Lila Sciences
| Lila Sciences | |
|---|---|
| Industry | Artificial intelligence, laboratory automation |
| Founded | 2023, Cambridge, Massachusetts |
| Headquarters | Cambridge, Massachusetts, United States |
| Products | AI Science Factory instruments |
| Website | lila.ai |
Lila Sciences, which writes its name LILA, is a company in Cambridge, Massachusetts, that is building what it calls "scientific superintelligence": an AI model joined to laboratories that run experiments without a person at the bench. It was founded in 2023 in the laboratories of Flagship Pioneering and unveiled on March 10, 2025, with the stated aim of building "fully autonomous labs for life, chemical, and materials sciences".[1][2]
Molecular Reality Corporation, a nanopore company, puts Lila Sciences second in its brain lab column, with the caption "AI for scientific research".[3]
Approach
Lila's home page describes the system in one sentence: "With advanced AI and autonomous labs LILA generates hypotheses, designs and runs experiments." The operating system, it adds, "executes the entire scientific method autonomously." The page gives each part of the system an organ. "LILA's advanced AI model is the brain", and the company's "proprietary AI Science Factory instruments are the body".[4] The instruments are the physical half of what the literature calls a self-driving laboratory.
The About page gives two design principles. One is that people should build "systems that can learn for themselves" in place of writing expert knowledge into tools. The other is a preference for "a single, general platform for autonomous science, rather than many narrow, domain-specific tools". Its short statement of method is "Scale is the key to accelerating the scientific method."[2]
The launch announcement of March 2025 listed early results that the company claimed for the platform, among them "optimal genetic medicine constructs that outperform commercially available therapeutics". It described the work as "generating hypotheses and testing them in real-world environments."[1]
Open-endedness
A post of September 10, 2026, introduced a research group that works on a different kind of search. The group "explores evolution-inspired algorithms that favor serendipitous discovery over narrow optimization." Its members developed "quality-diversity" methods that "simultaneously optimized for both variety and success", and the post says the algorithms are "applied to projects such as RNA therapeutics and quantum dot design". They belong to what the post calls "the entire 'wheel of science,' a never-ending feedback loop of hypothesizing and experimentation."[5]
Green hydrogen catalysts
The company's most detailed account of its own work is a post of September 25, 2026, about the oxygen evolution reaction (OER), which takes place at the anode of a water electrolyzer that makes hydrogen. OER "is typically performed at high potentials and strongly acidic environments, corrosive enough to dissolve most metals". The catalyst "best able to survive those conditions while remaining active is iridium oxide", and "Nearly all iridium, for example, is produced as a byproduct of platinum mining". Ruthenium "is another catalyst option, but it is less stable."[6]
In the campaign the AI model proposed compositions, and the laboratory synthesized 96 catalysts at a time "in parallel through physical vapor deposition". The samples were tested for activity and stability in acid, and "Results flowed back into the AI model, which generated a ranked list of candidate compositions to make next." Lila writes that its AI-directed lab screened 2,942 oxides "across 53 systems and 26 elements" in three months, and "identified a total of 6 palladium-based material families on or near the pareto front for OER", the set of compositions that no other composition beats on every measure. The lead candidate "has now been tested in acid for over 1,000 hours", and the top composition "performed as admirably as ruthenium". Lila compares its pace with that of an ordinary laboratory: it "screened catalysts 17x faster than a standard lab".[6]
The account is open about what was not automated. For this work "the company opted to use humans to transfer samples from one machine to another". A preprint describes the work in detail, and the post says "The novel OER catalyst isn't a finished technology."[6]
Public programs
On July 22, 2026, Lila wrote that it had been "selected to join the U.S. Department of Energy's Genesis Mission". Three proposals it submitted across the program's Lighthouse Challenges were chosen for Phase I awards, on problems in "manufacturing, materials, and the physics underneath the electronics".[7]
In Molecular Reality's usage
Lila gives itself the same anatomy that Molecular Reality uses for the whole column, a brain and a body, and the smaller company's materials add a building. One slide calls laboratories of this kind centralized "AI science cathedrals", and sets against them small sensors in homes, sinks and drains.[3] Lila is also on the list of eventual customers in those materials, under the heading "Megacap 'Brain' AI Labs".[3]
See also
References
- ^ a b "Flagship Pioneering Unveils Lila Sciences to Build Superintelligence in Science" (Press release). Flagship Pioneering. March 10, 2025.
- ^ a b "About". Lila Sciences. Retrieved October 25, 2026.
- ^ a b c "Company records". Molecular Reality Corporation. October 2026.
- ^ "Lila Sciences". Lila Sciences. Retrieved October 25, 2026.
- ^ "What's a science startup doing with an open-endedness team?" Lila Sciences. September 10, 2026.
- ^ a b c "How an AI-run lab cracked open green hydrogen's catalyst problem". Lila Sciences. September 25, 2026.
- ^ "Powering American Science: LILA to join DOE's Genesis Mission". Lila Sciences. July 22, 2026.