One row per milestone; forecast the date each milestone is…: When will a Discovery Loop system first be credited with a…

Part of: One row per milestone; forecast the date each milestone is first reached. If a milestone may plausibly never happen, express that through late percentiles. Shared background: Discovery Loop is a Delaware public benefit corporation announced August 5, 2026 by Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le, automating the full experimental loop of research (propose, run, evaluate, iterate), initially in ML research and engineering, acting as its own first customer, with stated applicability to the 14 NAE Grand Challenges. Seed round co-led by Radical Ventures and Khosla Ventures with Lightspeed, Kleiner Perkins, Doerr Capital, and Alphabet participating, terms undisclosed and unclosed at launch; Google Cloud supplies first-year compute, and Google announced a joint research framework with the company on ML systems. At launch: zero employees beyond the founders, no office, a single generic Member of Technical Staff posting in Palo Alto, no safety or policy or sales roles advertised, and a stated plan to hire a lean in-person founding team. Reference class: Google DeepMind's applied-science record shows fast wins where the experimental loop is digital with objective evaluation (AlphaFold at CASP, weather models, algorithm discovery) and slow or failed outcomes where the loop runs through the physical world or institutions (Streams shut down, disputed GNoME synthesis, Isomorphic reaching first trials only after five years). AlphaFold took roughly four years inside a fully staffed lab from project start to recognized breakthrough. Peer startups: Safe Superintelligence emphasized a small team and took nearly a year to reach roughly 20 employees and has shipped nothing public; Thinking Machines took about seven months to its first technical blog post and about eighteen months to a high-level safety document; Sakana AI took eight to twelve months to its first automated-discovery artifacts and saw its attribution claims disputed. — view all 4 rows

When will a Discovery Loop system first be credited with a substantive research contribution outside core machine learning?✓#2

milestone_dateAug 15, 2030

10% by
Feb '28
25% by
Dec '28
median
Aug '30
75% by
Jan '33
90% by
Mar '37

rationale:

What the criteria require and current status To qualify, Discovery Loop must secure authorship-level credit or be the central method in an independently verified, non-ML/non-computer-systems scientific advance. Independent experts must validate it as genuinely novel. As of its launch on August 5, 2026, the company has zero employees beyond the four founders, no office, and no domain scientists 3 sources. Critically, its stated sequencing is explicitly ML-first: the company will automate ML research and act as its own first customer before generalizing to the NAE Grand Challenges. Its initial Google Cloud partnership is heavily scoped to ML systems blog.google. This means early targets like chip design and compiler optimization are explicitly excluded by the resolution criteria.

Reference classes and timeline compounding History suggests a multi-year lag from founding to verified scientific breakthroughs. AlphaFold required roughly four years to achieve a recognized breakthrough within a fully staffed, mature lab 2 sources, while Isomorphic Labs took five years to reach first trials reuters.com. Even successful "AI scientist" efforts, like FutureHouse or Google's AI co-scientist, took 1.5–2 years from launch to peer-reviewed domain credit. Discovery Loop must first hire a team, build infrastructure, prove the automated ML loop, port it to a science domain, and survive peer review. Compounding these sequential steps places the median estimate around late summer 2030, roughly four years post-launch.

Paths to early resolution The fast tail (p10 in early 2028) hinges on Discovery Loop leveraging its founders' pedigree to quickly target a digitally evaluable science domain. If the company partners with academic institutions on computationally heavy fields—such as genomics, computational chemistry, or climate modeling—it can bypass the massive delays of physical lab build-outs. Fast wins are possible where the experimental loop is entirely digital with an objective scoring metric, and a high-profile collaboration could yield a preprint or accepted major-venue paper within about two years.

Risks of delay or failure to resolve The long right tail (stretching to 2037) reflects substantial probability mass on the company failing to cleanly meet the criteria for a decade or more. Discovery Loop may remain permanently focused on highly profitable ML and systems engineering. Commercial IP incentives could prevent them from pursuing open peer-reviewed publications, akin to Safe Superintelligence's lack of public shipping. Finally, the "domain experts treat it as a real advance" clause is a high hurdle. AI-originated discoveries frequently face intense scrutiny; past claims from systems like GNoME, Kosmos, and Sakana AI were heavily disputed regarding true novelty or scientific relevance 2 sources. Failure to win over domain skeptics, or a quiet acquisition by Alphabet, could permanently delay resolution.

Accounting for a roughly 22% chance of a breakthrough by August 2028 driven by a potential fast pivot into purely software-like or computational science domains slightly pulled forward the earliest percentiles, while leaving the median and long tail unchanged.