The Chairman Who Didn't Sell: A Decentralist Reads Oracle's Disclosure Gap

Two sentences arrived in my feed this week, filed under blockchain, worth exactly what they cost to publish. The chairman of Oracle Corporation has cancelled a planned sale of company stock. No replacement schedule was announced. That was the entire dispatch — no source, no date, no figure, no quotation — and it is worth more careful attention than most of the thousand-word threads I scrolled past to reach it, because its poverty is the point. A cancelled transaction, like a silence in a conversation, tells you not what happened but what someone would prefer you not to ask. Ever since 2017, when I translated Vitalik Buterin's Ethereum whitepaper into Portuguese and appended eighty pages of ethical commentary, I have measured every piece of financial news by a single question: does it let me verify, or does it ask me to believe? The Oracle dispatch asks me to believe. That is why it belongs in a blockchain feed, and that is why it does not belong there at all. To see why a routine stock sale by the founder of a database company deserves the scrutiny an evangelist normally reserves for a smart contract, you have to understand the machinery humming beneath the headline. In the United States, directors and officers of public companies who wish to trade their own shares operate inside a lattice of rules built to prevent them from trading on information the rest of us do not have. The load-bearing beam of that lattice is Rule 10b5-1, a provision of the Securities Exchange Act that allows an insider to establish a pre-arranged trading plan — in effect, a written contract specifying when, how many, and at what price shares will be sold — and, if the plan was adopted in good faith while the insider possessed no material non-public information, grants that insider an affirmative defense against insider-trading liability. The plan is a promise made in advance, designed to bind the future self to a decision the present self cannot game. The SEC rewrote the rule in December 2022, adding mandatory cooling-off periods — typically ninety days for officers and directors — before a plan may begin execution, and requiring companies to disclose the adoption, modification, and termination of such plans under Item 408 of Regulation S-K. Proposed sales are reported on Form 144; executed ones on Form 4, usually within two business days. The philosophy behind this is worth stating plainly, because it is a philosophy our industry claims to share. A market is a machine for aggregating beliefs into prices, and it works only if the beliefs are honest. When one participant knows something the others do not and trades anyway, price ceases to be a measurement and becomes a trap. Rule 10b5-1 was invented not to permit insider trading but to permit insiders to trade without the specter of suspicion — to separate the act of selling from the suspicion of cheating. It is, in other words, an attempt to build trust into a procedure, which is precisely the ambition of every consensus mechanism ever written. Nor is the rule as clean as its defenders pretend. The SEC's own 2021 study found that executives who traded under 10b5-1 plans captured significantly higher returns than those who did not — evidence that the mechanism was being used less as a firewall against suspicion than as a permit to profit. The 2022 revision, with its longer cooling-off periods and new certification requirements, was a response to that evidence: an admission that a rule designed to build trust had instead been built to launder it. This is the pattern I recognize from our own industry — mechanisms invented in good faith, exploited in bad, then patched after the fact, always one step behind the people who understand them best. Larry Ellison co-founded Oracle in 1977 and has never really let go of it. His stake in the company he built has at times approached forty percent of the equity, a concentration of ownership that grants him voting power almost no founder in the dual-class era can match. A plan by a figure like that to sell shares is never a private decision. It is a signal, and markets read signals the way sailors read a barometer — not for what they say but for what they foretell. When such a signal reverses, the reversal itself becomes the information. Consider the word at the center of the dispatch. "Cancelled." It arrives as a single settled fact, but it is a container for at least three mutually exclusive realities, each pointing the reader in an opposite direction. The first is the one the headline quietly encourages: a founder of incalculable personal wealth, staring at a company he believes the market undervalues, decided he would rather hold than sell. This is the confidence reading, and it is flattering to the company, to the market, and to the reader. It is also the least verifiable of the three. The second is more mundane and far more honest about how 10b5-1 plans actually behave. These plans routinely embed limit prices — thresholds beneath which shares simply will not be sold. If Oracle's stock slipped under that threshold, the plan would not execute, and the resulting non-event could be described, with perfect technical accuracy and complete narrative distortion, as a "cancellation." Under this reading, the news is not a show of confidence but a symptom of weakness: the shares fell to the level at which the seller's own pre-committed rule told him to stop selling. The third involves the trading windows and blackout periods that govern insider activity around earnings, financings, and mergers. If Oracle were approaching a material event, a compliance officer would suspend the plan, and again the outcome would be reported as a cancellation. This reading is neutral on sentiment but dense with timing: it hints that something is coming. In practice, a 10b5-1 plan is a set of instructions handed to a broker and executed without the insider's further involvement, which is precisely what gives it its legal force. The insider cannot intervene to sell on good news or to halt a sale on bad news, because the entire point is that the decision was made before the news existed. This is why a genuine cancellation is unusual: it requires an affirmative act, typically a formal instruction to terminate the plan, which itself triggers disclosure. A plan that simply fails to execute because its conditions were unmet produces no such act and no such disclosure. The word "cancelled," in other words, implies a decision that a technical non-execution does not, and the dispatch's use of that word is itself a claim the reader cannot audit. I have spent enough hours inside governance documents to know that the distance between these three readings is the distance between a buy signal, a sell signal, and no signal at all. And the dispatch, as published, contains not one of the details — not the size of the original plan, not its adoption date, not its limit price, not the reason for its termination — that would let a reader tell them apart. What it offers instead is a feeling, and the feeling has been manufactured by omission. This is where my own work presses against the frame. In 2020, during the first DeFi summer, I spent six hundred hours manually auditing the earliest scripts of what became Aave V2 and found three critical logic errors in its interest-rate models. I published a fifteen-thousand-word manifesto on GitHub, "Trustless but Not Careless," arguing that an audit which ignores the social contract surrounding the code is not an audit at all. The governance team adopted the findings, and a potential four-million-dollar exploit was closed. What I carried away was not the lesson that code is trustworthy. It was that trust is a system, and a system is only as honest as its worst-documented assumption. Apply that lens to the Oracle dispatch and the shine comes off the disclosure regime. On paper, that regime is a triumph: a mandated, standardized, machine-readable stream of insider activity, public within two days, free to anyone with a browser. Compare that, the evangelists say, with the opacity of a private ledger or a closed boardroom. But look at what the transparency actually delivers. It delivers the fact of a cancellation without the reason for it, the event without the mechanism. It is transparency of the artifact and opacity of the intent. This is the failure mode I have come to see as the central problem of both traditional finance and much of our own industry: we build systems that disclose what is easy to disclose, conceal what is consequential, and then congratulate ourselves on the disclosure. Here the comparison with on-chain governance becomes instructive, and not in the direction our maximalists prefer. A decentralized autonomous organization, in its idealized form, records every vote, every treasury movement, every parameter change on a public ledger, immutably and instantly. There is no two-business-day lag, no Form 4 filed by a compliance department, no gap between the act and the record. By the raw metric of transparency, a DAO beats the SEC's regime by an order of magnitude. And yet I have argued for years that most DAOs exist in a legal no-man's-land, and that this is not a rounding error but a foundational flaw. When a DAO's decisions cause harm — when a governance vote drains a treasury through a malicious proposal, when a contributor is injured by an automated policy no human reviewed — there is frequently no legal person to hold accountable. Members who voted may face unlimited personal liability under partnership doctrine in some jurisdictions, or none at all in others, and the ambiguity is itself the hazard. Radical transparency of process coexists with radical uncertainty of responsibility. The ledger shows everything; the law recognizes no one. The theoretical case is easy to make. The practical one has been made for us. In 2016, the collapse of The DAO — drained of roughly sixty million dollars through a reentrancy exploit — produced not a clean legal reckoning but a philosophical crisis that split Ethereum into two chains and left the question of responsibility permanently unsettled. In 2022, the Commodity Futures Trading Commission sued Ooki DAO, a successor to bZx, and won a default judgment against an entity that possessed no office, no address, and no officers — pursuing, in effect, a collective. The lesson was not that DAOs are illegal. It was that a structure without a legal person can still be held liable, and when it is, the liability falls on whoever the courts can reach. Transparency of process had done nothing to prevent the harm and everything to document it. So we have two systems, each transparent in the dimension the other neglects. The SEC regime documents the who and the when but buries the why beneath boilerplate. The DAO regime documents the what with perfect fidelity but dissolves the question of who answers for it. The Oracle dispatch sits where both failures meet: a corporate disclosure as thin as a governance proposal's rationale, asking the reader to supply the trust neither system has earned. Let me be precise, because precision is the only currency I have left after years of watching good technology sold on bad faith. The Oracle news is not a scandal. Nothing in it implies wrongdoing. Ellison may have cancelled his plan out of pure conviction that the company he built is worth more than the market pays — entirely plausible, especially as Oracle has reoriented its story around artificial-intelligence infrastructure in this exuberant market. My point is subtler and more damning: the dispatch gives you no way to know, and the absence of a way to know is itself the buried story. Consider what a genuinely informative disclosure would contain. The size of the original plan in shares and dollars. The adoption date and remaining duration. Whether a limit price existed, and where it sat relative to the current price. Whether the cancellation coincided with a trading window, a blackout, or an announced transaction. Whether the insider retained other plans, or pledged, donated, or transferred shares through other vehicles. Every one of these facts is already known to the company. Every one is material to a reasonable investor. None of them appeared. Here is the part that should trouble anyone who believes markets price information efficiently. If the dispatch is accurate and the cancellation was affirmative, then the company knows the reason and the reader does not — and the company decides whether to share it. If the dispatch is imprecise and the cancellation was technical, then the word itself is a distortion, and the reader has been misled by vocabulary. In neither case is the reader served. The disclosure regime has produced a fact that functions as a Rorschach test: it tells you more about the reader's prior beliefs than about the company's condition. This is not an argument for more regulation. It is an argument for a different kind of record — one that captures intent and context rather than merely tallying transactions. And here the decentralized toolkit, used with discipline rather than enthusiasm, has something real to offer. A zero-knowledge attestation could let an insider prove, without exposing the underlying terms, that a cancellation was voluntary rather than compliance-driven — a cryptographic commitment that dissolves the ambiguity while guarding privacy. A verifiable credential could bind a governance decision to a responsible party without revealing personal data. These are not fantasies. They are the primitives I helped build in 2024 through the Verifiable Humanity initiative, where we partnered with five AI companies and negotiated a five-hundred-thousand-euro grant from the EU Web3 Foundation to develop open-source toolkits that preserve human agency against algorithmic spam. The same cryptography that can verify a person can verify a motive. The distinction I am drawing is not centralized versus decentralized. It is between systems that make their reasoning auditable and systems that ask for faith. The SEC regime, for all its mandate, asks for faith in the sufficiency of the mandated facts. The DAO's on-chain record, for all its fidelity, asks for faith that someone will enforce the consequences. Code is law, but ethics is soul — and neither the SEC's forms nor a DAO's smart contracts contains the soul by default. This connects, uncomfortably, to the very reason a piece about Oracle arrived under a blockchain tag. The aggregation software that classified it saw a ticker, a financial event, and a feed optimized for velocity over accuracy. It did not see that the event is a governance event, and that governance — who decides, who answers, what is disclosed and what is hidden — is the actual subject of our industry. When we let machines sort our information, we inherit the machines' blindness. The dispatch is not a blockchain story mislabeled. It is a story about the machinery of trust, which is what blockchain was always supposed to be, and which we have allowed to be reduced to price action and press releases. Let me return briefly to what the news says about Oracle itself, because I have no interest in deconstructing a headline without saying something true about the company beneath it. Oracle's position rests on switching costs that approach gravity. Its databases and enterprise applications are woven so deeply into the operations of banks, telecommunications firms, and governments that replacing them is measured in years and in careers. That moat throws off the kind of recurring, predictable cash flow that makes a founder's long-term conviction rational rather than sentimental. In recent years the company has reoriented its narrative around AI infrastructure — the GPU compute and cloud capacity this bull market rewards with multiples untethered from present earnings — which makes its cash flow heavier, its capital expenditure larger, and its dependence on the patience of capital markets more acute. Ellison's relationship to control is instructive. He has twice left the chief executive's chair and twice reclaimed it, and even from the boardroom his fingerprints remain on the company's direction. A founder who cannot bear to relinquish operational control is, almost by definition, a founder who cannot bear to relinquish shares — because shares are votes, and votes are the last domain he has never surrendered. If the cancellation of his plan is read through that lens, it is less a market signal than an expression of character: the reluctance of a builder to let go of the thing he built. This is not a contradiction of my thesis. It is a fourth possible interpretation, and the dispatch gives the reader no more evidence for it than for the other three. It is precisely in such a period that a founder's personal trading decisions acquire outsized meaning. When a company asks investors to believe in a long, expensive transformation, the behavior of the person who controls its voting power becomes a referendum on that belief. Ellison holding rather than selling can be read as a vote of confidence. Ellison being unable to sell — because a limit was breached or a window closed — can be read as the opposite, dressed in the same headline. And the market, in a bull market above all, selects the flattering interpretation. This distortion is not unique to Oracle; it is the signature of a bull market. Euphoria is a discount rate applied to skepticism. In the current cycle, every token, every founder, every corporate press release is read in the most flattering light, because the alternative — that a signal is neutral or negative — imposes a cognitive cost that outperforming investors are reluctant to pay. The dispatch about Ellison was published into that environment and calibrated for it. Two sentences, no context, and a headline that lets the reader believe what they already wanted to believe. That is what euphoria is: the systematic preference for the reading that requires the least courage. There is a final irony worth naming, and it is the reason I find this dispatch almost funny. Oracle the corporation shares its name with one of the central unsolved problems of our field: the oracle problem, the impossibility of a purely on-chain system knowing anything about the world outside it. Every smart contract that touches reality needs an oracle to tell it what is true, and every oracle is a point of trust. The company called Oracle is, in this sense, the world's largest oracle — a centralized provider of truth to institutions that cannot verify it themselves. And here it is, once again, handing the market a truth it cannot verify: a cancellation without a cause. The name is not a coincidence. It is a confession. I have watched this pattern before. In 2021, I curated a digital exhibition called Soulbound Truths, gathering fifty artists who rejected speculative flipping in favor of community-building tokens, and we built a non-transferable credential system to prove that value could live in identity rather than liquidity. The project drew ten thousand visitors and zero secondary-market trades, which the crypto commentariat treated as failure and I treated as the point. Value that cannot be flipped cannot be faked, and a record that cannot be gamed is worth more than a record that merely cannot be hidden. The Oracle dispatch is the inverse: a record that hides the one thing that matters and hides behind the completeness of everything that does not. Now let me turn the knife on my own tribe, because the decentralized reflex deserves as sharp a challenge as the corporate one. The temptation among readers like us is to conclude that if only Oracle's governance ran on a public ledger, this ambiguity would vanish. It would not. A public ledger records actions with perfect fidelity and interprets them not at all. It would show that a wallet controlled by an Ellison entity failed to execute a scheduled transfer. It would not show why. The chain cannot distinguish confidence from constraint, or conviction from a compliance order, any more than a Form 4 can. Transparency is not the oxygen of trust; it is the light by which trust is either earned or exposed as absent. A brighter light on an ambiguous act leaves the act exactly as ambiguous and merely makes it harder to look away. The deeper contrarian point concerns the limits of disclosure as a governance technology. Both the SEC regime and the on-chain maximalist position rest on the same premise: that publishing enough information will produce the right conclusions. The evidence says otherwise. We have decades of proof that insiders trade through trusts, pledges, and derivatives the disclosure regime barely captures, and we have our own proof that on-chain visibility of treasury movements does not prevent governance capture — it merely documents it. More light does not manufacture accountability. Accountability requires a party who can be held to account, and that requires a legal and ethical architecture no quantity of data supplies. The evangelist who tells you that putting governance on-chain fixes the Oracle problem is selling the same faith the compliance department sells, in a different font. The question worth carrying out of this dispatch is not whether Larry Ellison sells. It is whether we, the builders of the alternative, will insist on recording not merely what happened but why it was permitted to happen, and whether we will admit that a record without a reason is only a more expensive kind of silence. The next governance system worth building will disclose its reasoning, not just its transactions. Until then, a two-sentence flash will keep doing the work that a page of intention should.