Martin Shkreli has criticized Anthropic’s reported drug-discovery claims involving its Claude artificial-intelligence system, declaring: “This is not impressive work.” The criticism puts a prominent skeptic against claims about AI-assisted pharmaceutical research, but the available account does not establish which experiment, result or benchmark prompted his objection.

The confirmed development is narrow: Shkreli rejected the quality of the work associated with Anthropic’s reported claims about Claude and drug discovery. His quoted assessment is categorical, but no accompanying technical argument, data analysis or description of the disputed task is available. It is also unclear whether he was challenging Claude’s output, Anthropic’s interpretation of that output or the way the work was presented publicly.

Anthropic’s underlying claim is not described in enough detail to evaluate independently. The available information does not identify a candidate drug, biological target, laboratory result, peer-reviewed paper, preprint or benchmark. It also does not establish whether Claude generated a new hypothesis, reviewed existing research, assisted with a computational workflow or performed another research task.

No response from Anthropic is included, and there is no confirmed indication that the company accepted or disputed Shkreli’s characterization. Without the company’s methodology and Shkreli’s full reasoning, the exchange remains a public criticism rather than a resolved technical finding.

Evidence Behind Claude’s Pharma Claims

The dispute matters because drug-discovery claims carry scientific and commercial weight. AI systems can assist researchers with literature review, molecular analysis and hypothesis generation, but a useful output is not the same as a validated treatment. Laboratory testing, reproducible methods and clinical evidence determine whether a proposed discovery has medical value.

Shkreli’s criticism also highlights the need to distinguish a model demonstration from verified scientific progress. Readers and investors cannot judge the value of the reported work from a forceful reaction alone. The relevant questions are what Claude was asked to do, how its performance was measured, whether specialists checked the output and whether independent researchers could reproduce the result.

That distinction affects how claims about AI in pharmaceutical research are understood. A system might reduce the time needed for one research step without discovering a viable medicine. Conversely, an apparently modest result could still be useful if it improves accuracy, cost or speed under controlled conditions. Neither possibility is established by the limited information now available.

From Model Output to Medicine

Drug discovery is a multistage process that can include identifying biological targets, screening compounds, evaluating toxicity and testing candidates in laboratory, animal and human studies. An AI model’s contribution must be described precisely because success at one stage does not confirm success across the full process.

Claude is identified in the dispute as Anthropic’s AI system, but no model version, research partner or experimental setting is specified. That missing detail matters because performance can vary by model release, prompting method, data access, external tools and the expertise of the people reviewing the results.

The phrase “not impressive work” communicates Shkreli’s judgment but does not reveal his standard of comparison. The account does not say whether he compared Claude with human researchers, specialist software, other AI systems or a baseline method. It also does not show whether his assessment underwent any outside review.

“This is not impressive work.”

— Martin Shkreli

Missing Data Limits the Dispute

Several core facts remain unknown. There is no detailed description of Anthropic’s drug-discovery claim, the evidence supporting it or the exact role Claude played. The available account also does not provide the date, platform or full setting in which Shkreli made his comment.

It is not yet clear whether any findings were peer-reviewed, released as a preprint or presented only as a company demonstration. No independent replication, laboratory validation or comparison with established methods is identified. Those omissions prevent a firm assessment of either Anthropic’s reported work or Shkreli’s rejection of it.

Shkreli’s qualifications for evaluating this particular result, any financial interest connected to the dispute and whether he reviewed unpublished material are also unspecified. His statement should be treated as an attributed opinion, not independent proof that Claude’s reported performance was weak.

Technical Details Could Settle Debate

The next meaningful development would be the release of Anthropic’s methodology and results, including the model version, task definition, comparison standard and validation process. A full response from the company could also clarify whether Shkreli addressed the work accurately and whether the disputed claim was scientific, commercial or promotional.

Independent specialists would then need to examine whether Claude produced original, reproducible value and whether any claimed advantage held up against existing research tools. Until those details appear, the disagreement remains unresolved and readers should avoid treating either side’s position as a settled scientific conclusion.

Source: Anthropic

Key Questions

What did Martin Shkreli say about Anthropic’s work?

Shkreli said, “This is not impressive work.” The available account links that criticism to Anthropic’s reported Claude drug-discovery claims but does not provide his full argument.

What did Claude reportedly do in drug discovery?

The specific task is not identified. It remains unknown whether Claude analyzed molecules, proposed a target, summarized research or assisted with another part of the drug-development process.

Has Anthropic responded to Shkreli?

No confirmed Anthropic response is included in the available information. The company’s position on Shkreli’s criticism is currently unclear.

Was the reported work peer-reviewed?

There is no confirmed information showing that the work was peer-reviewed, independently replicated or laboratory-validated. Its publication status and supporting evidence remain unspecified.

Does the criticism prove Claude failed?

No. Shkreli’s comment is an attributed judgment, while the underlying evidence is unavailable. Determining whether Claude succeeded would require clear benchmarks, methods and independent evaluation.

Source: Anthropic