Anthropic's "AI for Science" Event:
John Jumper, Claude, and Drug Discovery at 10x Speed

The short answer: On June 30, 2026, Anthropic hosts its biggest life sciences moment yet. Nobel laureate John Jumper — the scientist behind AlphaFold — left Google DeepMind to join Anthropic. Claude Mythos 5 designs drug candidates roughly 10x faster than human scientists in internal tests. Novo Nordisk cut Clinical Study Report writing time by 90%. CEOs from Novartis, Bristol Myers Squibb, and Genentech are on stage.

This guide is for life sciences engineers, computational biologists, and pharma CTOs. It covers the full briefing agenda, Jumper's background, Anthropic's 18-month timeline, Claude for Life Sciences integrations, Mythos 5 results, the ~$400M Coefficient Bio deal, industry cost curves, US export restrictions, five things to watch today, and a six-step action plan.

01 What is Anthropic's AI for Science event?

The Briefing: AI for Science runs June 30, 2026 at 10:00 AM PST (1:00 AM CST July 1) from San Francisco, streamed globally.

Event essentials
Item Detail
Format In-person San Francisco + global livestream
Host Anthropic (maker of Claude)
Agenda Life sciences vision, product demos, customer spotlights

Who's speaking?

Speaker Role
Vas Narasimhan CEO of Novartis; Anthropic board member
Chris Boerner, PhD CEO of Bristol Myers Squibb
Aviv Regev EVP and Head of R&D, Genentech
Lotte Bjerre Knudsen Professor, DMSc; former Novo Nordisk Chief Scientific Adviser
Eric Kauderer-Abrams Head of Life Sciences, Anthropic
Jonah Cool Head of Life Sciences Partnerships, Anthropic
Matthew Herper Senior Writer, STAT News (moderator)

When Novartis and BMS CEOs share a stage at an AI company's event, it signals where pharmaceutical R&D is actually heading — not a marketing stunt.

Common pain points:

  • 12–15 year timelines, $2.6B+ per drug: Only ~10% of clinical candidates reach approval.
  • Tool fragmentation: Literature, ELN data, and trial registries live in silos.
  • Model access blocked: Mythos 5 remains restricted under US export controls for most international teams.
  • Generic chatbots lack connectors: No native Benchling, PubMed, or FDA-compliant drafting workflows.

02 Who is John Jumper, and why does his move matter?

John Michael Jumper was born in 1985 in Little Rock, Arkansas. He studied math and physics at Vanderbilt (2007), earned an MPhil in physics at Cambridge on a Marshall Scholarship (2008), and completed a PhD in theoretical chemistry at the University of Chicago (2017).

Six months after his PhD, he joined Google DeepMind on a secret project: predicting how proteins fold into 3D structures — a problem that had stumped biologists since the 1970s.

At CASP14 in 2020, AlphaFold 2 achieved accuracy matching experimental methods in hours instead of months.

AlphaFold by the numbers:

  • 214 million+ protein structures predicted
  • 2 million+ researchers across 190+ countries
  • Accelerated drug discovery, cancer research, and fundamental biology worldwide

In 2024, Jumper and Demis Hassabis won the Nobel Prize in Chemistry (with UW's David Baker). At 39, Jumper became the youngest chemistry laureate in over 70 years.

On June 19, 2026, Jumper posted on X: "After nearly nine years, I have decided to leave Google DeepMind and join Anthropic." Hassabis responded publicly that AlphaFold had changed the world and shown AI's potential in science and medicine.

His announcement came 11 days before today's briefing. Anthropic has not disclosed his role. Given April's Coefficient Bio acquisition and Claude for Life Sciences momentum, he likely shapes foundational biology AI — possibly a next-gen protein tool along the lines of "ClaudeFold."

03 Anthropic's 18-month life sciences timeline and Claude for Life Sciences

How we got here (Oct 2025 – Jun 2026)
Date Event
Oct 2025 Claude for Life Sciences launches with Benchling, 10x Genomics, PubMed
Feb 2026 Research partnerships with Allen Institute and HHMI Janelia
Apr 2026 Coefficient Bio acquired for ~$400M (all-stock)
May 2026 Andrej Karpathy joins Anthropic's pre-training team
Jun 9, 2026 Claude Fable 5 + Mythos 5 launch with major science capabilities
Jun 12, 2026 US government orders Fable 5 and Mythos 5 offline (export control)
Jun 19, 2026 John Jumper announces move to Anthropic
Jun 24, 2026 Bloomberg reports AlphaFold co-authors Adler and Pritzel may also join (unconfirmed)
Jun 26, 2026 Mythos 5 partially restored for ~100 US critical-infrastructure orgs
Jun 30, 2026 "The Briefing: AI for Science"

Claude for Life Sciences is an enterprise stack built on Claude Enterprise with MCP connectors and Agent Skills wired into tools scientists already use.

Platform integrations
Platform What Claude can do
Benchling Draft protocols, SOPs, informed consent from ELN/LIMS data
10x Genomics Single-cell and spatial transcriptomics processing
PubMed / bioRxiv / medRxiv Search, retrieve, synthesize biomedical literature with citations
Open Targets Identify and prioritize therapeutic targets
Medidata / ClinicalTrials.gov Trial enrollment monitoring and registry queries
Wiley Scholar Gateway / BioRender Full-text access and scientific figure workflows

Coverage spans early discovery through regulatory submissions: literature synthesis, hypothesis generation, experimental design, genomics QC, clinical protocol drafting, and FDA query responses.

04 How good is Claude Mythos 5 at actual science?

Drug design: 10x faster, fully autonomous. In internal testing, Mythos 5 accelerated key drug design steps approximately 10x. With protein design and bioinformatics tools but no human assistance, the model:

  1. Chose binding sites
  2. Selected and ran protein design tools
  3. Recovered from failures on its own
  4. Produced viable drug candidates

Result: 9 out of 14 protein targets yielded strong candidates under investigation — immune checkpoints, growth-factor signaling, neurodegeneration, muscle disease, and complex structural targets.

On AAV capsid assembly prediction (Dyno Therapeutics data), Mythos 5 outperformed dedicated protein language models built for that task.

In blind comparisons, reviewers preferred Mythos 5 hypotheses roughly 80% of the time over the previous Opus generation. One hypothesis identifying an E. coli antimicrobial target has been validated in lab testing.

Left running for one week unsupervised on a genomics project, Mythos 5 assembled single-cell data from 138 animal species and millions of cells, trained a custom ML model 100x smaller than a recent Science publication — with better performance.

Novo Nordisk (Ozempic maker): CSRs took months to draft. They built NovoScribe on Claude via Amazon Bedrock + RAG with expert-approved templates. Waheed Jowiya, Digitalization Strategy Director: "Claude has helped us cut writing times on CSRs by 90%." Expansion covers device protocols, patient materials, and full Common Technical Document automation.

Other deployments: Sanofi, AbbVie, AstraZeneca, Genmab, Bristol Myers Squibb, Komodo Health, and Axiom (toxicity prediction via Claude Code + MCP).

Where AI compresses the R&D curve
Stage Traditional With AI
Target ID Months of literature review Hours of signal extraction
Compound design Lab high-throughput screening Computational design, tens of times faster
Regulatory docs Months of manual drafting Compliant drafts in a fraction of the time

05 Coefficient Bio, competitive edge, and the export-control cloud

In April 2026, Anthropic acquired stealth startup Coefficient Bio for ~$400 million in stock. Founders Samuel Stanton and Nathan C. Frey came from Genentech Prescient Design. Their goal: ASI for Science in biology. Investor Dimension reportedly achieved 38,513% IRR on the deal. The team now sits under Eric Kauderer-Abrams, bringing protein design and biomolecule modeling depth.

Anthropic's edge vs OpenAI and DeepMind:

  1. Safety-first culture — Constitutional AI aligns with pharma compliance needs
  2. Vertical integration — connectors, Coefficient Bio, and Jumper's scientific credibility
  3. Locked-in enterprise customers — Novartis, BMS, Genentech, Novo Nordisk

Export controls as of June 30:

  • Jun 12: US government forced Fable 5 and Mythos 5 offline; foreign nationals blocked
  • Jun 26: Commerce partially restored Mythos 5 for ~100 vetted US organizations
  • Fable 5 public access still down; general access not restored

International pharma teams face real operational uncertainty. Can Jumper replicate AlphaFold at Anthropic? Honestly: unknown. AlphaFold had DeepMind infrastructure, CASP benchmarks, and years of institutional partnerships. Anthropic is a commercial LLM company pivoting into scientific AI.

Five things to watch today: Jumper's appearance and title; Mythos 5 biology access expansion; new Claude for Life Sciences connectors; Fable 5 restoration timeline; paths for non-US researchers.

06 FAQ, six-step playbook, and CALMVPS wrap-up

Q: What is Anthropic's AI for Science event?
A: A June 30, 2026 livestream showcasing Claude across life sciences, with demos and case studies from Novartis, BMS, and Genentech.

Q: Who is John Jumper?
A: AlphaFold 2 lead and 2024 Nobel Chemistry laureate who announced his move to Anthropic on June 19, 2026 after nearly nine years at DeepMind. Role undisclosed.

Q: How fast is Claude at drug design?
A: Internal tests show Mythos 5 accelerating key steps ~10x; one fully autonomous run produced candidates for 9 of 14 protein targets.

Q: Can I use Mythos 5 for biology research right now?
A: As of June 30, only ~100 vetted US organizations have access. Fable 5 has been suspended for the general public since June 12.

Q: What is Claude for Life Sciences?
A: Anthropic's October 2025 enterprise stack connecting Claude to Benchling, PubMed, 10x Genomics, Medidata, and more — from literature to regulatory filings.

Q: Who acquired Coefficient Bio?
A: Anthropic in April 2026 for ~$400M. The sub-10-person team came from Genentech computational biology.

Six-step playbook:

  1. Subscribe to Anthropic life sciences updates and STAT News for post-event product and policy changes.
  2. Map your toolchain gaps against current MCP connector coverage.
  3. Verify Mythos 5 eligibility or plan fallback models via LiteLLM/OpenRouter.
  4. Pilot low-risk workflows first — CSR drafting and literature synthesis like NovoScribe.
  5. Build multi-model routing before relying on a single vendor under export control.
  6. Run agents on dedicated bare-metal Mac nodes for 24/7 isolation, audit logs, and GxP-friendly environments.

Citable data points:

  • Drug design: ~10x speedup; 9/14 (64%) targets with candidates (Anthropic internal research)
  • CSR writing: 90% time reduction at Novo Nordisk (official case study)
  • AlphaFold scale: 214M+ structures, 190+ countries, 2M+ researchers

Sources (re-check links after upstream updates):

Anthropic: Claude for Life Sciences

Anthropic: Claude Fable 5 and Mythos 5

Novo Nordisk customer case (NovoScribe)

John Jumper Nobel lecture (2024)

Shared cloud VPS and laptops struggle with multi-day unsupervised genomics runs, sleep-induced job kills, and weak isolation for regulated workflows. For production Claude Code, MCP connectors, and bioinformatics CI that must stay online through policy swings, CALMVPS bare-metal Mac Mini rental delivers dedicated Apple Silicon, 120-second provisioning, and flexible monthly billing — swap API endpoints on an isolated node without rebuilding your entire stack when Anthropic's access rules change.

Written ahead of the June 30 livestream. Post-event updates recommended: product announcements, Jumper's official role, Fable 5 restoration, and new pharma partnerships.