs

Anthropic launched Claude Science, an AI platform for researchers consolidating scientific tools.

#AI #Pharma #DrugDiscovery #Anthropic #ScientificAI

https://wellfunded.news/articles/anthropic-claude-science-drug-development
Anthropic Launches Claude Science — and Plans to Develop Its Own Drugs
Anthropic unveiled Claude Science, an AI workbench designed to unify fragmented scientific tools and datasets for researchers. But the company didn't stop there — it also announced plans to develop its own pharmaceutical drugs using AI.
wellfunded.news
July 3, 2026 at 3:00 PM