🤖 It’s Getting Better at Reading Your Thoughts

💭 It’s Getting Better at Reading Your Thoughts

⚗️ Anthropic’s New AI-Native Workspace for Scientists

📍 6 AI Stock Sectors

🛠️ 5 Trending Tools

🥪 4 Brief Bites

💰 3 Funding Frontlines

💼 4 Job Opportunities

BCIs establish seamless digital pathways between the human brain and external devices. In simple terms, they translate thoughts into actions. Imagine operating a robotic arm just by thinking about it. Tech billionaire Elon Musk’s Neuralink famously engineered the N1 Implant, a coin-sized chip that rests beneath the skull, equipped with flexible threads that insert into the primary motor cortex to interpret neural activity: the raw signals that carry everything we think, feel, and do. The key challenge with BCIs is that they typically require invasive brain surgery, which involves removing portions of the skull.

“B2Qv2” addresses this key challenge by leveraging MEG, a non-invasive magnetic field scanner that measures the electrical activity generated by neurons within the human brain. “B2Qv2” reads a person’s raw MEG signals and processes them through a fine-tuned Qwen2-4B language model, which aligns neural activity with likely words to reconstruct sentences.

“B2Qv2” was trained on the raw MEG signals of 9 volunteers who collectively spent about 90 hours typing approximately 22,000 sentences. It coherently recovered sentences from noisy neural inputs with a word accuracy rate of 61%, significantly improving upon the word accuracy rate of 8% from other non-invasive brain recording methods.

Imagine waking up fully conscious, with every thought clear, concise, and complete, yet finding yourself trapped within a motionless shell. Unable to move, your mind becomes a restless prisoner, alive and aware but painfully disconnected from reality. For people with severe paralysis, this nightmare defines their reality.

By bridging the gap between non-invasive brain recordings and advanced language modeling, “B2Qv2” paves the way for a future where thoughts aren’t trapped by physical limitations. These biomedical breakthroughs are enabling us to transcend the boundaries of our bodies.

It equips Claude with direct access to more than 60 scientific toolkits and medical databases, including PubMed, R Project, and Jupyter Notebook. For context, it’s estimated that global medical knowledge now doubles every 73 days, with 1.5 million new articles and citations added to PubMed every year.

It also features a multi-agent system, which deploys subagents to handle highly specialized technical tasks, such as modeling 3D protein structures or sequencing single-cell genomic data. For example, subagents leverage NVIDIA’s BioNeMo Agent Toolkit to engineer bimolecular models that simulate how DNA and RNA bind to specific cellular targets.

It also provides rendered LaTeX previews for drafting and revising scientific manuscripts. LaTeX is a widely used markup language for defining the content, structure, and format of scientific documents. For example, instead of clicking “footnote,” “italicize,” and “heading,” scientists type “\cite,” “\emph,” and “\section.”

Traditional scientific research is very time-consuming and highly labor-intensive. Scientists devote months, or even years, to analyzing empirical datasets, reviewing academic literature, and conducting controlled experiments.

Anthropic’s tackling one of the most complex challenges in scientific research: finding effective ways to organize, analyze, and synthesize raw data into actionable insights to drive new scientific discoveries.

Jarvie raised an $8.3M Seed Round to bring AI agents to group chats.

Pocket closed an $11M Series A to always take notes in the real world.

Omen AI landed a $31M Series A to fight data center coolant bacteria.

TikTok (London, UK): Data Analyst Project Intern, TikTok Live, Fall 2026

Snowflake (Bellevue, WA): AI Research Scientist, Agents, Entry-Level

Meta (Menlo Park, CA): Production Engineer, AI Systems, Mid-Level

Google (Kirkland, WA): Staff Data Scientist, Product, Senior-Level