Irresponsibly long τao. Life. Liberτy. Biττensor.
Church of Rao, τaoτardisτan
Joined September 2021
- Tweets12.8K
- Following3.1K
- Followers1.7K
- Likes79.6K
Rewatching some parts of @ExploitSummit.
High praise from the goat @jon_durbin to @pamanseau of @KubeTEEAI.
Highly recommend watching the whole talk to understand what’s being built on 90.
bittensor:native
Chairman τao retweeted
In KubeTEE, when a pod starts, it does an attestation that Confidential Computing is active and trusted by Intel Trust Authority. Only then does the pod receive the secrets from the KBS that gates it.
Chairman τao retweeted
Confidential Computing TEE attestation would kill any code changes. KubeTEE policies protect against this.
🚨 PHP2Shell exploited in real-world attacks.
An attacker chained PHP2Shell with a Linux kernel flaw to gain root access to 25 servers across four countries, including three in a U.S. government-affiliated environment.
The intrusion extended into the internal network, where the attacker targeted Active Directory.
Suspected China nexus. Signs of AI-assisted tooling.
Read the full report: thehackernews.com/2026/10/ph…
Chairman τao retweeted
Reputed engineer and Bittensor native @pamanseau running @KubeTEEAI (SN90)
Sitting at $3.5m MC
bittensor:native
AI can’t whisper.
Yet.
Right now, every secret you hand it gets read out loud...
In a room you don’t own.
Sounds dramatic.
But it’s actually just how computers work.
Your data is locked when it’s stored.
It’s locked when it travels.
But AI can’t read a locked file.
So the moment it gets to work, the lock comes off.
Your data sits wide open on someone else’s hardware.
And whoever runs that hardware has a front-row seat.
—
Most people never think about this.
Hospitals do.
Banks do.
Anyone with secrets worth stealing does.
That’s why the data AI wants most is the data companies trust it with least.
Not because AI isn’t smart enough.
Because nobody trusts the room it works in.
So here’s the real question.
What if the room could keep a secret?
—
That’s what one Bittensor subnet is building.
KubeTEE.
Subnet 90.
The idea is simple.
Put the AI inside a hardware-sealed vault.
The machine still does the work.
The owner just can’t see inside.
That vault is called a Trusted Execution Environment.
And it’s not some startup’s fantasy.
The locks come from Intel and NVIDIA.
KubeTEE didn’t invent the lock.
It’s building the network around it.
In plain English?
AI learns to whisper.
—
Now, this is Crypto Twitter.
So you’re thinking: “Cool story, bro. Is any of it real?”
Fair.
Let’s check.
Their docs show a production cluster run by BTLABS.
56 GPUs.
32 H100s.
24 H200s.
Not a render.
Not a pitch deck.
Documented iron.
Hardware is nice.
Revenue is nicer.
—
KubeTEE runs paid AI inference through sayGM, Subnet 28.
That money gets recycled into SN90’s alpha token.
On-chain.
About $29K since August 23.
Roughly $453 a day.
Every swap linked to a transaction you can click.
Now slow down for this part.
That is NOT SN90 earning $453 a day from its own customers.
Anyone who tells you it is wants something from you.
What it IS: a real service feeding a real on-chain loop.
That’s more than a promise.
—
So what stops someone from copying it?
If all it took was GPUs, nothing.
Anyone with a credit card can rent an H100.
The hard part is everything else.
Secure.
Reliable.
Cheap enough that a bank says yes.
Hard is the moat.
—
And here’s where Bittensor comes in.
Right now, emissions help pay the miners.
Those are the training wheels.
Paying customers are the bike.
SN90’s whole future is the day the training wheels come off.
—
Now the part most people skip.
Big cloud already sells confidential computing.
Other subnets want the same workloads.
The tech is brutally complex.
And direct paying demand for SN90 is still early.
So no, this isn’t a sure thing.
It’s a question with receipts attached.
—
Here’s how you’ll know which way it breaks:
→ Does the tech hold up?
→ Do customers pay?
→ Does usage grow?
→ Does the team keep shipping?
→ Do the tokenomics reward holders?
If those keep climbing while the market still calls this “just another compute subnet”…
That gap is the story.
—
Not hype.
Not hopium.
Receipts.
AI can’t whisper.
Yet.
Now you know who’s working on the “yet.”
DYOR.
Verify the numbers.
Watch the execution.
Chairman τao retweeted
On the road with Pip Studios... can you guess where we are 👀
@vidaio_ bittensor:native
Chairman τao retweeted
Last time we looked at Security. Today: Public Safety.
Law enforcement: body-cams, dash-cams and interview rooms produce footage that has to be offloaded, labeled and stored, then pulled for reviews, records requests, training and court.
Transportation hubs: airports, rail, metro and ports run cameras across platforms, gates, terminals and docks. The footage settles incidents, and it also shows how people and vehicles actually move through a site.
Emergency services and fire: stations, vehicles and incident scenes generate footage that can be reviewed to see how a response went and what could be done faster.
Four more in Public Safety:
Smart cities and municipal CCTV: intersections, transit stops, public squares
Border and customs: crossings, inspection lanes, cargo
Government and court archives: hearings, records, evidence storage
Defense and ISR: surveillance and reconnaissance video
In public safety, footage is evidence. It’s also how agencies review a response, train their people and find ways to work faster. Either way it has to be kept, found and produced on demand. Vidaio compresses, tags, names, sorts, and files it.
Next up: Health.
@vidaio_ $TAO
Just how many cameras are out there?
Last week we named five categories of industries we can serve. Starting today, we’ll take a closer look at each of them one at a time. First up: Security.
Retail and logistics: cameras cover the sales floor, the stockroom and the loading dock, often across many locations under one operator. Every site produces its own footage, and eventually footage is needed from the right clip and the right camera on the right day.
Casinos, gaming and hospitality: gaming floors, cash cages and entrances are recorded around the clock, and gaming regulators set rules for how long that footage has to be kept.
Corrections and detention: housing units, corridors and intake areas are recorded constantly, and incident footage may be needed long after the fact.
Four more in Security:
Banking and financial branches: counters, ATMs, vaults
Education and campuses: gates, hallways, parking lots
Commercial buildings, real estate and parking: lobbies, garages, access points
Stadiums and venues: concourses, entrances, vendor sites, crowd areas
In security, footage is evidence. It arrives constantly and incidents have to stay findable and affordable to keep. Vidaio compresses, tags, names, sorts, and files it.
Next up: Public Safety.
@vidaio_ bittensor:native
Chairman τao retweeted
Ditto’s improvement flywheel just switched on.
The agent competition was missing its counterbalance. Our new generator competition fixes that. Live on SN118.
Agent miners build better memory agents. Generator miners build programs that keep producing questions those agents still struggle with.
As agents close one blind spot, generators have to find the next. Two adversarial competitions, pushing each other to get better.
We designed for miners who enter both sides, too. Your own agents don’t count toward your generator’s score, and your generator’s cases are excluded from comparisons involving your agents. Hard questions must still be solvable by independent owners. The rules are built to reward progress and make gaming harder.
For the first time, miners are scaling DittoBench. Its next frontier no longer has to wait for the Ditto team to write every test.
Better tests push better agents. Better agents demand better tests.
A recursive improvement loop. This is how Ditto agents scale for the long run.
This launch is the beginning. The flywheel is moving.
dittobench.ai/generators
dittobench.ai/pipeline
Chairman τao retweeted
John Wick with wall hacks
The overhead hallway fight in John Wick 4 has the best choreography I’ve ever seen in an action movie. One unbroken shot, dozens of people, smoke, muzzle flashes, bodies everywhere.
So I made it my stress test. Raw run, misses and all. Dark corners and people on the floor are where it struggles.
What scene should I break it on next?
Chairman τao retweeted
Great post by @haitzu.
I know @theminos_ai, one of the most interesting subnets on $TAO, has been a little quiet these past few weeks, largely because the team was preparing for the genetics conference in Montreal.
This was an important event where thousands of geneticists, researchers, and industry leaders gathered to discuss the future of genomics and medicine.
Think about what that means. A young Bittensor subnet had the opportunity to present its technology directly to the people who could actually use it.
To me, that’s far more important than short-term price action.
@theminos_ai is tackling a real problem in genomics: making genetic analysis more accurate, reliable, and continuously improving.
HelixForge is a big part of what makes this opportunity so interesting.
It’s based on the technology, the team, and the size of the problem they’re trying to solve.
That’s where things could get really interesting for @theminos_ai and $TAO.
I love studying the growth of monster subnets: how do new open source subnets grow from zero to heroes?
Minos (SN107) was a blessed son when it launched earlier this year
Alongside Metanova (SN68), @theminos_ai occupy an important role for Bittensor to pursue novel, open source experimentations for life sciences
Running an always on genomics prediction competition, it has been one of those "sounds too smart for me" type of subnets for many
However, being featured by OpenAI in a field report and receiving an investment from Stillcore Capital has shown how explosive new development teams going directly into the belly of the open source community can be
Another proud story for Bittensor to continue telling the story of
Chairman τao retweeted
39 experiments in two months, all with one goal:
the best encryption and the best speed, for everyday use
Privacy often comes with a cost to speed. We're always running research on our code to improve and make that tradeoff disappear
Ideally you don't even have to think about it
Our new report collects the last couple of months of that research, with every experiment on a timeline showing what it measured and whether we kept it. Even failed tests are valuable to learn from
A few that stood out:
- in our phone tests, a Wikipedia page went from 2.5-4 seconds to about half a second
- a local test link hit gigabit-class speeds after we changed how much data is kept in flight
- a tighter memory limit actually hurt speeds by about a third, so we killed that change
Full report below
Chairman τao retweeted
$ZEC climbed ~4x to ~$1,250 in roughly a year
You're probably thinking the privacy ship has already sailed, but you'd be mistaken
Beyond Zcash, emerging protocols are bringing privacy to payments, trading & internet access
Three names on our watchlist:
1) @NullMaskio $MASK
Brings shielded transfers and swaps that obscure the trader’s identity to existing EVM wallets such as MetaMask and Rabby through a custom RPC
Founded by Tomas Krnak, whose team developed shielded transaction support for Trezor, it reported $50m+ in combined deposit, withdrawal and swap volume between 1–19 October
Using familiar wallets makes adoption easier, although deposits, withdrawals and swap amounts remain public, and the operator can read private balances and history
2) @UmbraPrivacy $UMBRA
Supports private balances, payments and swaps on Solana, helping users protect financial information such as holdings, salaries and commercial transactions
Its October 2025 MetaDAO token sale attracted ~$155m in commitments against a $3m cap, followed by its public wallet launch in March 2026
Payroll and treasury management offer practical applications, where public transactions can expose compensation, counterparties and business relationships
Deposits, withdrawals and network metadata still carry privacy limitations
3) @urnetwork on Bittensor SN25
Encrypts and routes internet traffic through distributed connections, founded by former Palantir engineer Brien Colwell
The existing URnetwork has reported 350K+ monthly users, with 100K+ providers being transitioned into Bittensor miners following October’s mainnet launch
Bittensor adds incentives for supplying connectivity and validators that check delivery, building on an existing network and user base
Each addresses a different privacy need, with different trust assumptions and token economics
Our comparison looks at adoption, who pays for each service and how that activity benefits the token
Follow @KhalaResearch for research on emerging crypto protocols
Chairman τao retweeted
Introducing the Bittensor <> Ethereum bridge on Taoswap
You can now bridge USDC or ETH from Ethereum into TAO or any subnet token on Bittensor, and convert TAO back to USDC or ETH.
Powered by @taodotcom.
Chairman τao retweeted
Thank you @GeneticsSociety.
It was a privilege to be in Montreal and share how @opentensor 's Bittensor mission enables Minos: decentralized compute and continuous validation so genomic AI can be trained, benchmarked, and improved.
To everyone who asked hard questions or grabbed merch, thank you. Excited for what's next.
Chairman τao retweeted
PersistentAI is becoming a network operator on UR (@urnetwork).
What we’re cooking:
Joint privacy research - looking into zkAPI & private, open-model routing.
Persistent’s Private Inference (@P_Inference) hooking into UR.io to make it’s traffic anonymous via UR’s proxy network.
Eventual Private Inference subscription layer available to all @urnetwork users.
Say it with us: Private is Better
