Table of Contents
Mark Creaser, CEO of DSV Fund, published a new essay arguing that Bittensor's rapid economic and governance changes are making the network harder for builders and allocators to underwrite.
The essay, titled "Changing the Table Mid-Hand", arrives roughly one month after DSV was reportedly raising $20 million for its TAO-native investment strategy focused on staking, subnet exposure, and dynamic allocation across the Bittensor ecosystem.
Bart Hillerich
Creaser's primary concern is that the rules governing subnet economics have changed so quickly that teams and investors can no longer plan around them with confidence.
"TAO trades at about $190 today (4th August 2026), roughly two thirds off its high. Meanwhile Dropbox is piloting one subnet's product, Intel has co-authored a paper with another, OpenAI with a third. Why does the market refuse to believe? I've spent 18 months invested across this network, and I think I know: Nobody can tell you what the rules will be next Tuesday."
The essay is a critique of Bittensor's recent emission changes, deregistration mechanics, conviction incentives, and concentrated decision-making, balanced against Creaser's view that the network still contains one of the strongest collections of incentive-funded AI teams in crypto.
The Babelbit Situation
Creaser opens with Babelbit, the team behind Bittensor subnet 59, which focuses on low-latency speech-to-speech translation. He describes the project as an example of the kind of technically serious team that can still end up at risk under Bittensor's current market-based pruning system.
Babelbit, according to Creaser, is led by Matthew Karas, a longtime speech technology executive and former launch CTO for ITV Player, and Tom Horner, who previously built systems for environments including flight simulators and trading infrastructure. Creaser points to recent work on real-time dubbing while noting that the subnet was sitting first in the pruning queue when he wrote.
His concern is that deregistration does not directly measure whether a subnet is shipping useful technology, generating revenue, or advancing the network's capabilities. Instead, he argues, the mechanism ultimately reflects market price.
"As I write, Babelbit sits first in the pruning queue, waiting for the axe to be swung. It is not being removed for failing. Bittensor's deregistration mechanism doesn't measure failure, only price."
That distinction runs through the essay.
Creaser acknowledges that deregistration has a legitimate role in keeping the network from filling with low-quality projects or passive yield extractors. But he argues that the current selection process can also remove teams that are building real products if their subnet tokens do not price strongly enough relative to the rest of the market.
The Emission Debate
Much of the essay focuses on the pace of Bittensor's recent protocol changes. Creaser cites several 2026 upgrades, including net TAO flow, price-plus-burn, and the Emission Gate, which shifted passive emissions away from lower-ranked subnets and concentrated more rewards near the top of the subnet market.
He argues that the problem is not only the substance of the changes, but the limited warning given to teams that had already made capital, hiring, infrastructure, and token decisions under prior rules.
Creaser points in particular to registration economics. In his telling, the network previously encouraged teams to pay meaningful sums to register subnets, only for later release notes to describe the old registration model as a failure and move toward the view that a subnet slot should cost little or nothing.
"The network sold tickets at boom prices. Then it announced they had always been mispriced. No refunds."
The problem, Creaser says, is that Bittensor keeps revising its economic design after teams have already committed capital under the previous one.
The same critique applies to conviction locking. Creaser writes that Conviction was introduced after concerns that founders might sell subnet alpha and leave. Teams were encouraged to lock alpha to demonstrate commitment. But shortly after conviction went fully live, the Emission Gate changed the economics for subnets below the cutoff.
Bart Hillerich
Creaser argues that this sequence put teams near the bottom of the rankings in a bind. They locked the token to prove commitment, then watched the funding model get repriced while the locked position stayed unavailable.
"Everyone below the line is now holding a promise they cannot unmake, on a token they cannot sell, inside a business the network has stopped funding."
Price, Governance, and Builder Confidence
Creaser's essay also takes aim at how Bittensor currently values subnet activity.
He says the network has seen examples of teams with public founders, real products, or revenue still being deregistered. He names projects including Vericore, Bitrecs, Bitads, Leoma, Vocence, and Djinn.
The current structure, he thinks, may be pushing serious builders and institutional capital to hesitate, even when they remain interested in Bittensor's technology.
He quotes DSV partner Siam Kidd saying "dTAO is basically uninvestable" under the current pace of rule changes. Creaser also says one serious AI project that had considered building inside Bittensor cut the network from its plans because the protocol had become "a variable they cannot control."
From a governance standpoint, Creaser argues that Bittensor's economic layer remains heavily shaped by the influence of Const and the Opentensor Foundation, despite the network's decentralization ambitions. He does not dismiss the rationale for rapid iteration, but he says the same discretion that allows fast fixes also creates uncertainty for investors and founders.
"A casino at least keeps the same rules all night. Bittensor changes the table mid-hand and calls it an upgrade."
Why Creaser Still Sees Value in Bittensor
Despite the criticism, Creaser is not leaving. He says he remains in Bittensor because the network still contains significant technical work that he believes the broader market is mispricing.
He points to examples across the ecosystem: MinosAI appearing in an OpenAI field report on scientific computing with agentic AI, Lium reporting meaningful annual GPU rental demand, and Macrocosmos pretraining a 100-billion-parameter model across distributed single-GPU machines.
Bart Hillerich
Those examples, Creaser argues, show that Bittensor works as a live network of independent AI teams experimenting with incentive-funded infrastructure, inference, compute, data, and research, well beyond a speculative subnet market. He wants the mechanism to protect and strengthen that builder base rather than make long-term planning harder.
He also concedes that many of the recent changes target real problems. Slots may have been mispriced. Passive emissions may have encouraged extraction. Root staker auto-selling may have created damaging sell pressure. Deregistration may be necessary.
Creaser's stance is not that Bittensor should stop evolving, but that it needs a clearer process for evolution.
His proposed remedy is about the process for change.
- Give more notice before major economic changes
- Establish a cadence that businesses and investors can plan around
- Give proposals such as widening the emission gate from 32 to 64 a full hearing
- Hold still for six months. Let the system run long enough for teams to build under stable assumptions.
"That is not a revolution. It is a pause between revolutions, and it would do more for the price of TAO than another six upgrades."
