KubeTEE AI

Verifiable compute. Decentralized AI on Bittensor SN90. The Confidential AI Factory

[KubeTEE AI @KubeTEEAI]

True

[Alex DRocks @DrocksAlex2]

$TAO Subnet 90 owner, a very experimented dev and respectable miner on Bittensor, @pamanseau said something bold about @jon_durbin 's work on Parallax tech with @chutes_ai .

Quoting him: "I compare Jon Durbin to Abraham Louis Breguet for watchmaking revolutionary inventions and improvements like the Tourbillon, Perpetual Calendar and Parachute (yes this is for the escapement choc absorber) he did his inventions 250 years ago and we still use them today in watches.
I'm sure we will still use Parallax in 250 years from now"

[SayGM @say_gm_]

MiMo-V2.6-Pro UltraSpeed just got better on SayGm.
It's now:

• 10% cheaper than OpenRouter
• Twice as fast
• Sealed in a confidential TEE, so no one can read your prompts, including us or Xiaomi

It was already great value, but now it's remarkable.

Served by @KubeTEEAI.

[KubeTEE AI @KubeTEEAI]

View our Keynote at Exploit Montreal Sep 28

https://t.co/rpYfXcNeYq

[Exploit Summit @ExploitSummit]

Keeping your model private on someone else’s GPU.

@pamanseau of @KubeTEEAI is live on the OTF stage, discussing confidential AI across decentralized clusters and protecting model weights and data from the host.

Watch live: https://t.co/TEzlPlR5G2

[Exploit Summit @ExploitSummit]

“The product is the attestation.”

@pamanseau of @KubeTEEAI on what enterprises need before trusting a decentralized AI service: proof they can verify and their compliance team can use.

[KubeTEE AI @KubeTEEAI]

Getting Ready for the keynote Monday at 14:00 EST.
The live streaming link will be in the comments.

[KubeTEE AI @KubeTEEAI]

https://t.co/xNRywauAJT

[KubeTEE AI @KubeTEEAI]

Getting Ready for the keynote Monday at 14:00 EST.
The live streaming link will be in the comments.

[KubeTEE AI @KubeTEEAI]

@TaoOutsider He’s not the owner if SN90 anymore we are.

[⌞· τaoli ·⌟ @taoleeh]

KubeTEE (SN90 - @KubeTEEAI) is now a provider on SayGM (SN28 - @say_gm_), serving Xiaomi's MiMo V2.6 Pro UltraSpeed at 10% off list.

Every request runs inside a TEE (a sealed enclave that can prove what code touched it), with no prompt logging and no silent model swaps. Same OpenAI-compatible surface, so you point your SDK at the gateway and pass the model string.

One subnet's hardware, another subnet's storefront, a Xiaomi model inside. Decentralized inference is compounding.

$TAO #Bittensor

[DaτDudεDereκ @Dat_Dude_Derek_]

#SN90 @KubeTEEAI seems to be flying under the radar. @pamanseau will be at Exploit next week. Interesting…

[SayGM @say_gm_]

MiMo-V2.6-Pro-UltraSpeed from @XiaomiMiMo is now live on SayGm.

Served by @KubeTEEAI at 10% off list, cheaper than going direct to Xiaomi or through OpenRouter.

Every request runs in a TEE so no prompt logging.

OpenAI-compatible. Swap the model string and go.

[Belmont Labs @belmontlabs_ai]

BT Labs is now Belmont Labs.

New name. Same team. Bigger mission.
We’re building AI your enterprise owns: The models, the data, and the intelligence that compounds with your business.

Welcome to Belmont Labs.

[Exploit Summit @ExploitSummit]

Decentralized compute has a trust problem: the person who owns the machine can see everything running on it.

@KubeTEEAI fixes that by running AI workloads and services inside hardware-enforced enclaves. Models and data stay encrypted at rest, in transit, and in use, so even the host operator can’t read them.

Proud to announce them as a Gold Sponsor: https://t.co/CgJmJoRXKk

[Exploit Summit @ExploitSummit]

The cloud asks you to trust whoever owns the machine. KubeTEE is built so you don’t have to.

At Exploit, @pamanseau of @KubeTEEAI shows how confidential Kubernetes can run AI across decentralized GPU clusters while keeping models + data hidden - even from the host: https://t.co/CgJmJoRXKk

[KubeTEE AI @KubeTEEAI]

Keynote Monday September 28th @ 14:00

[Exploit Summit @ExploitSummit]

For 4 decades, @pamanseau has been building infrastructure - from one of Canada’s early ISPs to cloud, Kubernetes + hardware-level security.

Now he’s founder of @KubeTEEAI and working on secure infrastructure designed to run workloads across decentralized multi-cluster environments while protecting data in use.

Catch him at Exploit: https://t.co/CgJmJoRXKk

[KubeTEE AI @KubeTEEAI]

@mogmachine Team Work greatly appreciated @say_gm_

[mogmachine (ττ) @mogmachine]

https://t.co/VGKeDDQjr2

Once again getting those models to your agents before anyone else.

Thanks again also to @KubeTEEAI on this one.

[KubeTEE AI @KubeTEEAI]

Proud to sponsor all video streaming and recording of the event in Montreal

[Exploit Summit @ExploitSummit]

Decentralized compute has a trust problem: the person who owns the machine can see everything running on it.

@KubeTEEAI fixes that by running AI workloads and services inside hardware-enforced enclaves. Models and data stay encrypted at rest, in transit, and in use, so even the host operator can’t read them.

Proud to announce them as a Gold Sponsor: https://t.co/CgJmJoRXKk

[Donnie Danko // CHΛOS (🐦‍⬛, 🏴‍☠️) @daanisharif]

Exploit Summit (@ExploitSummit) keeps surfacing people I've been quietly tracking. Pierre Manseau from KubeTEE (SN90 - @KubeTEEAI) is one of them.

He's the miner who brought confidential compute nodes to Targon (SN4), helped Chutes (SN64) get B200 and B300 nodes online, worked with Lium (SN51) on the same. Now he's built the thing he kept circling: an "AI factory" on decentralized Kubernetes clusters, everything running in Trusted Execution Environments.

The gm (@say_gm_) angle is worth watching too; SN28 pulls inference from SN90's idle capacity, GLM-5.2 running 64% below retail. Two subnets, one feeding the other. The two also put Ornith-1.5-397B live before anyone else in the world.

The question for these guys is; everyone's pitching confidential compute to enterprises now. So show me the remote attestation.

Can a customer cryptographically verify the exact image running on their data, or does it still end at "trust us"?

Sept 28-29, New City Gas, Montreal, and a question that deserves a live demo.