Google AI co-scientist claims it independently discovered a leukemia drug and novel antibiotic mechanisms — hosts urge caution

Summary

Google's AI co-scientist system claims to have independently discovered a new leukemia drug candidate that tested successfully in vitro, along with novel liver fibrosis drug targets and bacterial antibiotic mechanisms. The company announced the development as part of a broader push into AI-assisted scientific research, inviting researchers into a trusted tester program.

The hosts treat the announcement with significant skepticism. One core issue: Google's credibility on AI demos has eroded after the company was caught speeding up a previous demonstration. That history now colors how claims like this land.

Beyond trust deficit, the substantive question is whether this represents genuine scientific breakthrough or expensive compute applied to well-trodden ground. In vitro success is described as "easy" — most drug failures happen later in development, after animal models and clinical trials. The comparison the hosts draw is instructive: when DeepMind solved protein folding with AlphaFold, it solved one of biology's hardest computational problems. Yet biotech stock markets barely moved. It turned out the problem was hard but not economically fundamental — existing machines and PhD students running X-ray crystallography could do the work. The breakthrough didn't reshape pharma profitability.

The same uncertainty applies here. A reasoning agent that generates hypotheses and screens compounds is genuinely useful tooling for research labs. But whether it represents a structural shift in drug discovery economics, or just another speedup in a process with many bottlenecks downstream, remains unclear. The hosts note they'll watch biotech equity performance as a real-time test: if this is truly transformative, markets should reflect it, though public pricing may lag.