Adam Benenson

Ambient AI layer for the browser that resurfaces relevant information before you know you need it. The problem is structural: knowledge workers lose hours daily not to hard problems but to context reconstruction — finding the thing you already found, rebuilding the thread you already held. Most AI tools respond to prompts; Forethink anticipates context, proactively surfacing connections across your entire digital history. All processing happens on-device, with no data extraction or centralized indexing. Privacy isn't a policy — it's the substrate.

A recursively self-building entity — a factory for autonomous agent swarms that continuously rewrites its own dispatch logic, capability map, and authority boundaries as evidence accumulates. It maintains a queryable self-model: not just a record of what it has done, but a structured representation of why decisions produced the outcomes they did — enabling causal attribution and counterfactual reasoning. Each autonomy expansion is backed by a promotion certificate: a proof object recording the evidence trail, admissibility checks, and rollback witness that authorized it. Candidates move through a typed validation pipeline before earning authority; clawback on degradation is mandatory and automatic. The governance layer isn't a constraint on growth — it is the growth mechanism. A system that can only expand by proving it should.

A personal digital twin — built to reconstruct and extend my own causal reasoning, belief structure, and intellectual trajectory. Ingests arbitrary data: conversations, documents, reading history, behavioral traces, structured profiles. Extracts and consolidates claims across ten psychological dimensions; maintains an 18,000-episode store, 12,000-entry profile database, and queryable knowledge graph. Context engineering is formalized as a Markov Decision Process (DT-MDP) with a contrastive IRL retrieval policy. The system runs continuously, improving its own models between sessions.

A recursively self-learning alpha compounding engine for prediction markets. Hypotheses enter as replay-scored candidates, survive shadow and paper validation, earn size through bounded live review, and get clawed back when evidence degrades. Every fill updates a market factor graph; the graph reshapes candidate selection; better candidates generate more evidence. The system autonomously ratchets size as edge compounds — no operator in the loop.

A formal theory of the relationship between information and energy, and how organized structure emerges from their interaction. Starting from a single unavoidable constraint — absolute indifference is incoherent, because any representation of nothingness introduces a distinction — SFT derives first a variational principle governing how information organizes on the induced geometry, then a three-component tensor decomposition: Integration, Complexity, Uncertainty. Developed in Lean 4 on Mathlib; key theorems sorry-free.

The finite Čencov–Petz uniqueness theorem: any continuous monotone metric family on the probability simplex is a scalar multiple of the Fisher information metric. Proved via splitting invariance at rational points + density argument.

The spectral theorem for compact self-adjoint operators on Hilbert spaces: a Hilbert basis of eigenvectors exists.

Rellich–Kondrachov compact embedding: H¹ → L² is compact on compact Riemannian manifolds. Euclidean heart via Fréchet–Kolmogorov; manifold glue via finite chart decomposition.

Led the product, engineering, data, and design departments at an enterprise analytics company serving hedge funds, sovereign wealth funds, pensions, and family offices — 140+ institutional investors managing $3T+ in assets. Full org ownership from strategy through execution; member of the executive team.

Analyst on a $2B+ systematic equity book (Sharpe above 2) through the 2008–2009 financial crisis; promoted to PM and managed own $200M+ portfolio in parallel.