Anthropic's Claude Science Targets Drug Discovery
Anthropic unveiled Claude Science at a closed event for pharmaceutical executives, biotech founders, and academic researchers — positioning it as the scientific equivalent of Claude Code, the company's autonomous software engineering agent.
Like Claude Code, Claude Science can execute complex, multi-step tasks from high-level natural language instructions. Key capabilities include:
- Tools purpose-built for computational biology and drug development
- Autonomous execution of research workflows with minimal human hand-holding
- Internal use by Anthropic itself to accelerate research into rare and neglected diseases
The launch signals a deliberate strategic bet: Anthropic believes AI can meaningfully compress the drug development pipeline, which currently averages more than 10 years and billions of dollars per approved therapy.
California's Carbon Manure Math Doesn't Add Up
California runs a subsidy program that pays cattle farmers to capture methane emissions from manure and convert them into pipeline-ready natural gas. It has become enormously popular — primarily because the financial incentives are exceptionally generous.
But new research suggests the program is a case study in the limits of carbon offsetting schemes:
- Rather than forcing polluters to reduce emissions directly, the system shuffles climate responsibility between industries and regions
- Critics argue it allows industries to continue polluting by paying others to offset elsewhere
- Modeling indicates the approach could lock in additional warming over the long term
The broader concern is that incentive-based offset markets, however well-intentioned, can substitute accounting gymnastics for actual emissions reductions.
The Dark Matter Hunt Gets a Reboot
For decades, physicists chased weakly interacting massive particles (WIMPs) as the prime dark matter candidate. That search has hit a fundamental wall: the so-called "neutrino fog."
Particles streaming from the sun and distant stars now drown out potential dark matter signals in detectors — but researchers aren't giving up. New detection strategies under consideration include:
- Quantum sensors with greater sensitivity
- Liquid-helium detectors for lower-threshold detection
- Atmospheric searches, including in Jupiter's atmosphere
The field is effectively widening its aperture, betting that novel detector physics can circumvent the neutrino noise problem.
Headlines Worth Your Attention
- US lifts export restrictions on Anthropic's Mythos and Fable models after extended negotiations — though analysts warn the crackdown already gave Chinese AI rivals an opening
- The Vera C. Rubin Observatory in Chile has begun the most detailed survey of the universe ever attempted, using the largest digital camera on Earth
- Tech talent exodus: H-1B visa instability is pushing skilled workers to consider Canada, the UK, and the Gulf states
- Trump reported $635 million in meme coin royalties in 2025, with total crypto-related income exceeding $1 billion
- The UN is warning that rapid AI adoption risks deepening global inequality, and is proposing a shared governance framework
- Netflix and ElevenLabs used AI to recreate Gene Wilder's voice for a new Willy Wonka series — with the estate's blessing
Quote of the Day
"Caveman save you token, save you money."
— The README for a GitHub plugin that strips LLM outputs down to minimal, terse language as a cost-cutting measure. A senior OpenAI employee was among its contributors.
AI Drug Discovery: Promise Meets Practice
A growing cohort of startups is deploying machine learning models to predict how candidate drug compounds will behave in the body — discarding dead-end molecules before they ever reach a lab.
The pitch is compelling: slash the 10+ year, multi-billion dollar development timeline by front-loading computation. But the field remains early-stage, with many companies making claims that outpace their evidence. The technology is beginning to deliver real results, but it is not a silver bullet.
