Adeniyi Abiodun, co-founder and chief product officer of Sui, wants the record to show that his team said no. No to forking the EVM. No to the venture capitalists and institutional backers who, during the chain's formative years, pushed what has long been the fastest shortcut in crypto: clone Ethereum's execution environment, inherit its developer base overnight, and ship a compatible chain under a fresh ticker. Abiodun's retrospective arrives with the serene confidence of a man who believes history has already ruled in his favor. "Copying a flawed design," he says, was never a serious option. Sui, in his telling, was and remains "the only viable choice."
That is a remarkable sentence โ not because it is true, but because of what it discloses about the strategic posture of a non-EVM layer-1 in 2026. When a founder of Abiodun's seniority reaches for the word "only," he is not describing a competitive landscape. He is describing a psychological one. The distance between those two things is where the real story lives.
I have spent seventeen years watching crypto founders narrate their own decisions, from the editorial desk to the bleeding edge of on-chain forensics. The most informative part of any retrospective is rarely the claim itself. It is the silence wrapped around it โ the data that goes uncited, the rivals who go unnamed, the economic questions that go unasked. What follows is an attempt to read Sui's silence and to stress-test the "only viable choice" thesis against the technical facts on the ground. This is not a hatchet job on Sui. The architecture deserves the respect it gets. It is a forensic reconstruction of a decision the market has been told to celebrate without ever being shown the arithmetic.
Context: The Diem Ghost and the Non-EVM Bet
To understand why Abiodun's refusal matters, you have to understand what he was refusing โ and what it cost.
Sui is a layer-1 blockchain built by Mysten Labs, a company assembled in 2021 by a cluster of engineers who had worked on Meta's ill-fated Diem stablecoin project. That lineage is not incidental. Diem, for all its regulatory death, produced two durable technical artifacts: the Move programming language, created by Sam Blackshear, and a generation of engineers who had spent years thinking about how to move digital assets safely at scale. When Meta wound the project down, those engineers scattered. Some founded Aptos. Others founded Mysten Labs. Both bet on Move. Both bet against the EVM. And both launched mainnet within roughly seven months of each other โ Aptos in October 2022, Sui in May 2023.
The EVM โ the Ethereum Virtual Machine โ is the execution standard that defines what most of the crypto industry calls "compatible." Fork the EVM and you inherit, in a single stroke, Solidity, the tooling, the wallets, the auditors, the indexers, and a decade of accumulated developer habit. Chains from BNB Smart Chain to Polygon to Avalanche's C-Chain to Fantom built their early momentum on exactly this inheritance. It is, in capital-efficiency terms, the single highest-leverage decision available to a new layer-1. You are not building a chain. You are buying distribution.
Abiodun says his team was advised to take it. "Risk capital and institutions," in his framing, suggested forking the EVM. The team declined. They chose instead to build a chain from first principles around the object model, parallel execution, and Move โ accepting, in Abiodun's own words, "the risk and the time cost."
That is the claim. Now let me examine the arithmetic behind it, because a decision this consequential deserves more than a victory lap.
The Economics of the Shortcut
Before I audit Sui's refusal, let me be fair to the shortcut it refused. The EVM fork is not a lazy choice. It is a rational one, and the market has repeatedly rewarded it.
When BNB Smart Chain forked the EVM in 2020, it did not need to convince a single Solidity developer to learn a new language. It needed only to change a chain ID and point a wallet at a new RPC endpoint. The entire EVM toolchain โ Hardhat, Foundry, MetaMask, Etherscan, the audit firms, the Stack Overflow answers โ came for free. Within months, BSC was processing more transactions than Ethereum itself, and its TVL ballooned into the tens of billions. It bought distribution with compatibility, and the trade worked. Polygon did the same thing more elegantly, positioning itself as a scaling layer rather than a rival. Avalanche built its C-Chain on the EVM and used subnet architecture to differentiate. Fantom chased raw throughput on an EVM-compatible base. None of these chains invented a new execution model. All of them captured real economic value.
This is the counterfactual Abiodun's retrospective never confronts. The fork was not a trap. It was a proven strategy with a decade of evidence behind it. When a founder says he rejected the consensus advice, the interesting question is not whether he was brave. It is whether the advice was actually wrong. And the historical record says the fork-first strategy was, for most chains, correct. The exceptions โ Solana, Sui, Aptos โ are the chains that had something the EVM genuinely could not express. That is a much smaller club than "every project with an independent streak," and membership has to be earned on technical grounds, not asserted in a press cycle.
So the burden of proof sits with Abiodun. To justify refusing the shortcut, he has to demonstrate that the capabilities Sui gained were worth more than the distribution it forfeited. That is the audit I am about to run.
The Technical Substance of "Flawed Design"
Let me start with a confession of method. I don't trust founder retrospectives that don't show their work. In 2017, I broke a story about a state-variable race condition in a Solidity 0.4.19 contract โ a reentrancy bug in the successor to The DAO's fork โ by reading raw commit diffs before the public audit landed. The exposรฉ forced three exchanges to pause listings, and it taught me a permanent lesson: the truth of a technical decision lives in the code and the numbers, never in the press release. So when Abiodun tells me the EVM is "flawed," I want to see the flaw. When he tells me Sui is "the only viable choice," I want to see the denominator.
Let me give him the benefit of the doubt first. Because on the narrow question of whether the EVM has architectural limitations, Abiodun is on defensible ground.
The EVM was designed in 2014 for a world that had not yet imagined DeFi. Its execution model is account-based and globally serial: every transaction mutates a single shared state trie, and the ordering of those mutations is a consensus-critical concern. This design choice has three consequences that any systems engineer would flag.
First, execution is effectively serial. Two transactions that touch completely disjoint parts of the state cannot be processed in parallel without elaborate and error-prone dependency analysis, because the EVM's semantics assume a single, totally ordered history. Ethereum has spent years chasing throughput through rollups and data availability layers rather than through the base execution model, precisely because parallelizing the EVM is a research problem, not a configuration flag. You do not simply turn on parallelism in a machine that was built to be a single-threaded state machine.
Second, the account model is hostile to safe asset handling. In the EVM, a token balance is just a number in a storage slot owned by a contract. There is nothing intrinsic to the language that prevents that number from being mutated by a reentrant call, a delegatecall, or a logic error. The entire discipline of "checks-effects-interactions" exists because Solidity gives developers a loaded gun and asks them to point it away from their feet. The reentrancy bug I chased in 2017 was not an exotic failure. It was the EVM behaving exactly as designed.
Third, the EVM conflates the identity of an asset with the storage slot that holds it. There is no first-class notion of an object with an owner and a lifecycle. Everything is a key-value pair in a giant hash map, and meaning is imposed by convention rather than enforced by the type system. This is why the EVM's history is littered with approval bugs, allowance exploits, and proxy-upgrade disasters โ each one a case of a developer believing a convention that the machine never guaranteed.
These are real criticisms. I have made versions of them myself, often, and usually to the annoyance of whoever was paying me. So Abiodun is not bluffing when he calls the EVM flawed. The question is whether the flaw is fatal or merely inconvenient โ and whether Sui's alternative actually solves it.
Move's Resource Safety, Explained
Sui's answer to the reentrancy problem is Move, the language Mysten inherited from Diem. Move's central innovation is the concept of a "resource" โ a type that is subject to linear logic. A resource cannot be copied and cannot be implicitly discarded. It can only be moved, and it must be explicitly destroyed. This is enforced at the bytecode level by the Move virtual machine, not by developer discipline.
The practical upshot is that the class of bugs that has drained billions from EVM protocols โ the reentrancy exploits, the double-spend-via-reentrancy, the "forgot to update the balance before the external call" family โ is largely structurally impossible in Move. You cannot re-enter a function and duplicate a resource, because the type system forbids duplication. You cannot leave a resource dangling, because the compiler forbids it. This is a genuine advance. It is the kind of design decision that makes security engineers, a tribe I have spent a lot of time with, sit up.
I stress-tested this claim the way I stress-test everything: by trying to break it. In 2020, I executed a fifty-thousand-dollar flash loan arbitrage between Uniswap and Sushiswap โ not for profit, but to map the millisecond latency of oracle manipulation. I spent weeks scripting Python bots to trace the exact path of malicious capital through lending protocols, and I documented a two-million-dollar drain on a lesser-known protocol for what became "The Anatomy of a Flash Loan Attack." What that exercise taught me is that most DeFi exploits are not clever. They are boring. They are the same three or four state-machine errors, replayed against contracts that were audited and still wrong. Move removes an entire shelf of those errors from the toolbox. That is real.
But โ and this is the pivot Abiodun's retrospective skips โ resource safety is a solution to a problem that Sui's architecture creates partly by not having the problem in the first place. Which brings me to the object model.
The Object Model and Parallel Execution
Sui's defining architectural choice is its object-centric data model. Where Ethereum models the world as accounts with balances, Sui models it as a collection of objects, each with an owner, a type, and a version. An object can be owned by a single address, shared across the network, or made immutable. This is a fundamental departure, and it is what makes parallel execution tractable.
Here is the mechanic, stripped of marketing. When a transaction touches only objects owned by a single address, it does not need to go through global consensus at all โ it can be executed locally and finalized fast. When transactions touch disjoint sets of shared objects, they can be executed in parallel, because there is no total ordering requirement between them. Only transactions that contend for the same shared object need to be serialized. Sui's consensus layer is engineered to exploit this.
This is a genuinely clever design, and it is the strongest technical argument for the non-EVM path. The parallelism is not bolted on. It is a consequence of the data model. You cannot get this by forking the EVM, because the EVM's account model does not expose the object boundaries that parallelism requires. This is what Abiodun means when he says the team could not have taken the shortcut. On this specific point, he is right: the shortcut would have foreclosed the architecture.
So the technical case is coherent. Sui built something the EVM could not have given it. The flaw in the retrospective is not the technical claim. It is everything the technical claim is being used to obscure.
The Consensus Layer: Narwhal, Bullshark, Mysticeti
There is a second technical layer to the refusal, and Abiodun's statement gestures at it without naming it. Sui did not just build a different execution model. It built a different consensus model, and the two are inseparable.
Sui's original design separated the mempool from consensus. Narwhal handled data dissemination โ a directed acyclic graph of transactions that guaranteed availability without requiring agreement on order. Bullshark sat on top of that DAG and produced the actual ordering. This split matters because it decouples two hard problems: getting the data out to everyone, and deciding what order it goes in. Later, Mysten replaced this stack with Mysticeti, a consensus protocol explicitly engineered to push latency toward its theoretical limits by committing blocks without the extra communication rounds that earlier DAG protocols required.
None of this is achievable by forking the EVM, because the EVM does not have a separate data-availability layer to fork. Its execution and its consensus are entangled in the same state machine. To build what Sui built, you have to start from a blank page and design the consensus and the execution model together. That is the deepest sense in which the team's decision was irreversible: the fork would not merely have changed a language. It would have locked in an entirely different set of engineering possibilities.
I will be blunt about what this means for the retrospective's rhetoric. The technical achievement is real. But the existence of a real achievement does not retroactively make the strategic bet correct. Building something hard and building something valuable are distinct claims, and Abiodun's statement collapses them into one. The consensus design is evidence of capability. It is not evidence of adoption.
The Developer Tax
Let me now do the arithmetic Abiodun did not do.
Every chain that refuses the EVM pays a tax, and the tax is denominated in developers. When BNB Smart Chain forked the EVM, it did not need to convince a single Solidity developer to learn a new language. It needed only to change a chain ID and point a wallet at a new RPC endpoint. The entire EVM toolchain came for free. That is why BSC captured billions in TVL within months.
Sui, by refusing, agreed to pay for distribution the hard way. Move is a different language with a different mental model. Its toolchain, while improving, is younger and thinner than Solidity's. Its auditor pool is smaller. Its developer mindshare is a fraction of the EVM's. Every DApp on Sui had to be built natively, by a developer who first had to learn Move. This is the "time cost" Abiodun mentions in passing โ and it is not a rounding error. It is the single largest strategic cost of the decision, and it compounds.
I have seen this tax paid in other contexts. When I traced the 2021 NFT metadata crisis โ the analysis I called "decoding the heuristic break in 2021 NFT metadata" โ the pattern that emerged was that infrastructure constraints, not artistic merit, determined which collections survived. The same logic applies here. Developer availability is infrastructure. A chain that cannot hire from the dominant talent pool grows slower than a chain that can, no matter how elegant its architecture. Sui chose elegance over access. That is a defensible choice, but it is a choice with a price, and the retrospective quotes the price without ever tallying it.
The Aptos Problem
The Diem ghost cuts both ways here. Sui's closest competitor is not Ethereum. It is Aptos โ another Move chain, built by overlapping alumni of the same Meta project, launched within months of Sui, chasing the same developer pool with the same language.
This is the single most important fact the retrospective omits, and its omission is structural rather than accidental. When two teams inherit the same language and the same lineage, they do not just compete for users. They compete for the scarce resource of Move developers, and they cannibalize each other's ecosystem narrative. Every developer who learns Move and chooses Aptos is a developer Sui did not get. Every enterprise evaluating a Move chain has two near-identical options and must be given a reason to pick one. The differentiation that Abiodun claims for Sui has to be argued against a rival that shares its DNA, not against the EVM it already dismissed.
A story about being "the only viable choice" cannot survive contact with the existence of a near-identical rival. So the retrospective does not mention it. It frames the decision as Sui versus the EVM, a binary that flatters the team, when the actual competitive set is Sui versus Aptos versus Solana versus every parallelizing EVM chain on the horizon. I have watched this move before. In 2026, when I investigated the AI-agent fraud ring I documented as "The Synthetic Pump," the operators of the scheme never mentioned the competing bots in the same niche. They presented their token as the only game in town, because a crowded field is a harder story to sell. Founders and fraudsters share one instinct: narrow the frame until the claim becomes true.
The Token Gap Nobody Mentioned
Now the part that should worry any reader treating this statement as a bullish signal.
Abiodun's retrospective is entirely about technical correctness. It says nothing about token economics โ no supply, no emissions, no unlocks, no value capture, no staking economics. This is a conspicuous silence, and it maps onto a well-worn pattern in crypto communications: when the technical story is strong and the economic story is weak, you talk about the technical story.
I have watched this pattern destroy retail investors more times than I can count. The most painful version was Terra-Luna. In early 2022, I published a pre-mortem series called "The House Always Wins (Until It Doesn't)," in which I dissected Anchor Protocol's yield sustainability and identified a negative feedback loop in the collateralization ratio. The market laughed at me. I stood my ground in viral threads, defending the math against the FUD label. Forty-eight hours later, the de-peg began. The lesson I took from that experience was not that I was clever. It was that a beautiful mechanism and a valuable token are two different things, and founders are rarely motivated to point out the difference.
A layer-1 can be technically excellent and its token can still fail to capture that excellence. Value accrues to a token only if the token is structurally required to use the network and if the network's activity translates into demand for the token. Sui's fees are paid in SUI, and SUI is staked to validators, so there is a value-capture path. But the strength of that path depends on activity that Abiodun's retrospective does not quantify: TVL, active addresses, transaction volume, developer count. Without those numbers, "the route was correct" is a statement about engineering, not about economics. And the market prices economics.
This is where I part ways with the retrospective's implicit framing. The EVM fork decision was a bet that differentiation would eventually be worth more than distribution. That bet may still pay off. But the retrospective asks you to grade the bet on the strength of its logic rather than its outcome โ and the outcome, measured in ecosystem data, is the one thing the statement refuses to show. This is the same move I have criticized in the NFT market, where dynamic metadata and programmable royalties are sold as innovation while the artists who need stable buyers are handed a more complex tech stack instead. Complexity is not value. It is a cost that someone downstream has to absorb.
Contrarian: The Silence Around "Only"
Here is the angle nobody covering this statement will take, because it requires reading what is not there.
The most important word in Abiodun's retrospective is "only." "Sui is the only viable choice." Set aside whether this is true โ it plainly is not, in any literal sense. Solana processes more transactions than any chain in the industry and is not an EVM fork. The EVM L2 ecosystem settles more value than any non-EVM chain. Aptos runs the same Move language and the same object-adjacent architecture. Monad and Sei are actively parallelizing the EVM itself, which quietly dismantles the premise that parallelism requires abandoning the EVM. If "viable" means "can host applications at scale," then Abiodun's sentence is not a claim. It is a symptom.
The symptom is this: a founder does not reach for the word "only" when he is winning. He reaches for it when he is being asked to justify why he isn't.
I have seen this before, and I have learned to treat it as a diagnostic. In 2021, during the NFT boom, I ran a script across ten thousand top collections and found that roughly fifteen percent would lose their images entirely if a handful of centralized IPFS gateways failed. I published the analysis as "The Fragile Canvas," arguing that most NFTs were, functionally, broken hyperlinks. The response was instructive. Founders who were confident in their infrastructure engaged with the data. Founders who were not attacked the messenger and doubled down on the rhetoric of decentralization. The volume of the rebuttal was inversely proportional to the strength of the architecture. I came to call this decoding the heuristic break โ the moment a claim's confidence outruns its evidence, and the overreach itself becomes the tell.
Abiodun's "only viable choice" is a heuristic break of exactly this shape. It is a maximalist claim deployed in defense of a decision that, by his own admission, took years to vindicate and is still being proven. A confident winner says "we were early and the market caught up." A founder who feels the ground shifting says "we were the only ones who could have done this." The grammar of the second sentence is defensive, not declarative. It is the language of a person arguing with an invisible skeptic, and the skeptic is usually the market.
And notice what the retrospective asks you to focus on: the moment of decision, years ago. It draws your attention backward, to a fork in the road that has already been taken, and invites you to admire the courage of the choice. This is a classic narrative maneuver, and it is worth naming. When a project's forward story is strong โ rising usage, shipping milestones, a clear roadmap โ founders talk about the future. When the forward story is uncertain, they retreat to the past, where the choices are already made and the scoreboard can be curated. Abiodun's retrospective is a retreat to the past. "The next few years will be very interesting," he says, which is what you say when you cannot yet say what will be interesting about them.
There is also a quieter omission worth flagging. The retrospective names no investors. It refers only to "risk capital and institutions" that advised the fork. In a decision framed as intellectual courage against capital, the identity of the capital matters enormously โ because it tells you whether the advice was naive or informed, self-interested or genuinely strategic. The anonymity is convenient. It lets the story cast the team as independent visionaries without ever having to defend that framing against the actual records of the funds involved. I am not accusing anyone of dishonesty. I am pointing out that a narrative about resisting capital is more persuasive when the capital has no name.
Finally, the retrospective never explains why now. This is the question I always ask first, and it is the one the statement refuses to answer. A founder does not spontaneously publish a defense of a years-old decision. He publishes it because something prompted him โ a conference panel, a competitor's milestone, an internal review, a moment of market attention that needed to be captured. When a message cannot say why it arrived today, the safest assumption is that its real driver is reputation management rather than information. The retrospective is not news. It is positioning dressed as reflection.
The Regulatory Non-Question
The retrospective is also silent on regulation, and that silence is worth a moment, because it is the kind of silence that shapes how a project is perceived by the institutions that now dominate crypto capital flows.
Abiodun's statement is a public act of narrative construction. It paints Sui as a technical project โ an engineering achievement โ rather than a financial one. That framing is not neutral in a regulatory context. When the SEC and its global counterparts apply the Howey test, they look for a common enterprise, an expectation of profit, and reliance on the efforts of others. A layer-1 that presents itself primarily as infrastructure has a rhetorical advantage over one that presents itself as an investment product, and founders know it. This does not make the framing deceptive. But it does mean the retrospective is doing double duty: it is a technical argument and a regulatory posture at the same time.
I would caution readers against reading any compliance benefit into the non-EVM route itself. Whether a chain forked the EVM or built from scratch is irrelevant to its regulatory exposure. That exposure lives in token distribution, staking mechanics, foundation structure, and marketing โ not in the virtual machine. And I would extend the same caution to the broader geography of crypto regulation, where jurisdictions compete for the appearance of legitimacy more than the substance of it. Hong Kong's licensing regime, for instance, is less about embracing innovation than about seizing the position Singapore has occupied in Asian finance. The competition is between cities, not between philosophies. Sui's technical autonomy buys it nothing in that contest.
What Proper Diligence Would Require
If you want to actually evaluate Abiodun's claim, here is the work the retrospective leaves undone.
First, pull the chain's fundamentals. Total value locked, active addresses, transaction count, and fee revenue โ compared not against Ethereum but against Solana and Aptos, the two chains that most directly bound Sui's market. A technical story that cannot be confirmed by usage is a story without a scoreboard.
Second, measure the developer flow. The Electric Capital developer reports track which ecosystems are gaining and losing contributors. If Move's share of total developers is flat or declining while the EVM's parallelizing variants grow, then the "route correct" thesis is being falsified in real time, regardless of how elegant the architecture is.
Third, read the unlock schedule. SUI's emission curve and vesting cliffs determine the supply pressure that any demand-side narrative must overcome. A chain can be growing and its token can still be falling, and the difference between those two facts is almost always on the unlock calendar.
Fourth, find the original post. What the retrospective summarizes is likely a longer, more specific statement. If the original contains data โ numbers, comparisons, milestones โ then the summary understates it and the analysis should be redone. If the original is as data-free as the summary, then the statement is what I think it is: positioning.
None of this is exotic. It is the minimum a serious reader should demand before converting a founder's reflection into an investment thesis. The retrospective's greatest failure is not that it is wrong. It is that it invites you to skip this work by substituting conviction for evidence.
Takeaway: The Fork Not Taken
So what should you actually do with Abiodun's statement?
Treat it as a narrative artifact, not a fundamental. It contains no new technical disclosure, no on-chain data, no token mechanics, no regulatory filing. It is a founder reaffirming a decision his team made years ago, in language calibrated to shore up community conviction. That is not nothing โ conviction is a real force in a market that prices stories โ but it is not evidence about the protocol's health, and it should never be entered into a valuation model.
The bet Sui made is real and it is defensible: a from-scratch architecture around objects, parallelism, and Move buys capabilities no EVM fork can replicate. The cost of that bet is also real and less discussed: a developer ecosystem that must be built natively, a direct rival in Aptos that shares its lineage, and a token whose value capture remains unproven in the absence of the data the retrospective declines to supply. The fork not taken was the right call for the architecture. Whether it was the right call for the token is a question the statement is careful not to raise.
So read the silence. Watch for the numbers Abiodun left out โ TVL, active addresses, developer share against Aptos, and the SUI unlock schedule โ because those will settle the argument that rhetoric cannot. The EVM is being parallelized from within. The non-EVM club is not shrinking because of Sui's courage but because of engineering that Sui does not own. And when the next founder tells you his project was the only viable choice, ask him the same question I would ask anyone making a maximalist claim in a market with no monopolies: if you were truly the only option, why do you sound like you are still trying to convince me?