One row per AI lab; part of a comparable battery across…: When will Discovery Loop first have a model at the AI…

Part of: One row per AI lab; part of a comparable battery across new AGI-style labs with multibillion-dollar war chests, gauging whether any could rival OpenAI or Anthropic. Question per row: when will the named lab first have a frontier model? Resolution: the earliest date at which either (a) a model from the lab appears in the top 5 of the Artificial Analysis Intelligence Index (or its successor flagship overall-quality leaderboard) at any snapshot, or (b) for unreleased or unlisted systems, credible reporting plus at least one independent expert assessment establishes that the lab has trained a model broadly regarded as frontier-class, comparable to the contemporaneous top models from OpenAI, Anthropic, or Google (the way Anthropic's unreleased Mythos-class model inside Project Glasswing was publicly established before launch). If the lab is acquired or wound down first, or the event seems unlikely for many years, express that through late percentiles. Anchors: the current index top-5 is occupied by Anthropic and OpenAI variants; frontier generations currently require training runs costing billions of dollars and multi-hundred-megawatt compute. — view all 6 rows

When will Discovery Loop first have a model at the AI frontier, per the resolution criteria?✓#3

frontier_dateJun 1, 2036

10% by
Aug '29
25% by
Sep '31
median
Jun '36
75% by
Jan '45
90% by
Jan '60

rationale:

The Nature of Discovery Loop Discovery Loop was announced today (August 5, 2026) as a public benefit corporation by Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le, with the mission of automating the experimental loop of ML, science, and engineering research 2 sources. The seed round is co-led by Radical Ventures and Khosla Ventures, with Alphabet participating 2 sources. Notably, the company explicitly frames itself as consuming frontier AI models and large-scale computational infrastructure to orchestrate automated loops, rather than producing a flagship general chatbot 2 sources. The initial compute agreement with Google Cloud is only for one year 2 sources, and early analyses view the venture as an infrastructure and discovery layer that is deliberately orthogonal to general-model leaderboards 2 sources.

The Resolution Bar and Plausible Paths Placing in the top 5 of the Artificial Analysis Intelligence Index is a brutally high bar, currently occupied by flagship models from Anthropic, OpenAI, and Moonshot 2 sources. Because incumbents dominate multiple slots, top-5 entry requires a broad, multi-billion-dollar general model. Discovery Loop’s most plausible route to resolution is via independent expert assessment of an unreleased model (the second resolution path). If their recursive ML-automation thesis works, a headline demonstration model could be quietly trained and vouched for by experts. However, this path still demands that the system be broadly comparable to contemporaneous OpenAI, Anthropic, or Google flagships, rather than merely an impressive agentic harness on top of Gemini.

Base Rates for Elite Labs Even for elite-founder labs explicitly built to win the general-model race with enormous capital, reaching the frontier takes years. xAI took ~28 months to reach the frontier with Grok 4 and only returned to the top tier with Grok 4.5 in July 2026 artificialanalysis.ai. SSI, heavily funded and valued at ~$32B, had released nothing two years after its June 2024 founding turingpost.com. Moonshot AI took roughly three years to debut Kimi K3 at number 3 in July 2026 artificialanalysis.ai. A lab that is not currently aiming at the general frontier should be expected to take even longer, setting an absolute floor of roughly three years if they pivot immediately and raise billions for a massive training run.

Upside Risks for Early Resolution There remains a credible minority chance of an early resolution (reflected in the early-to-mid 2030s estimates). The founding team is arguably the world's strongest for large-scale training and ML infrastructure — Vinyals co-led Gemini, while Dean and Ghemawat built Google's ML infrastructure. Star-founder AI labs can easily raise multi-billion-dollar war chests within months. Furthermore, the Alphabet partnership could quietly underwrite a very large training run with dedicated TPU capacity the-decoder.com. If recursive ML-automation produces a sharp algorithmic jump, a frontier-class model is a natural demonstration artifact.

The Heavy Tail and the "Never" Scenario The median and late percentiles are pushed deep into the 2030s and beyond to reflect a substantial probability that the resolution criteria are never met. The company’s mission may be completely satisfied by orchestrating existing frontier models. Furthermore, Dean's own "1% rule" advises founders to avoid competing on tasks that general models already do well teahose.com. Given the lean team and friendly Alphabet relationship, a terminal state where Discovery Loop operates as an infrastructure layer building on Gemini, or is eventually reabsorbed by Alphabet, is highly plausible. Consequently, the modal outcome is that the company succeeds (or fails) without ever fielding a model that triggers these specific frontier criteria.