SN2 Dsperse has now surpassed 1B+ proofs completed.
Proof throughput continues to accelerate, with peak volume reaching 19.4M proofs/day as larger models become increasingly practical to verify at scale.
Verifiable inference infrastructure is moving into production.
https://t.co/qv3NBYgmZy
SN2 Dsperse has now surpassed 1B+ proofs completed.
Proof throughput continues to accelerate, with peak volume reaching 19.4M proofs/day as larger models become increasingly practical to verify at scale.
Verifiable inference infrastructure is moving into production.
https://t.co/qv3NBYgmZy
@Trillion_Tao Incentives is exactly what $TIG needs.
$TAO can fix this.
I tried mining a couple months back and it's basically impossible to do solo. Even with very hefty hardware.
There's also no good mining pools, so the existing 17 (I think) miners is kneecapping the entire project imo.
@sebyrubino Weird af how $tao gets excluded by most of these guys. Would love to understand why...
4 potential subnets on that list and 1 playing AI 'cause it's cool
Insurance companies now use CV models to decide claims. The claimant can't verify which model ran. The adjuster can't either.
A score decided it. No record of how. Verifiable inference fixes this at the root.
https://t.co/pchEKDmwBF
@inference_labs PLEASE get rid of these fake replies!!!
Such a bad look.
Always ~200 of them and awful slop replies.
Amateur from one of $tao's best subnets.
@inference_labs PLEASE get rid of these fake replies!!!
Such a bad look.
Always ~200 of them and awful slop replies.
Amateur from one of $tao's best subnets.
@brt2412 Sounds like a possible locale issue with the input field. Happens often. Special characters like comma (,) or fullstop (.) can get swapped based on region/locale.
A closer look at the Enigma revenue model:
> Each challenge sponsor will be required to stake directly in the subnet, elevating all other stakeholders and encouraging them to produce high-quality challenges.
> Miners will be required to pay a small submission fee each time they submit a solution. These fees will go right back into the subnet in the form of alpha purchases. This mechanism reduces "junk solutions" and increases inflows to the subnet.
> @qBitTensorLabs is given the right to license IP from submissions, giving an additional stream of revenue in the event of a truly groundbreaking discovery.
When a milestone is achieved, the winner will have two options:
1. Claim their prize in alpha -- This is the preferred method, as it keeps sell pressure low. In this case, the winner will get to keep 100% of the advertised prize.
2. Claim their prize in USD -- The winner will be awarded 80% of their prize, with 20% taken by qBitTensor Labs for conversion fees. In this event, liquidation will be done through DATs or other liquidity providers to be as minimally damaging to the subnet as possible.
With these Bittensor-first mechanisms in place, we anticipate that the subnet will flourish, giving holders the stability they need to keep holding and miners adequate rewards for their work.
@numinous_ai just released a new Signals API.
It brings structured data from multiple sources into one place, designed for agents and real-time workflows.
It will be available on #Desearch Playground soon.
👉🏻 Check it out here | https://t.co/wN75ABlVs6
@Altcoinbuzzio Def not the biggest contributor to bittensor either. Chutes, Ridges, Hippius etc. Actual startups. #Sn3 was an amazing showcase of the network’s ability.
@Robin_T100 I guess stating they’re building drug discovery is in part the reason they are here. Communication.
Not at all what they’re doing, but hardly anyone knows what they’re actually building.
🚨 This is one of the strongest enterprise adoption subnet stories we’ve had lately because it moves beyond AI "cool", into regulated budget spend, which institutions would actually pay for.
This is https://t.co/hmlFpx7ddf by $TAO's SN2 @inference_labs.
The only computer vision solution with cryptographic proof of every prediction.
This is end-to-end computer vision verification infrastructure for industries where mistakes carry real legal and safety consequences.
Think about the markets targeted here:
▫️Civil aviation: runway incursion, debris, aircraft inspection
▫️Healthcare: diagnostic imaging, pathology, surgical workflow
▫️Energy: pipeline integrity, power lines, PPE compliance
▫️Agriculture
▫️Manufacturing
▫️Warehousing
▫️Logistics
📷 Deploy Vision AI on any camera feed.
Even cameras you do not own.
Describe what needs to be detected.
Connect the stream.
Get predictions that are cryptographically verified.
From idea to auditable proof in minutes.
Anyone can verify the result without access to the model weights or the underlying private data.
No new hardware.
No changes to existing models.
No specialized TEEs.
Runs on consumer hardware.
That is massive.
The live demo below and in thread, says everything: airport tarmac footage uploaded, prompt set to “Trucks, Aircraft,” 33 frames auto-annotated, 338 annotations generated, model trained, deployed, and every bounding box output is cryptographically verifiable.
Real planes.
Real trucks.
Real compliance value.
This solves the biggest enterprise AI bottleneck:
AI scales at the speed of compute. Enterprises scale at the speed of compliance.
That gap traps billions in delayed deployment.
That is not speculative utility.
That is product-market fit territory.
$TAO