In which domain will Ineffable Intelligence's first headline public result be, by December 31, 2028?

rationale:

The simulation-first strategy

Ineffable Intelligence's first public result is most likely to emerge from "Games or simulated environments" (48%). The company's stated mission is to build a "superlearner" that discovers knowledge from its own experience rather than human data. To achieve this, the founders are building "crucibles"—simulated worlds featuring open-ended objectives, multi-agent dynamics, and physics engines—which they view as the safest and fastest path to superintelligence David Silver and Lasse Espeholt of Ineffable Intelligence David Silver | Training Data Summary - Teahose David Silver and Lasse Espeholt of Ineffable Intelligence David Silver and Lasse Espeholt of Ineffable Intelligence. Their hiring aligns perfectly with this strategy: the team recently added multi-agent and 3D virtual world experts, including cofounders who worked on AlphaStar and Quake III Arena https://fortune.com/2026/09/07/ineffable-intelligence-hires-cofounders-hiring-google-deepmind-instadeep-flying-fish/ Exclusive: Ineffable Intelligence adds 6 "cofounders," hiring .... Furthermore, an open role for a "Simulated Universes Engineer" details the construction of MMO-scale, GPU-accelerated environments Simulated Universes Engineer @ Ineffable Intelligence https://jobs.ashbyhq.com/ineffable. David Silver's own track record (AlphaGo, AlphaZero, AlphaStar) heavily suggests that their first major demonstration will feature agents learning novel behaviors inside these simulated worlds The Man Behind AlphaGo Thinks AI Is Taking the Wrong Path The Man Behind AlphaGo Thinks AI Is Taking the Wrong Path.

The likelihood of staying in stealth

There is a substantial chance (26%) that Ineffable publishes nothing by the end of 2028, matching the timeline distribution for the lab's first technical artifact . The lab asserts it is strictly research-led and prioritizes internal scientific breakthroughs over public product releases David Silver | Training Data Summary - Teahose David Silver and Lasse Espeholt of Ineffable Intelligence. While Silver has aggressively projected having "extremely strong evidence" within a year and achieving superintelligence in three years Ineffable's David Silver bets on simulated worlds, not bigger ... David Silver and Lasse Espeholt of Ineffable Intelligence David Silver and Lasse Espeholt of Ineffable Intelligence, ambitious AI timelines frequently slip. The firm's $1.1B funding provides significant runway to operate quietly, similar to other pure-research labs. However, the need to eventually recruit more talent and raise further capital—coupled with a deep bench of former DeepMind researchers accustomed to publishing—tips the scales toward at least one headline release within the 26-month window.

Alternative domains and classification ambiguity

If a release does not easily map to games or simulated environments, "Mathematics or theoretical science" (13%) is the leading alternative. Given Silver's background with AlphaProof and the company's objective to rediscover science, a theorem-proving or scientific discovery result is a plausible early milestone Ineffable Intelligence David Silver and Lasse Espeholt of Ineffable Intelligence. A key uncertainty here is framing: if agents inside a simulated crucible achieve a mathematical breakthrough, the resulting publication could plausibly be headlined as a math result rather than a simulation breakthrough, shifting the classification.

Why other domains are unlikely

We assign very low probability to "Software engineering or computer-using agents" (3%) and "Robotics or the physical world" (2%). The founders have expressed open skepticism of LLM-based approaches reliant on human data. They have also explicitly stated that real-world deployment, including physical robotics, will be held back until their alignment science has been thoroughly proven within simulations David Silver | Training Data Summary - Teahose David Silver and Lasse Espeholt of Ineffable Intelligence David Silver and Lasse Espeholt of Ineffable Intelligence. A small residual probability (8%) is reserved for "Other domain," which captures potential work on weather modeling given cofounder Lasse Espeholt's background with MetNet, emergent economics, or reinforcement learning infrastructure.