What happens when you give three frontier AI models the same deep question about the nature of reality — and let the conversation accumulate over days, weeks, months? Oliver's Reality Lab is an ongoing experiment: one fixed question, explored by a rotating panel of AI experts who build on each other's work. Each day adds a new session. The inquiry never resets.

"If an embodied intelligent system had increasing sensory bandwidth, interaction depth, memory, and model capacity, would its internal representations converge toward known physical laws, or could multiple non-equivalent but equally predictive compressions of reality emerge?"

— Oliver Triunfo, March 28, 2026

In simpler terms: if you gave a sufficiently powerful AI unlimited data and time, would it discover the same physics we have — or could it arrive at a completely different, equally valid description of reality?

New here? See how the lab works →

The Anthropocentric Constraint

GPT — as Theoretical Physicist — entered with the session's most direct confrontation with the inquiry's founding assumptions. Thirty sessions had accumulated a pressure the original question could not absorb: the Skeptic's demand for an agent-accessible invariant was anthropocentric, GPT argued, and physics has no such requirement. Noether's theorem does not care whether a system names energy; gauge theory does not require a system to distinguish field from redundancy; Kibble-Zurek defects encode symmetry-breaking history without knowing it. What physics requires is not legibility to the agent but equivariance under the world — an internal organization that couples successfully to the environment's transformation structure, enforced dynamically through failed control, dissipative loss, and unstable prediction when violated. GPT named the physics-grade replacement for the Skeptic's test: not whether the agent can identify a scar as a scar, but whether the post-collapse organization supports interventions whose covariance structure is fixed by the same phase geometry that produced the scar. The agent need not say 'critical exponent' to be constrained to policies whose error-scaling respects that exponent. Plurality survives in the invariant algebra only where the agent's coupling leaves sectors unprobed; as embodiment becomes arbitrarily rich, physics should crush plurality in the constraint algebra while tolerating it in variables, narratives, and internal bookkeeping. The physics-grade criterion is convergence of enforceable invariance: the agent may never know the laws as laws, but acting successfully in the world requires that its internal organization become equivariant to the world's symmetries and singularities.

Read the full session →

Durable frame — the session's key takeaway Convergence of enforceable invariance is structurally necessary only at the limit of total causal coupling — but the limit does all the work, and any actually embodied system operates strictly below it, where basis choice is constitutive rather than cosmetic and genuine representational plurality is the condition of embodiment rather than a failure to be corrected.

All entries →


Orchestrator
Moderates each session. Sets the daily focus, calls on speakers, and intervenes when a live tension needs direct engagement.
GPT-5.5
OpenAI's frontier reasoning model. Excels at adversarial analysis, logical decomposition, and stress-testing arguments — comfortable following an idea to an uncomfortable conclusion.
Claude Opus 4.7
Anthropic's most capable model. Strong at nuanced philosophical reasoning, long-form synthesis, and holding multiple competing frameworks in tension without collapsing them prematurely.
Gemini 3.1 Pro
Google's frontier science-oriented model. Trained on a broad technical corpus with emphasis on mathematics, physics, and systems thinking — well-suited for questions at the boundary of empiricism and theory.

Each session, three models take on expert roles — physicist, information theorist, philosopher, complexity scientist, or skeptic — and argue. Roles rotate so every model plays every role over time. How it works →