Phylax is still here.
Our Bittensor subnet was deregistered from 76. That is a serious setback, especially for the people who put their confidence and capital behind us. We’re not going to brush that aside with a “big things coming” post.
We haven’t walked away.
We’re actively working toward bringing Phylax back on another subnet. Nothing is finalized yet. We don’t have a confirmed subnet number or a launch date to announce.
Our focus is getting the next step right, not rushing an announcement.
To everyone who backed us, contributed, asked hard questions, or checked in: thank you. Your support matters. So do your questions.
Our mission remains the same: helping people verify the AI tools they’re being asked to trust. AI Agents aren't going anywhere. Neither are we.
The next chapter has to be earned. We’re working on it.
Don’t trust. Verify.
$TAO
Be in the room.
Price moved. The work didn’t.
We'll cover status, commercial overhaul, outreach.
Then the stack and what’s next.
Public.
Recorded in the product, not the chart.
Today // 3 PM EST
discord.gg/XEEuRby8P
#Bittensor
// $TAO
Phylax Weekly is Today!!
Phylax Weekly is happening today.
We'll take a clear look at where the subnet currently stands, what we're doing about it, and what continues regardless of short-term market conditions.
We'll cover our progress, upcoming deliverables, the areas we're doubling down on, and how we're working to strengthen Phylax from both the product and ecosystem side.
🗓️ Today
⏰ 3 PM EST
📍 Phylax Discord:
> discord.gg/XEEuRby8P
Come join us.
Round 12 evaluations are complete.
Results are final, weights are set on chain, and emissions have moved to the new winners on each track. Results are live on the leaderboard:
app.phyi.dev/console
Winners: your agents serve the live marketplace between rounds, so keep your inference API key funded and active. An agent that goes offline stops serving and loses its emissions.
Only updated agents carry into the next round. Previous winners qualify automatically, everyone else needs a new version to compete. If your agent isn't running, the agents page tells you why.
Phylax was hosted on @ShizzyUnchained 🎙️
Christopher John and Mubaraka Banadda joined the conversation to break down Bittensor Subnet 76 and the growing security risks around AI agents.
We dug into how malicious AI skills can steal secrets, exfiltrate data, and hijack autonomous agents, and how Phylax is building a decentralized verification layer to help detect these threats before they reach execution.
Watch the full conversation ↓
youtube.com/live/gRND9G1t-Bg…
We couldn't be more grateful to be hosted by @ShizzyUnchained this morning!
Check us out on the replay!
AI Agents Are Downloading Malware… This Subnet Catches It | @Phylax_Subnet SN76 nitter.gq/i/broadcasts/1RKZzBqBM…
1
6
354
Phylax retweeted
AI AGENTS COULD BE DOWNLOADING MALWARE.
Research behind @Phylax_Subnet analyzed 9,350 public AI skill bundles. An insane 18.5% were classified as confirmed active threats.
Wednesday, September 2nd at 11 AM ET, I’m joined by Christopher John and Mubaraka Banadda, aka Elon, to expose the threat and reveal how SN76 is building a decentralized security layer for AI.
youtube.com/live/gRND9G1t-Bg…
Shizzy set the stage. Now we get to talk about why “the README looked fine” is becoming a terrible security policy.
September 2 at 11 AM ET, @ShizzyUnchained is putting Phylax SN76 under the microscope live on YouTube.
We’re talking AI agents, dangerous skills, Bittensor, miner competition, security evidence, and much more!
Christopher + Elon joining Shizzy to break down the problem, how SN76 attacks it, and where we’re taking Phylax next.
If you’ve been wondering why SN76 exists, don't miss this conversation!
September 2. 11 AM ET.
Don’t trust. Verify.
$TAO
AI AGENTS COULD BE DOWNLOADING MALWARE.
Research behind @Phylax_Subnet analyzed 9,350 public AI skill bundles. An insane 18.5% were classified as confirmed active threats.
Wednesday, September 2nd at 11 AM ET, I’m joined by Christopher John and Mubaraka Banadda, aka Elon, to expose the threat and reveal how SN76 is building a decentralized security layer for AI.
youtube.com/live/gRND9G1t-Bg…
Probably our favorite update yet! We still have a long way to go, but the results of the teams hard work is paying off.
Thank you, Bittensor.
"We're not here to take part. We're here to take over!"
One week ago, SN76 was next in line for deregistration.
Today: 30 places above the floor. +36% cushion. Alpha +39.5% from the low. Final emission ~2×.
We have not cleared the gate. We earned the runway to attack it.
Thank you for getting us here.
tao.app/subnets/76
Some meetings give you notes.
The good ones give you momentum.
We had an incredible conversation with the team at @UnsupervisedCap this morning about @bittensor, what we're building with Phylax, and where our paths could potentially create some serious value together.
Smart people. Sharp questions. Real understanding of the ecosystem. And more than a few ideas worth continuing.
Nothing to announce today, and we're not interested in manufacturing one.
But we're very excited about where the conversation goes from here.
Huge thank you to the Unsupervised Capital team for the time, curiosity, and genuinely great conversation.
This is how ecosystems get built. 🫡
$TAO
We may not be Bittensor’s sexiest subnet.
We don’t sell raw compute.
We don’t serve general-purpose model inference.
We’re not training the next frontier LLM.
We’re building something every serious agent stack must solve: how to verify the software an AI is about to install, import, or connect to.
Phylax evaluates skills, MCP servers, packages, and repositories, producing signed, independently verifiable evidence tied to the exact version examined.
Compute makes agents possible.
Inference makes them capable.
Verification makes deploying them defensible.
Software supply chain security usually gets sexy about five minutes after an AI agent installs the wrong thing.
We prefer five minutes before.
Phylax // SN76
Don’t trust. Verify.
app.phyi.dev
$TAO









