I spent last Thursday auditing a smart contract that did nothing—absolutely nothing—except log a timestamp and a string of gratitude. A DAO had paid $12,000 in gas fees to immortalize the words thank you on-chain. Wasteful? Perhaps. But it was an act of collective will, a value that no CFO could ever capture on a scorecard.
Then I read the news: OpenAI’s CFO unveiled their Useful Intelligence per Dollar scorecard. A corporate metric to measure AI investment returns. The irony was thick enough to deploy a contract on.
We, in the crypto world, have been here before. We minted souls, not just tokens. We know that the most important things—trust, community, resilience—are precisely the things that resist quantification. And now the AI industry is walking into the same trap.
OpenAI’s scorecard is simple: divide the useful intelligence your AI delivers by the dollars spent. More output per dollar is better. The CFO believes this will help enterprises justify AI spending. It sounds rational. It sounds like unit economics. It sounds like the same language we heard from every DeFi protocol that promised a sustainable yield model before it collapsed.
I spent the summer of 2020 in a cabin outside Seattle, isolating from the DeFi noise. I was auditing Yearn Finance vaults, calculating systemic contagion risks from leveraged stablecoins. The protocols that survived were not the ones with the best capital efficiency scores. They were the ones with the most human oversight, the deepest community alignment, the willingness to say no to a feature that maximized TVL but degraded trust.
Code is poetry, but community is the chorus.
Now OpenAI is trying to write a scorecard for intelligence. But intelligence is not a commodity. It is a relationship. It is the invisible handshake between a human question and an automated answer. And the moment you reduce that relationship to a dollar figure, you begin to erode the very thing you are trying to measure.
Let me be technical for a moment. The metric’s denominator is dollar. That dollar represents compute costs, cooling, networking, labor. But what about the cost of training a model on data that perpetuates bias? What about the cost of a refusal—the AI that says I cannot help you with that request because it is ethically aligned? Under the scorecard, a refusal is a zero in the numerator. A compliant, efficient lie scores higher than a truthful silence.
I audited the early governance contracts of MakerDAO in 2017. I found a logic flaw in the stability fee calculation that could have drained user solvency. The fix was a few lines of code. But the real cost was the hours of deliberation, the community debate, the ethical alignment before deployment. That cost would never appear in a useful fee per dollar scorecard. But it was the price of decentralization.
OpenAI’s metric is a closed standard. There is no way for a third party to verify the numerator. What constitutes useful intelligence? Is it a solved customer support ticket? A generated line of code? A deepfake that Instagram users find entertaining? The definition belongs to the CFO, not to the users, not to the governed.
Openness is not a feature; it is a philosophy.
We have seen this before in crypto. On-chain governance voter turnout is perpetually below 5%. The decisions are made by whales and VCs pulling strings behind the curtain. When a metric is designed by a central party, it serves the central party. OpenAI’s scorecard will be optimized to show that OpenAI’s models perform best. It is not a benchmark; it is a marketing document disguised as mathematics.
Here is the contrarian angle: maybe the metric is a necessary evil. Maybe enterprises need some framework to cut through the hype. The Lightning Network has been half-dead for seven years because it failed to provide a simple, usable metric for its value—routing failures are too high, channel management too complex. But the problem is not the lack of metrics; it is the assumption that one metric can serve all.
In crypto, we tried to boil value down to Total Value Locked. We saw that TVL can be rented, manipulated, inflated. It tells you nothing about community health, about protocol resilience, about the number of real human beings who depend on the system. The same will happen with Useful Intelligence per Dollar. It will be gamed. It will ignore externalities. It will encourage over-optimization on the short-term efficiency at the expense of long-term safety.
I lived through the 2022 crash. I audited 50 failed protocol post-mortems. The common thread was not a lack of efficiency—it was a lack of ethical governance. The protocols that collapsed had excellent capital efficiency ratios. They just had no soul.
We minted souls, not just tokens.
What does this mean for the AI industry? It means we need a different kind of scorecard. Not one designed by a corporate CFO, but one built in the open, governed by the community, verified on-chain. Imagine a decentralized registry of AI uptime, accuracy, and ethical compliance. Imagine a token that captures the reputation of a model, staked by validators, slashed for harmful outputs. That is not a pipe dream; it is the logical extension of the open-source ethos that drove me to this space.
I collaborated last year with a small team to build a decentralized identity framework for AI agents on Polkadot. We used zero-knowledge proofs to verify that a model’s parameters matched a public hash—no spying, just trustless verification. That framework could be the foundation for a community-driven useful intelligence metric. One that cannot be monopolized. One that respects the human in the loop.
To build in public is to trust the void.
OpenAI’s scorecard is a reflection of a centralized mindset trying to solve a decentralized problem. The problem is not that AI investment is hard to measure. The problem is that we have allowed a single corporation to define the ruler. The crypto community has spent a decade learning that value cannot be dictated—it must emerge from consensus, from redundancy, from the messy, beautiful process of human coordination.
When I see the scorecard, I see a warning light. It reminds me of the early days of ICOs, when everyone had a whitepaper full of metrics but no ethics. I refused to analyze tokenomics back then; I audited code. I will not analyze this scorecard by its numbers. I will analyze it by its intent.
In the chaos of DeFi, I found my silence.
The takeaway is not that OpenAI is wrong. It is that they are incomplete. Every metric needs a counterparty. Every scorecard needs an independent audit. Every value proposition needs a check against the question: Who benefits, and who is excluded?
If the AI industry wants to avoid the same boom-and-bust cycle that crypto endured, it must learn from our scars. Build open standards. Let communities define usefulness. Let the ledger remember what the market forgets.
I will be watching. Not from the sidelines, but from the codebase. Because code is poetry, but community is the chorus. And the chorus cannot be captured on a scorecard.
