The post was Dario Amodei responding to criticism of AI regulation and public messaging. His core claim was that public distrust is real, that marketing will not fix it, and that the only durable answer is to produce obvious public benefit. He pointed to biology and medicine as the place Anthropic hopes to prove that case, with “curing cancer” as the emblematic example. He also repeated a familiar Anthropic line that AI tends to concentrate power for structural reasons, so regulation should focus on frontier labs while carving out room for smaller players.
People were not buying it. The main reaction was that this reframes concrete present-day complaints into a future promise. Job destruction, higher power and hardware costs, copyright fights, military use, scams, market froth, and general concentration of power are the reasons many people distrust AI companies now. Saying the answer is to push even harder until medicine breakthroughs arrive sounded less like accountability and more like a demand for patience while the same firms deepen the very dynamics people object to.
The “cure cancer” example landed especially badly because many readers took it as a category error. Biology is slow,
wet-lab and trial constrained, and full of bad data and messy feedback loops. Several people who were otherwise positive on AI still said this is not like math or coding, where outputs can be quickly checked and iterated. The more grounded pro-AI view in the conversation was narrower. Models can help with parts of drug discovery, administration, coding, literature review, and experiment planning. They do not remove the long clinical and physical bottlenecks, so any real impact is likely to be incremental, indirect, and credited to scientists and institutions using the tools rather than to Anthropic “curing cancer.”
A second theme was trust in the messenger. Even some people who thought Amodei may be sincere said Anthropic has wrapped itself in paternalistic safety rhetoric while keeping models closed, expensive, and tightly controlled. Critics saw the regulation argument as obvious incumbent self-interest dressed up as governance.
Open weights were defended not as a full solution, but as one of the few available checks on dependence. Others accepted Amodei’s structural point that compute-heavy AI naturally centralizes power, yet concluded that this makes access and
antitrust more urgent, not less.
Under all of it was a broader mood shift. The conversation read less like technical disagreement about model quality and more like a labor and political backlash from people who think AI leaders are asking society to absorb immediate costs in exchange for speculative future miracles. Even commenters who use and like the tools often framed them as “very useful” rather than civilization-bending, which made the grand promises sound like financial and regulatory positioning rather than a credible roadmap.
twitter.com