🖥️The answer remains unclear, senior biopharma correspondent Andrew Dunn reported from one of the world’s major AI research gatherings. AI has already upended areas like computer science, where LLMs thrive at tasks with clear instructions and examples of success and failure. But science, which often explores the unknown, typically doesn’t have those clear instructions and examples. 

Some companies, like Lila Sciences, haven’t been shy about making bold claims, Andrew wrote. The company compares some of its earliest research findings on RNA to a “Move 37” moment, referring to a significant time for AI when a Google DeepMind algorithm made a move in the board game Go that no human expert would have considered. It’s now building a 250,000-square-foot “AI science factory” in Cambridge, MA, that Lila co-founder Ben Kompa thinks can be like another game, or a “gym,” for AI. “You have to set the game board and allow AI to play,” Kompa said.

Others question if LLMs alone will be enough in science. Patrick Schwab, GSK’s vice president of AI for science, sees LLMs as more like orchestrators of a suite of biology-focused, task-specific models. Schwab said he’d like to see the field focus more on causality. It hasn’t been a priority for the LLM world, and he believes that non-LLM models are the way ahead. You can read Andrew’s full analysis here. 

👀As the two best-selling checkpoint inhibitors approach their patent cliffs, the biopharma industry is coming to terms with the fact that it may not be able to repeat its success. The result is that the “search for a magic bullet has faded,” Evan Rachlin, co-founder and managing partner of the life sciences VC firm Ascenta Capital, told Endpoints News.

With Merck’s Keytruda and Bristol Myers Squibb’s Opdivo both expected to lose their market exclusivity around 2028, “segmentation” of both cancer treatments and patient populations is coming. Rachlin, whose firm invested in the recently acquired Tubulis, said that would likely involve matching specific patient populations with well-suited drug regimens without the one-size-fits-all approach of anti-PD-1 drugs. 

💉Eli Lilly pulled off three acquisitions at once this week for up to $3.8 billion total, aiming to increase its presence in infectious disease. The pharma is buying Curevo, LimmaTech Biologics and Vaccine Company. The trio of takeouts represents the eighth, ninth and 10th acquisitions Lilly has made so far this year. And we’re not even halfway through 2026.
 

The Zepbound maker hasn’t historically had a major presence in vaccines or infectious diseases. But now flush with cash from its obesity and diabetes franchise, Lilly is in an all-out sprint to broaden its activities into genetic medicines, sleep disorders and other areas it hasn’t had a large presence in until recently.

🔬In a pair of late-stage trials testing a chronic hepatitis B treatment, GSK said 19% of patients treated with bepirovirsen achieved a functional cure. The data are a striking advance over standard of care nucleoside or nucleotide analogs (NAs), which are associated with a 1% to 5% functional cure rate. The results met the benchmark set by Jefferies analysts, who wrote that functional cure rates of 15% to 20% would mark a “major improvement” over the standard of care.
 

🧬Cell and gene therapy is a “challenging market at the moment,” Charles River Laboratories CEO Birgit Girshick told Endpoints News. Three months ago, Charles River announced it would shed its manufacturing unit, which has a focus on cell and gene therapies. According to Girshick, other companies are likely considering doing the same thing. In the future, she thinks there could be a consolidation of smaller manufacturers in the space, so most players would be the larger service providers.