When a corporation claims it can measure the value of intelligence per dollar, we must ask: who defines 'useful'? And at what cost to the unfathomable depth of the human mind? This month, OpenAI CFO Sarah Friar unveiled a ‘useful intelligence per dollar’ scorecard, a metric designed to quantify the return on AI investment for large enterprises. It sounds rational, almost inevitable—a CFO’s job is to justify spend. But to those of us who have spent years in the trenches of decentralized governance, this is not a simple efficiency tool. It is a declaration of war on the very concept of sovereign value.
I have been here before. In 2017, auditing the Parity Wallet library, I saw how a single line of flawed code could drain millions—not because the technology was malicious, but because the governance around it was blind. Later, during the MakerDAO debates of 2020, I helped craft a whitepaper arguing that stablecoins should serve as public goods, not profit centers. We fought over what ‘useful’ meant for a decentralized currency. Now, OpenAI is attempting to define ‘useful intelligence’ from a single boardroom. This is the same battle, fought on a different front.
Context: The scorecard is a cost-efficiency ratio. Numerator: ‘useful intelligence’—a vague term OpenAI has not yet operationalized. Denominator: dollars spent on compute, energy, infrastructure. The goal is to help enterprise clients build a business case for AI, to move conversations from ‘what can AI do?’ to ‘what can AI do for my bottom line?’ On the surface, it’s a mature move by a company facing pressure to show ROI after $13+ billion in funding. But as a practitioner who believes decentralization is a practice of radical empathy, I see deeper signals.
Tracing the code back to the conscience, I find that this metric is a Trojan horse for centralized control. By defining ‘useful intelligence’ unilaterally, OpenAI places itself as the arbiter of value in an entire industry. In crypto, we learned the hard way that value is emergent from community consensus, not decreed by a single entity. The MakerDAO governance proposal I pushed through required 15 rational actors to agree on what collateral was ‘useful’ for stability. It was messy, slow, and beautiful. OpenAI’s scorecard is the opposite: fast, clean, and ultimately autocratic.
Core analysis: Let me dissect the implications for our domain—the intersection of blockchain and AI. First, the metric incentivizes cost-cutting over safety. If ‘useful intelligence per dollar’ becomes mainstream, any safety feature that adds latency or compute cost (alignment tax, red-teaming, edge-case handling) will be viewed as inefficiency. We saw this dynamic in DeFi: protocols that optimized for TVL growth often neglected security audits until after a hack. The same pattern will repeat in AI. Second, it creates a hierarchy of intelligence. High-cost, high-capability models for drug discovery will be in one bucket; low-cost, high-efficiency models for customer service in another. This stratification mirrors the digital divide we fight against in Web3. Third, it accelerates the centralization of AI infrastructure. To achieve the best ‘useful intelligence per dollar’, you need massive scale, custom hardware, and access to cheap energy—exactly what OpenAI and its backer Microsoft control. Small, community-run AI projects cannot compete.
From my 2022 retreat in Hanoi, after the FTX crash, I wrote the ‘Ho Chi Minh Trust Manifesto’, arguing that true resilience comes from community verification, not algorithmic guarantees. The same applies here. The scorecard will be used by enterprise buyers to justify locking into OpenAI’s ecosystem, creating a moat that is harder to cross than any technical benchmark. Governance is not a vote; it is a vigil. We must watch how this metric is applied—not just as a pricing tool, but as a means of excluding alternative value systems.
Contrarian view: Some in the crypto space might welcome this as a ‘reality check’ for AI hype, a move that forces startups to focus on real-world use cases. I have sympathy for that. During the 2020 DeFi summer, I saw countless projects promise ‘algorithmic banking’ that crumbled under scrutiny. A clear value metric can help separate signal from noise. But the problem is not the existence of a metric; it is who owns it. Open-source models like Llama or Mistral, with zero licensing cost, could theoretically achieve stellar ‘useful intelligence per dollar’. But without a brand and enterprise salesforce, they remain invisible to corporate procurement. The scorecard, as defined, serves the institution, not the innovator.
Furthermore, there is a spiritual blind spot. Intelligence is not a fungible commodity. The AI that writes your code and the AI that holds your hand through grief are not comparable on a per-dollar basis. We build bridges from the ashes of belief—the belief that technology can augment human dignity. But a cost-efficiency ratio treats all intelligence as the same, stripping it of context, ethics, and soul. In my work on a Human-First Proof of Personhood protocol in 2026, we insisted that identity should be self-sovereign and privacy-preserving. We rejected the notion that a centralized authority could measure the value of a person. Similarly, we must reject the notion that one corporation can measure the value of intelligence.
Takeaway: The blockchain community must respond with its own value framework for AI—one that measures not just cost efficiency, but resilience, decentralization, and alignment with human sovereignty. We can draw from the Ethereum-based ‘Proof of Stake’ metrics that weigh security against energy, or from the radical transparency of on-chain data. The question is not whether to measure, but who gets to define the ruler. The protocol must serve the human spirit, not the quarterly report. I urge every builder in crypto to study this scorecard, not as a competitor, but as a mirror. It shows us what centralization looks like when it tries to quantify the infinite. Our job is to offer an alternative: a practice of radical empathy where value is not commanded, but co-created.