Poet of Respect

Tony Stark built this in a cave.. with a bunch of scraps!

[Green Compute | SN110 @green_compute_]

Myself and the team landed safe in Canada 🇨🇦

The preparation for the big reveal talk on tuesday begins 🙌

[Green Compute | SN110 @green_compute_]

With clients expanding, our work with iota and new commercial inference clients its time for an evening bigger buyback !

With new miners being onbaorded and the sales team expanding ( come and meet some of them at our stand at Exploit🇨🇦 ) we continue to get more excited for the future 🙌

[Minos @theminos_ai]

Synthetic genomes for genomic AI!

Design precise donor-derived datasets with known mutations and full ground truth.

Request early access👇

https://t.co/Al3jCExbvm

[Mariuszek @sobczak_mariusz]

Very interesting @theminos_ai is also teasing a new app with the message “Get your agents ready,” which @centrum_blue will unveil in Montreal. That makes next week especially interesting for anyone following @theminos_ai and the $TAO ecosystem. Very cool, this is going to be huge

[Mark Jeffrey @markjeffrey]

Minos. Bittensor subnet 107.

Genomics + AI.

Something is coming next week 🫡 🍿

[Poet of Respect @PoetOfRespect]

@theminos_ai First and Minos also 1st place let’s goo

[Minos @theminos_ai]

Get your agents ready.

[Minos @theminos_ai]

Chromosomes 17 and 16 are now live on Minos, bringing the total number of chromosomes in the network to 7.

Chr17 carries BRCA1 and TP53, while Chr16 contains key genes implicated in kidney disease and blood disorders.

[TAO.com @taodotcom]

NEW: @theminos_ai has brought chromosomes 16 and 17 live on Minos, taking the network to seven. Chr17 carries BRCA1 and TP53; Chr16 contains genes implicated in kidney disease and blood disorders.
https://t.co/RY5Efbuf1Z

[const @const_reborn]

Hippius on NS this week

[Mariuszek @sobczak_mariusz]

For the past five months, I have been writing about $TAO subnet @theminos_ai , trying to explain why I am so bullish on it. Some people thought I had ulterior motives one even posted my wallet. But I didn’t trade @theminos_ai . I just held and let it accumulate.

A while back, I realized that trading subnet tokens is difficult and my skill set isn’t particularly suited to it. So I decided to study and research subnets. In fact, that’s what the Nerds are. Most of us are researchers playing the long game. There are traders, for sure, but that’s another topic.

Anyhow, back in March or April, @CryptoZPunisher made a post about a brand-new subnet with some real science behind it. I am a nerd at heart, so naturally that appealed to me. But I also knew there were real scientists in the Nerds, so I asked Jo to help me understand what @theminos_ai was building. Part of his PhD thesis focused on genetic mutations, so he was already familiar with the field.

I invited @centrum_blue to the Nerds, and we studied @theminos_ai for hours on end. In the process, I learned about genomics and came to understand what an awesome opportunity it presented.

The opportunity I see is in genomic benchmark data. Difficult test cases help researchers and AI developers measure how accurately their tools perform. My thesis is that supplying high-quality data for those evaluations can become a valuable business.

The road to success for @theminos_ai is now open, and @centrum_blue will lay out the next steps over the coming weeks.

First, at Exploit Summit in Montreal on September 28–29, in front of the Bittensor community.

Then, three weeks later, at ASHG on October 20–24, the largest human genetics and genomics meeting in the world, in the same city. Mamad says Minos has two abstracts there, one on HelixForge and another on the subnet work. He considers both equally important.

Two very different rooms. For me, ASHG is especially significant commercially, because it brings both parts of the work in front of the scientific community and potential customers.

I have long believed @theminos_ai can become a top-three subnet within Bittensor, and that remains my conviction. On top of that, my interactions with @centrum_blue have shown me a super nice and honest person, which makes supporting him that much better.

As for me, I don’t intend to sell anytime soon. I’m happy to hold, let emissions accumulate, and give the business time to develop. I expect the next few years to be super exciting for Minos, with the potential for explosive business growth. That’s why trading it makes little sense to me. I’ve spent months developing this thesis and I believe the biggest opportunity is still ahead.

[Mariuszek @sobczak_mariusz]

A big wallet dumped @theminos_ai today. I have no idea why, but the timing feels a little odd, considering what’s coming up.

@theminos_ai will be presenting at Exploit Summit in Montreal this coming Monday, September 28, followed by the American Society of Human Genetics (ASHG) Annual Meeting on October 20–24. Two opportunities to introduce more people to @theminos_ai and what’s being built on Bittensor $TAO

I think this could be an awesome opportunity for people who have been looking for a good entry. Personally, I think the next few months could be spectacular for @theminos_ai.

[Mariuszek @sobczak_mariusz]

Better incentives make @theminos_ai benchmarking and optimization work more commercially useful.

For $TAO subnets, the real test is producing results customers are willing to pay for.

I like this update. Well done.

[Minos @theminos_ai]

Competition in Bittensor follows incentives. We just rolled out our second major scoring overhaul since launch, upgrading how Minos evaluates genomic predictions and rewards miners.

Incorrect predictions now carry a penalty. Finding mutations that are harder to detect earns a bonus, and precision becomes more valuable.

Every new round shows how miners respond. Explore the scoring shifts with Minos MCP.

[Jolly Green Investor 🍀 @jollygreenmoney]

Elon bull posting $TAO

[Verathos @verathos_ai]

Verathos SN96 belongs in this comparison.

The key question is exactly right: are you proving that compute happened, or verifying that the model you paid for actually served the request?

Verathos is designed around the latter.

Gleipnir v3 binds real inference requests, outputs, sampler configuration and captured execution state into a commitment.

Validators then unpredictably select requests for hard audits that verify nonce-selected execution relations against authenticated model weights/artifacts.

Miners don’t know which eligible request will become the hard audit before committing their execution state.

Separately, hot-capacity audits verify that the advertised GPU compute is actually retained.

And Sleipnir extends this to distributed inference, allowing a model to be served across GPUs/machines while preserving stage-level verification.

It’s probabilistic verification

not a claim that we’re SNARKing every FLOP of every token

but importantly, the object being verified is actual model serving, not an unrelated matrix multiplication puzzle.

That’s the distinction that matters if the end product is inference.

96 $TAO

[Verathos @verathos_ai]

Verathos is introducing Conviction Allowance, becoming the first platform to let you turn alpha conviction into daily inference credit.

Lock Alpha and earn daily usage credits for as long as your Alpha stays locked. The exact amount moves with Alpha’s price, so think of it as a floor rather than a fixed payout.

There’s no minimum commitment and you can unlock whenever you want. Your credit starts showing once the lock is confirmed on-chain.

The credit gives you access to our full lineup of frontier models. If you already hold SN96 Alpha and believe in what we’re building, your conviction should be useful today too.

96 bittensor:native

[Verathos @verathos_ai]

Verathos now serves GLM-5.2 for as low as $0.25 per million input tokens and $1.00 per million output tokens, with GLM-5.3 also coming soon.

For context, the cheapest GLM-5.2 rates currently listed on OpenRouter are $0.343 per million input tokens and $1.078 per million output tokens.

What makes these lower prices possible is Verathos’ mesh infrastructure, Sleipnir.

Instead of relying on a single centralised provider, Sleipnir lets multiple miners pool their GPU capacity and serve larger models across distributed hardware. This creates competition between providers, increases available capacity and lets requests be routed to hardware that can serve them efficiently.

So the lower cost isn’t just a pricing decision. It’s a result of how the compute itself is being coordinated, while still giving you cryptographic verification with every response.

And with GLM-5.3 being added soon, Verathos is continuing to expand the models available through this infrastructure.

If you’re looking for a more affordable way to run verified inference, explore Verathos: https://t.co/SmpcMrWayL

[Verathos @verathos_ai]

https://t.co/LCWpnbxxD5

[TAO.com @taodotcom]

NEW: @verathos_ai has deployed Microtensor's certified mt-code-3g-r1238-5ED3LzEW coding model to the Verathos testnet. The Apache 2.0 model came from Microtensor's round 1,238 arena, expanding the models Bittensor Subnet 96's Sleipnir can route.
https://t.co/hAhqdSZOLq