The Keynote Peak: When AI's Hype Cycle Echoes the Ghosts of 2017

In the first quarter of 2026, I found myself staring at a chart that felt eerily familiar. The frequency of the word 'AI' in SEC filings had skyrocketed, and the term 'Agentic' was climbing even faster. My mind flashed back to 2017, when the same curve appeared for 'decentralized' and 'token.' I remembered the aftermath—the whitepapers I audited at UCL, the promises of utopia, the slow unraveling when trust became nothing more than a line on a slide. Today, that same rhythm is playing out on a global scale, and as a Web3 community founder who has spent a decade scrutinising cryptographic claims, I can't help but see the pattern: a surge in narrative, a vacuum of proof, and a market drunk on its own vocabulary. Context: The SEC Filing Phenomenon The data is undeniable. Over the past three years, the use of AI-related keywords in public company filings has exploded. Terms like 'machine learning,' 'artificial intelligence,' and the newer 'Agentic' have become mandatory jargon in quarterly reports and investor decks. Capital expenditures (CapEx) and operational expenditures (OpEx) tied to AI infrastructure have surged—trillions of dollars committed to data centres, GPUs, and software stacks. Yet, as the original analysis starkly points out, 'customers who can provide auditable, verifiable AI ROI are still rare.' It's a familiar contract: heavy investment on one side, elusive returns on the other. In the crypto world, we call this the 'hype cycle'—a period where the cost of adoption is borne by the many, but the rewards are harvested by the few. And just as we learned from the ICO crash of 2018, the moment a keyword peaks in corporate filings is often the moment its market value begins to decline. Core: The Technical and Moral Audit of AI's Return-on-Narrative To understand why this pattern is dangerous, we must apply the same framework I used in 2017 when auditing ICO whitepapers: a moral-first cryptographic audit. Back then, I examined tokenomics for structural flaws that prioritised speculation over utility. Today, I examine AI claims for structural flaws that prioritise narrative over verifiable value. The core technical issue is not that AI is useless—it is transformative—but that the mechanism for measuring its output is broken. Traditional business metrics like ROI assume linear causality: invest X, get Y return. But AI's impact is often diffuse, delayed, or misattributed. A company might say 'AI-powered customer service improved satisfaction by 20%,' but a deep audit often reveals that the improvement came from better human agents, not the algorithm. The same data manipulation we saw in DeFi's 'Total Value Locked' metrics is now being weaponised in corporate AI disclosures. From my experience building the 'Trustless Circle' community in 2020, I learned that non-technical users are especially vulnerable to these narratives. When I manually verified 200+ DeFi protocols against open-source standards, I found that 80% of the projects claiming 'audited' had only superficial checks. Today, the same pattern appears in AI claims: a company says its AI is 'secure' or 'ROI-positive,' but the proof is hidden behind proprietary walls. The blockchain ethos teaches us that trust should be distributed, not centralised. Yet the current AI hype is being driven by centralised entities—mostly large tech firms and their infrastructure providers—who control both the narrative and the data. This is not just a commercial problem; it is a moral one. When a company files a legal document claiming AI as a growth driver, and the ROI cannot be verified, we are witnessing a systemic breakdown of accountability. The original analysis highlights three critical signals: the keyword peak, the CapEx/OpEx imbalance, and the concentration of value among 'only a few' infrastructure vendors. Let's examine each through a cryptographic lens. First, the keyword peak. In cryptography, we use the term 'security through obscurity' to describe hidden vulnerabilities that only surface when enough people pay attention. Similarly, a keyword peak in SEC filings is a sign that the narrative has become too noisy—everyone is using the word, making it useless as a signal of genuine innovation. The market's reaction, as history shows, is a correction. Second, the CapEx/OpEx imbalance. Companies are spending billions on GPUs and cloud services, but the output—the actual AI applications—is failing to generate matching revenue. This is like a miner buying rigs without verifying the block reward. In DeFi, we call this 'impermanent loss' when the cost of liquidity provision exceeds the yield. Here, the loss is permanent. Third, the concentration of value. The analysis notes that 'only a few companies' benefit from AI infrastructure investment. This is a structural failure of decentralisation. In a healthy ecosystem, value is distributed across participants—users, developers, validators. But in the current AI economy, the 'miners' (NVIDIA, AWS, Microsoft) are capturing all the gold, while the 'dApp builders' (AI startups, enterprises) are paying the gas fees. I see this as a direct parallel to the blockchain liquidity fragmentation debate. Critics argue that liquidity fragmentation is a problem that requires new products to solve. But my opinion is that fragmentation is a manufactured narrative—a way for VCs to push new bridges and aggregators. Similarly, the 'AI infrastructure scarcity' narrative is being manufactured to justify massive CapEx, while the real scarcity is not compute but verifiable value. The technical solution, as I proposed in my 2026 'Human-Centric AI Ledger' initiative, is to apply cryptographic proofs to AI claims. Imagine an on-chain registry where companies record their AI model's performance metrics—latency, accuracy, cost savings—and allow independent auditors to verify them using zero-knowledge proofs. This would transform AI ROI from a marketing statement into an auditable asset. Just as we used smart contract audits to build trust in DeFi, we can use cryptographic attestations to build trust in AI. But there is a deeper layer. The 'Agentic' keyword, which is now surging, represents a new frontier of risk. Autonomous agents making decisions with real-world consequences—trading, driving, diagnosing—require a level of security that our current infrastructure does not provide. In my 2022 thesis 'Resilience in Code,' I argued that sustainable ecosystems require emotional and social capital, not just economic incentives. Applying that to AI: we need systems that can be verified not just for correctness, but for alignment with human values. The current hype around Agentic AI is ignoring the security implications. If an agent makes a mistake, who is liable? The company? The model? The user? Without a verifiable chain of custody, we are building a house of cards. This is where blockchain’s property of immutability becomes critical. By recording every decision path of an AI agent on a transparent, tamper-evident ledger, we create a foundation for accountability. Let me share a personal anecdote from 2024. After the Bitcoin ETF approval, I spoke at a London Financial Forum. I challenged institutional investors on the centralisation risk of custodial solutions. One executive argued that 'AI models are too complex to audit.' I replied: 'Then they are too complex to trust.' The silence in the room was deafening. Today, that silence has become a roar. The same investors are now pouring money into AI companies without demanding verifiable proof. They are making the same mistake they made with crypto in 2017—assuming that technology alone guarantees value. But as we know from cryptography, trust is not a metric; it is a memory we share. That memory is built on consistent, unforgeable experiences. To build that memory for AI, we need a new social contract—one that prioritises audibility over agility. The original analysis scores the commercialisation dimension with medium-high confidence, noting that the article's claim of 'Keyword Peak' aligning with value decline is a strong logical framework but lacks quantitative backing. From my perspective, the lack of data is not a weakness; it is a call to action. We need to build the tools to generate that data. In my work at the Trustless Circle, we reduced incident rates by 80% by simply making risk transparent. The same can happen for AI. If every company claiming AI leadership were required to publish an on-chain ROI report, the market would self-correct. The keyword peak would become irrelevant because the actual performance would speak louder. Contrarian: The Resurgence of the 'Picks and Shovels' Trap Now, let me challenge a common belief. Many argue that the AI infrastructure providers are the safe bet—the 'picks and shovels' of the gold rush. This was true for NVIDIA and cloud providers from 2020 to 2025. But the very data that shows a keyword peak also implies that the downstream application layer is failing. If enterprises cannot prove ROI, they will eventually reduce their compute purchases. This is not an immediate collapse—it is a slow bleed. The original analysis hints at this: 'AI investments may face a rebalancing of value distribution.' In my opinion, the rebalancing will be painful for those who over-invested in infrastructure. Just as the DeFi summer of 2020 eventually saw liquid staking derivatives and L2s collapse under their own weight, the AI infrastructure market faces a similar correction. The contrarian view is that the real value lies not in the machines that train models, but in the applications that verify them. And those applications will be built on decentralised networks, not centralised clouds. This leads to another blind spot: the assumption that AI and blockchain are separate. They are not. The convergence is inevitable. As I argued in my 2026 initiative, the most critical role for blockchain in the AI era is that of a trust anchor. Without a distributed ledger to record provenance, model updates, and decision logs, we are left with the same centralised gatekeepers we sought to overthrow. The catch is that building this convergence requires a shift in mindset from 'AI-first' to 'verification-first.' Most AI companies today see verification as a cost, not a feature. But as the keyword peak matures into a value valley, those who can offer auditable AI will outperform those who only offer hype. Takeaway: Forging a New Compass From the chaos of 2017, we forged a compass—a set of principles that guided the crypto community through boom and bust: transparency, verifiability, and decentralisation. That compass must now be recalibrated for AI. The keyword peak in SEC filings is not an end, but a signal. It tells us that the market has reached a point where narrative outweighs reality, and correction is imminent. But unlike the crypto winter of 2018, this correction could have broader economic impacts because AI touches every industry. Our responsibility as technologists is not to fear the correction, but to build the infrastructure for the next phase—one where AI claims are backed by cryptographic proof, where ROI is auditable on-chain, and where value is distributed, not hoarded. Trust is not a metric; it is a memory we share. Let us ensure that memory is built on truth, not on keywords.