rationale:

Current State of AI in Physics

As of late October 2026, no experiment has tested a physics prediction made by a general-purpose AI system AI Agent Swarms as Researchers: Progress, Challenges ... October 8, 2026 - Nathan Benaich AIR STREET CAPITAL.. While AI systems have demonstrated profound capabilities in theoretical physics—such as the claimed construction of 4D Yang-Mills theory Working with academic communities to build scientific ... and high-loop super Yang-Mills amplitudes AI makes its first meaningful breakthrough in theoretical ...—these are purely theoretical and mathematical milestones. Recent achievements by LLM agents, like calibrating superconducting qubits https://arxiv.org/abs/2606.22376 or exploring melting hierarchies in real quantum materials AIMS: an AI experimentalist turns uncertainty into quantum ... AIMS: an AI experimentalist turns uncertainty into quantum ..., represent impressive laboratory integration and the explanation of known phenomena. However, they do not cross the threshold of predicting a quantitatively new, unmeasured physical effect.

The Required Sequence and Bottlenecks

For this milestone to be reached, a sequence of events must occur publicly: a general-purpose AI must generate and publicly post a quantitative prediction of a new phase, particle, or effect, and then an independent experimental group must execute and publish a confirming measurement reaching 3σ or broad field acceptance. Currently, open-ended hypothesis generation benchmarks (CritPt) have been flatlining since July 2026, suggesting that AI systems remain much better at mathematical derivation than selecting physically fruitful, novel hypotheses. Furthermore, the criteria explicitly exclude discoveries of new materials or compounds AI Accelerates Scientific Discovery | by Kosi Gramatikoff, and any academic attribution will be strictly scrutinized to ensure the AI genuinely supplied the "central idea" rather than just acting as an execution tool.

Fastest Paths to Resolution

The earliest breakthroughs will likely bypass traditional particle accelerators or complex condensed-matter fabrication. Instead, programmable quantum simulators, photonics, and the reanalysis of archived astronomical or collision data offer the tightest turnaround times. Historically, the lag from prediction to experiment in condensed matter has ranged from about a year (for the HgTe quantum spin Hall effect Quantum Spin Hall Effect and Topological Phase ... Quantum spin hall insulator state in HgTe quantum wells) to seven years (for twisted-bilayer flat bands A review of experimental advances in twisted graphene moiré ...). With the rapid integration of public AI tools into physics workflows—evidenced by a five-fold increase in math preprints crediting AI between June and September 2026—and the availability of cloud-based quantum hardware, the time from a solid prediction to an independent measurement could realistically fall to 6–18 months.

Conclusion and Timing

My median date for the first qualifying discovery is October 1, 2028. This allows roughly a year for a clear, public, AI-credited prediction to be formulated, followed by another 6 to 18 months for an independent laboratory to execute the experiment and report the findings. The expectation that the first confirmed physical predictions will drive a modest accumulation of such successes by the end of the decade is consistent with this timing, rather than requiring a rapid wave of discoveries . The tight early percentiles—June 15, 2027 at 10% and January 15, 2028 at 25%—account for the possibility that an unnoticed, qualifying prediction is already sitting in recent preprints and simply awaiting a rapid quantum-device test. The extended right tail, with 75% probability by June 1, 2030 and 90% by September 1, 2032, reflects the friction of academic attribution norms, the slow speed of experimental physics, and the risk that early candidates might be disqualified as mere material discoveries or theoretical proofs.