Sure! I was building a workflow that assesses human gene model accuracy. Which part is supported by what evidence? Important for drug development; if we want to change $THING it better exist in the first place.
I came up with a set of rules, collated external databases, and then slowly (Claude-code assisted) built a Nextflow pipeline that gives me an automated report, which I then manually expand by 'human' assessment of the evidence.
Claude Science prompted with 'assess this gene' came up pretty much with the same rules, built a report, and did the manual assessment of evidence pitfalls for about 10% of a Claude Max subscription's tokens in about 15 minutes. Some details differ from my report - a different tool here or there - but overall, what we needed out of these reports is in the Claude Science report.
Consider that if this was ran from your machine - it would have all of your memories and might be relying on the past 2 months. I'd be curious what would happen if it was run independently.
(I am assuming they didn't train on your prompts, which is always a worry)
I came up with a set of rules, collated external databases, and then slowly (Claude-code assisted) built a Nextflow pipeline that gives me an automated report, which I then manually expand by 'human' assessment of the evidence.
Claude Science prompted with 'assess this gene' came up pretty much with the same rules, built a report, and did the manual assessment of evidence pitfalls for about 10% of a Claude Max subscription's tokens in about 15 minutes. Some details differ from my report - a different tool here or there - but overall, what we needed out of these reports is in the Claude Science report.