In September, NEAR Intents replaced a cryptographic routine inside its signing path. No changelog fanfare. No governance post. Six weeks earlier, at the end of July, a DFINITY patent application crossed into grant. The distance between those two events is roughly thirty days. That window is now the entire factual surface of a dispute that has generated more public text than the code it concerns.
Dominic Williams made the accusation. Haseeb Qureshi answered it from Dragonfly's side of the table. Neither has published a claim chart. Neither has cited a patent number in the open. We have an allegation, a rebuttal, and a September algorithm change with no stated motive. That is not a legal record. It is a press cycle dressed as one.
Strip the personalities. What remains is a question the industry has never had to answer at scale: when a cryptographic primitive migrates from a CC BY paper into a production chain, who owns the right to operate it?
Threshold ECDSA is the primitive. The construction is t-of-n: no single party ever assembles the full private key, and signing requires a quorum of shares. The mathematics entered the public domain through a cluster of papers published between 2019 and 2020 — Gennaro and Goldfeder among them, Canetti's group among them — all released under permissive academic licenses. DFINITY built chain-key signing on that lineage and shipped it to mainnet. NEAR Intents built an intent execution layer that leans on threshold signing to move assets across chains. Both are production systems. Both carry the same structural cost: threshold signatures are slow. Seconds, not milliseconds. That latency is not a bug. It is the price of removing a single point of key custody.
The dependency graph is worth drawing explicitly, because it explains why this argument is louder than its stakes. Upstream sits the academic primitive — public, permissively licensed, unowned. Midstream sits the L1 implementation and the intent/middleware layer, which is where the two parties collide. Downstream sit the bridges, MPC wallets, and cross-chain DApps that consume the signing operation without ever touching the paper. Upstream is the prior-art defense. Midstream is the lawsuit. Downstream is the collateral.
So the dispute is not about whether threshold ECDSA exists. It is about what, precisely, got patented.
My working read: the patentable surface is the implementation envelope, not the primitive. If DFINITY's claims reached the underlying mathematics, the blast radius would include Fireblocks, ZenGo, and every cross-chain bridge running a signing committee. Those products are not in the crosshairs. That asymmetry is diagnostic. Narrow claims, or claims that DFINITY's own counsel is not confident enough to test against a well-funded defendant.
Qureshi's defense has two legs. They are not equally strong, and the louder one is the weaker one.
The weak leg is CC BY. A Creative Commons Attribution license is a copyright instrument. It grants the right to reproduce and distribute a paper. It does not grant the right to practice a claimed invention. Conflating the two is a category error, and it is the error currently being amplified across the timeline. A permissive license on a document is not a patent license on a method. That distinction has been settled law for decades.
The strong leg is prior art. If the underlying papers published before DFINITY's priority date, the claims lack novelty and the patent falls. That is a far more lethal argument than any license defense. But it requires dates, and nobody in public has laid them out side by side. Notice matters here too. Under 35 U.S.C. §154(d), provisional rights can reach back to publication of an application — provided actual notice was given and the granted claims are substantially identical to what was published. The one-month framing may be optimistic. It assumes exposure began at grant. It may have begun earlier.
What would actually settle this is unglamorous. A claim chart. Element by element, mapping each limitation of each independent claim onto a specific function in the accused implementation, with priority dates attached. Until someone produces that document, every public statement is a position, not a finding. In my audit work I have never once seen a specification survive contact with an implementation without a gap. The gap is where the money is.
Then there is the remedy asymmetry everyone keeps skipping. For an infrastructure primitive, damages are the small number. An injunction is the large one. A stop-use order against a signing layer is not a fine. It is an outage. That is why the September algorithm swap matters more than any public statement: it is the only concrete act of mitigation on the record.
I have seen this gap before, from the other direction. In 2017 I spent six months reverse-engineering the Casper FFG specification and built a Python simulator to stress finality conditions against adversarial message schedules. I found three edge cases in the slashing logic that the prose specification did not cover. The lesson stuck: specifications and implementations are different objects. Patent claims read like specifications. Accused code is an implementation. Litigation is won in the gap between them, and that gap is invisible until someone produces the chart.
The same discipline applied during my Terra/Luna forensics. The death spiral was not a surprise. It was a circular dependency that anyone reading the mechanism design could have drawn on a napkin. What made the analysis useful was refusing to narrate. Strip emotion, keep the state transitions. This dispute deserves identical treatment. The state transitions are: application filed, application published, patent granted in late July, algorithm modified in September, accusations exchanged in the weeks after. Everything else is commentary.
Now the capital efficiency layer, which is where this stops being a timeline argument and starts being a cost model. Threshold signing sits under a large share of institutional crypto plumbing — custody, cross-chain settlement, intent routing. Every one of those products is priced off a signing operation. Insert a licensing fee into that stack and you are not taxing a feature. You are taxing throughput. A patent on a signing primitive is a toll booth on liquidity infrastructure, and tolls get passed downstream. Institutionally, that shows up as a basis-point drag on custody products and a widened spread on cross-chain settlement. Small numbers. Compounding ones.
I watched the same mechanism work in the opposite direction in 2024, when I priced the structural efficiency of spot Bitcoin ETFs against direct custody. The wrapper did not change the asset. It changed the friction, and friction is a fee line that institutions will pay to avoid. Legal uncertainty is the same species of cost. An institution sizing an MPC custody product will now apply a haircut for unresolved patent exposure — not because the exposure is large, but because it is unquantified. Unquantified risk gets priced at the worst plausible case. That is the real bill here, and it is being written before any court is involved.
Which brings me to the part of this story almost nobody is arguing correctly.
The loudest defense in the room is the weakest instrument available. CC BY will not save anyone. Prior art might. Claim construction almost certainly will, because claim construction decides the overwhelming majority of patent outcomes and no one outside counsel has read these claims. Meanwhile the party with the most to lose is not NEAR. It is DFINITY. Consensus is not a feature; it is the only truth — and in this community, the consensus on cryptographic primitives is that they belong to everyone. A chain that plants a flag on a signing routine inherits a label it cannot unwind with a blog post. Developer sentiment is a lagging indicator with a long memory. It shows up eighteen months later, in grant applications that never get filed.
There is a second blind spot. DFINITY chose public accusation over filing. Read that as a signal, not a strategy. Parties confident in claim scope file quietly and serve papers. Parties uncertain negotiate in the open. The accusation itself is evidence about the strength of the patent. That inference is not proof. It is base-rate reasoning, and a market of narrative traders consistently underprices it.
Watch three things over the next ninety days. A second accused party. A post-grant review or inter partes petition challenging validity. A defensive patent pool forming around MPC and threshold signing. Any one of those converts a press cycle into a structural repricing of the signing layer. None of them, and this was noise — expensive noise, but noise.
The peg is imaginary. The liquidity is real. So is the exposure window. Thirty days of unlicensed operation does not build a case. It builds a negotiating position. The question is who blinks first, and whether the rest of us are standing inside the blast radius when they do.
