After the panel had spent much of the episode arguing about whether AI progress should be slowed, Peter Diamandis moved the conversation to the subject his audience had voted second most important: health and longevity. He opened with a disclosure. Insilico Medicine, the company he was about to praise, is one of his portfolio companies; its founder Alex Zhavoronkov is, he said, "a dear, dear friend," a participant in his longevity XPRIZE, and someone Diamandis backed as one of his earliest investors and advisors.
The company's platform, he explained, does not simply screen molecules that already exist. It invents new ones. And its lead compound now had two pieces of news attached to it in the same week.
The drug and the disease
The compound is rentosertib — a name the panel mocked before conceding it was at least pronounceable, and that the FDA constrains how drugs are named anyway. What makes it unusual, in Diamandis's account, is that the AI identified the disease target and then designed the molecule to hit it.
The disease is idiopathic pulmonary fibrosis, or IPF: a scarring of the lungs, more common with age, that Diamandis described as killing most patients within three to five years of diagnosis. He said the New York Times had reported that rentosertib has advanced to phase three trials — the last stage before regulators decide on approval — and that no AI-designed drug had reached that stage before.
The second item was the one that made him sit up: an analysis of the earlier, smaller phase 2a trial in which six different protein-based aging clocks all pointed the same way. Treated patients' blood signatures, he said, looked like those of people three to six years biologically younger. Aging clocks, he acknowledged, "are all over the map" — but six agreeing, in his reading, "that's real signal."
What an aging clock is, and what this one measured
An aging clock is a statistical model. It takes a measurement from the body — here, the levels of many different proteins circulating in blood — and returns an estimate of how old a person looks by that measurement, which can differ from the number of birthdays they have had. Some clocks are trained to predict chronological age. Others are trained to predict mortality risk. They are not the same instrument, and they do not have to agree.
The published analysis, released on 7 September 2026 by Zhavoronkov and colleagues, sits inside a 12-week randomized, blinded, placebo-controlled IPF trial run in China between July 2023 and June 2024. Seventy-one patients were randomized and 55 completed treatment; 43 consented to the proteomic work and 42 had the measurements the analysis required. Blood serum was taken at the start and at weeks 2, 4 and 12, and six clocks were applied to those protein readings.
The movement was dose-specific. At week four, patients taking 60 mg once daily showed reductions of 2.71 to 3.46 years on four clocks trained on chronological age. Neither mortality clock changed significantly in that arm. The 30 mg twice-daily arm showed broader agreement across clocks. And the signal did not keep falling: by week 12 it had plateaued. The authors also note that the clocks share fibrosis-related proteins among their inputs, which complicates reading their agreement as independent confirmation — a drug that calms lung scarring may move several clocks at once because they are partly watching the same thing.
The authors treat the whole exercise as exploratory. Their proposal is that the result be replicated outside IPF, and then tested against clinical endpoints or put through formal biomarker qualification, before anyone calls it geroprotection.
The panel's bolder reading
The panelist Diamandis turned to, addressed on air as Alex, went further than the paper does. He used the finding to "double underline" a conjecture he had made before: that longevity escape velocity is already here, but "it's spiky, so it's only visible in subpopulations." He expects the term to follow the path of the Turing test or AGI — passed quietly, then argued about for years, with people disagreeing about the definition long after the line has been crossed.
His arithmetic came from the week-four peak. Four weeks of treatment, three to four years of apparent biological age reversal on the proteomic clocks. "Four weeks of input, negative three to four years of output," he said — which, against his working definition of longevity escape velocity as more than a year of output per year of input, he called an example on the margin. He then stacked his own caveats: it is only a subpopulation, it is a clinical study, "caveat, caveat, caveat." He described it as an AI-triggered narrow spike that looks like escape velocity in a portion of the population, and predicted more such sparks, with the debate about whether the threshold has been passed still running twenty years from now.
The study's own numbers put limits on that reading. The clocks that moved in the 60 mg arm were the ones estimating chronological age, not the ones estimating risk of dying; the effect did not compound over the trial's twelve weeks; and nobody was followed long enough to see whether any of it translates into extra time alive.
Why the regulator is the bottleneck
The conversation turned quickly to approval. Asked what the biggest roadblock to longevity drugs is, Diamandis's answer was that the FDA does not recognize aging as a codable condition — there is no box to tick, and so no straightforward path to trial and approval for a drug whose purpose is to slow aging itself. He mentioned that Aubrey de Grey is convening a summit, to which Diamandis is sending a video, on building specialized regulatory regions that could approve such drugs rapidly. A sovereign state, a city, a state — "basically, medical charter cities," he said, and expected people from everywhere to travel to them for early access.
A panelist named the version of this that has already happened: "regulatory arbitrage in this regard. There's a reason that Alex did this of first trials in China." Then the register changed. "My dad's actually got IPF," he said, "so I can't wait for this drug to hit the market."
That is the gap the paper leaves open. The phase three trial now running is testing rentosertib against lung disease, which is what a patient with IPF needs it to do. The aging-clock result is a separate, exploratory reading of blood proteins in 42 people, and its authors' next step is not a longevity claim but a replication in people who do not have the disease.