Anthropic “splashed” $100,000 on a hackathon, using real scientific research data to put Claude Science to the test.

Anthropic, in collaboration with the Gladstone Institutes in San Francisco and the community organizer Cerebral Valley, will host a global online hackathon, "Built with Claude: Life Sciences," from July 7 to 13. It features two tracks: the lab track and the development track, with an estimated 500 finalists selected. Winning teams will share a total of $100k in API credits and usage quotas. This is the first major real-world application of Claude Science, the AI science workstation recently launched by Anthropic.
(Previous context: Anthropic spent $400 million to acquire AI biotech startup Coefficient Bio, directly challenging OpenAI)
(Background supplement: Google suffers talent loss! Nobel laureate John Jumper leaves DeepMind to join Anthropic)

Table of Contents

Toggle

• Clear division of labor between the two tracks
• The datasets provided by Gladstone this time are not practice problems
• 500 spots up for grabs first

Key Takeaways

• The hackathon "Built with Claude: Life Sciences" runs from July 7 to 13, with applications closing this Sunday (July 5)
• Choose between two tracks: the lab track uses real scientific data, and the development track builds scientific tools
• 500 finalists each receive one month of Max 20x plus $200 in credits; winning teams share $100k

Less than a week after the science workstation launched, Anthropic has thrown real immune T-cell sequencing data, DNA regulatory activity predictions, and protein interaction network data into the hackathon arena. Participants are practicing with real data, not submitting assignments with simulated datasets, directly pitting the newly launched tool against real laboratory data.

According to Anthropic's announcement, the organizing community Cerebral Valley is not new; previous Claude online build weeks were all run by this team, including the earlier Built with Opus 4.6 Claude Code hackathon. This time, they are collaborating with the Gladstone Institutes in San Francisco, targeting researchers, doctors, bioinformatics experts, and biotech developers.

Clear division of labor between the two tracks

The lab track targets those who want to do hands-on research. The rules are to use the science workstation Claude Science to pick a specific biological problem and ultimately produce reproducible analysis results or a trained model; presentations won't pass. The development track takes a different approach: contestants use the development tool Claude Code to build software tools that can be readily used by researchers, clinics, or biotech companies. The two tracks compete separately with non-overlapping judging panels.

The datasets provided by Gladstone this time are not practice problems

The Gladstone Institute opens three types of real scientific data to participants: immune T-cell sequencing, DNA regulatory activity prediction, and protein interaction networks. The sequencing data tests whether participants can extract signals from immune responses; DNA regulatory activity prediction involves gene expression mechanisms; and protein interaction networks directly touch on core issues in drug development. All three datasets are challenging. Opening private databases to external participants for practice is not a routine action for research institutions; Gladstone's move this time is a rare exception.

For context, Claude Science is an AI science workstation released by Anthropic on June 30, currently in beta, supporting macOS and Linux, and available to Claude Pro, Max, Team, and Enterprise users.

It integrates workflows that researchers previously needed to switch between multiple software packages into a single agent-driven work environment, with over 60 built-in scientific skills covering genomics, proteomics, single-cell analysis, and cheminformatics, as well as 3D protein structure and molecular visualization. It includes a reviewer agent that catches citation and calculation errors. Every figure comes with underlying code and message logs for verification and reproducibility. CNBC calls this Anthropic's "AI drug discovery" initiative.

500 spots up for grabs first

The organizers will select 500 participants from applicants. In previous build weeks of the same series, they selected 500 spots from over 20,000 applicants; this selection ratio is likely to follow the same logic. 20,000 people competing for 500 spots—the competition begins at the moment of registration. Regardless of ranking, each finalist receives one month of Claude Max 20x access plus $200 in API credits to practice. The real prize goes to the final winners, with winning teams sharing a total of $100k in API credits and usage quotas. Applications close this Sunday (July 5); to register, you must submit your information on the registration page.

This is Anthropic's first step in moving Claude Science from launch to real laboratory data.

Frequently Asked Questions

How do I register for the Built with Claude: Life Sciences hackathon?

Submit an application on the official registration page; the deadline is July 5 (Sunday). 500 finalists each receive one month of Claude Max 20x and $200 in API credits.

What is Claude Science?

An AI science workstation released by Anthropic on June 30, currently in beta, supporting macOS and Linux, with over 60 built-in scientific skills and a reviewer agent. CNBC calls it Anthropic's AI drug discovery initiative.