Generated by Codex with GPT-5

What this article is about

This summary covers The Economist’s April 11th, 2026 Science & technology article listed in the contents as Curing all diseases with AI and published under the headline Panacea seer.

The article is built around a large claim: that artificial intelligence may eventually become a general-purpose engine for inventing new medicines, rather than just a tool for speeding up narrow research tasks. The immediate focus is Demis Hassabis, the head of Google DeepMind, and his spin-off company Isomorphic Labs. The broader point is that AI drug discovery is moving from promise and publicity towards something more concrete: actual candidate drugs that can be tested in the real world.

From protein prediction to drug design

The article treats AlphaFold, DeepMind’s protein-structure system, as the crucial starting point rather than the final breakthrough. Predicting how proteins fold was scientifically important and won Sir Demis a share of the Nobel prize in chemistry in 2024, but the real ambition always went further. The goal was to use that foundation to design therapies for disease.

That is where Isomorphic Labs comes in. The company, which Google spun out five years ago, is described as an AI-powered pharmaceutical firm built to turn DeepMind’s research into medicines. According to the article, it now has 19 programmes across cancer, cardiovascular disease and immunology, combining internal projects with partnerships involving large drugmakers such as Eli Lilly, Novartis and Johnson & Johnson. The message is that the lab is no longer just refining research tools. It is trying to produce a repeatable system for generating drugs at scale.

Why the next model matters

The article argues that AlphaFold alone was not enough for that job. Knowing a protein’s shape is useful, but drug design requires more detailed predictions about how molecules will behave when they meet biological targets. The new internal model highlighted here, called IsoDDE, is meant to handle more of that work. In particular, it predicts biochemical interactions such as binding affinity, which matters because a promising drug must attach strongly and selectively enough to do something useful in the body.

That distinction matters because it shows where AI’s commercial and medical value may lie. The biggest gain is not simply mapping biology more elegantly. It is reducing the trial-and-error burden of pharmaceutical research by helping scientists rule out weak candidates earlier and identify stronger ones faster. The article presents this as the first real step toward what Hassabis describes as a generic technology for drug discovery: a platform that, once built, could be turned against many diseases in quick succession.

Science, power and gatekeeping

The most interesting tension in the piece is that it is not only about scientific progress. It is also about control. The article notes that IsoDDE is being kept proprietary, partly because of commercial incentives and partly because of biosecurity concerns. A model that can help design beneficial compounds could, in principle, also help bad actors reason about harmful ones.

The Economist uses that point to widen the lens. It compares DeepMind’s caution to Anthropic’s decision, in the same week, to restrict access to a powerful model because of cyber-risk. That leaves a small number of lab leaders making consequential judgments about what capabilities should be released, to whom and when. Hassabis presents himself as a scientist trying to build the best tool for science, but the article suggests that modern AI chiefs are also becoming unelected gatekeepers over technologies with public-health and security implications.

The takeaway

The article does not claim that AI is on the verge of literally curing every disease. Its argument is subtler and more credible than that. It says the field may be crossing from a phase of impressive demonstrations into one where AI systems begin to generate real drug candidates and reshape how medicines are discovered. If that happens, the effect could be large even before any grand dream of universal cures is realised.

In plain English: The Economist sees AI drug design as one of the first areas where frontier models might create obvious practical value outside software and chatbots. The real test now is no longer whether these systems can produce exciting scientific outputs on paper, but whether they can help deliver medicines that work, and whether the people controlling them can be trusted to manage both the opportunity and the risk.