Dr. Alex Wissner-Gross

Dr. Alex Wissner-Gross treats frontier artificial intelligence as a priced commodity and a traded risk, writing about how cognition, compute, and superintelligence are being pulled into the language and infrastructure of financial markets. As curator of The Innermost Loop, he covers the fast edge of AI through the lens of prices, indices, energy, and regulation rather than product launches alone, asking what each new model or policy means for the cost of intelligence and the structure of the AI economy. A physicist, computer scientist, entrepreneur, and investor, he brings a technical background in AI together with a markets-focused perspective, and his finance coverage often follows the moment when abstract ideas like “tokens” and “throughput” become benchmarks, contracts, and traded exposures.

High-velocity intelligence from the event horizon

The Innermost Loop is built as a rapid-fire briefing rather than a traditional column, with most issues framed as “Welcome to [date]” dispatches that open with a sharp thesis about the current state of the Singularity and then move through tightly curated items on models, markets, and policy. Its tagline describes the project as “high-velocity intelligence from the event horizon,” and that speed and framing are visible in how he writes about frontier models outrunning their own launch cycles or the Singularity moving so quickly that conventional timelines break down. Each edition typically blends concrete news—such as specific model releases, energy contracts, or regulatory decisions—with a one- or two-sentence synthesis that treats them as signals about where intelligence, and the ability to pay for it, is going next. The running structure gives communications teams a clear sense of cadence: he returns repeatedly to the same touchpoints—capabilities, prices, and constraints—so stories that illuminate one of those levers have the cleanest fit.

The financial system is attempting to price in superintelligence

Wissner-Gross is unusually explicit about the way financial markets digest AI progress, writing that “the financial system is attempting to price in superintelligence” and then tracking how that shows up in instruments, indices, and capital allocation. In his coverage of OpenAI’s GPT-5.x series, he not only lists capabilities like native computer use, million-token context, and GDPval scores, but also connects them directly to economic benchmarks, noting when a model “matches or exceeds human professionals at knowledge work 83% of the time” and what that implies for the valuation of human versus synthetic labor. When he writes about GPT-5.2 beating Pokémon Emerald after spending 71% of its time “thinking,” the punchline is not the stunt but the way such milestones change expectations for automated problem-solving and thus how investors and risk managers discount future cash flows. Across these pieces, his finance beat is less about stock-picking and more about regime change, asking how quickly full software engineering automation could arrive—he cites estimates as early as 2027–2028—and how that timeline feeds back into everything from wage structures to the cost of capital.

Once cognition is just throughput, it gets priced like crude oil

One of his defining themes is that once AI cognition becomes scalable throughput, it gets priced per unit like any other commodity, “by the token,” with direct analogies to crude oil and power markets. In “The First Frontier AI Token Price Index,” he introduces the Ornn Token Price Indices (OTPI) as a benchmark that prices frontier-lab tokens based on actual transaction volume rather than posted rate cards, positioning it as the first attempt to give the AI economy a transparent reference price for output tokens. He explains how OTPI aggregates paid inference across models into a single daily figure in dollars per million tokens, paralleling how commodity indices roll diverse physical grades into a tradable benchmark. That work sits alongside coverage of GPU rental rate indices and the first compute futures market, where CME Group and a data provider plan contracts tied to daily GPU rental benchmarks so traders and AI builders can hedge compute price volatility. Together, these pieces show him treating GPU time as “a hedgeable commodity” and AI tokens as the output leg of a new spread between input compute and priced cognition, a level of microstructure detail that goes well beyond generic tech or finance reporting.

The Singularity has reached the point where government clearance is scarce

Regulation and geopolitics enter his finance coverage as hard constraints on the supply of usable intelligence, not just as policy color. He writes that “the Singularity has reached the point where the scarcest input to frontier intelligence is no longer compute but government clearance,” and then walks through episodes such as the US government lifting its block on Anthropic’s Claude Mythos 5 and redeploying a strong cybersecurity model to around 100 trusted companies and agencies defending critical infrastructure. In another edition he describes how the question of who writes the values baked into frontier AI is becoming “a geopolitical fault line,” highlighting Anthropic’s public refusal to let a government client dictate the system’s values and treating that stance as a live risk factor for deployment and investment. He also tracks how national commemorations and institutions respond, noting, for example, a Museum of American Finance using an interactive AI recreation of Alexander Hamilton, which he frames as part of a broader shift in how financial history and identity are mediated by AI systems. Across these stories he returns to the idea that access, licensing, and value alignment decisions are starting to ration effective intelligence just as much as data centers and GPUs do, a perspective that matters for anyone pitching stories on AI regulation, security-clearance issues, or cross-border AI policy.

Energy, infrastructure, and the cost of intelligence

Another recurring strand in his work is the physical infrastructure that underpins the AI boom, especially long-term energy contracts and their effect on the marginal cost of cognition. He calls out deals like Ormat’s 150‑MW geothermal power purchase agreement with NV Energy to supply Google’s Nevada data centers through 2030, treating it as an example of how hyperscalers secure low-carbon baseload power to de-risk AI expansion. He follows Microsoft’s aggressive push to “tokenmaxx” products like Copilot Cowork with cheaper token pricing, linking commercial moves directly to underlying compute and energy costs and to the pricing benchmarks he covers elsewhere. When he writes that frontier models are outrun before their launch weeks are out, or that the Singularity is moving faster than traditional infrastructure cycles, the implicit question is whether there is enough power, hardware, and capital to keep up—and how those constraints will show up in AI prices per token, per task, and per synthetic worker. This blend of infrastructure detail and pricing logic marks his finance beat as one that starts from physical capacity and ends at per-token economics.

Outside The Innermost Loop, Wissner-Gross presents himself as a physicist, computer scientist, entrepreneur, and investor working at the intersection of artificial intelligence and other complex systems, and his newsletter has accumulated a substantial subscriber base. That background shows in his preference for benchmarks, indices, and quantitative measures—GDPval scores, token prices, GPU rental rates—over anecdote, and in his habit of treating every new model or policy as a shift in an evolving market structure for intelligence.