S

[⌞· τaoli ·⌟ @taoleeh]

Verathos pays referrers up to $0.30 of emission for every dollar a customer tops up, spread over 30 days.

Any registered hotkey can hold a code, whether it serves models or not. Referring yourself loses money by design.

Two more lanes are nearly done: an Apple private lane that lets a Mac serve inference privately and earn more per token, and a training competition scored on technique rather than GPU count.

Follow Verathos (SN96): @verathos_ai

$TAO #Bittensor

—

Everything important that happens in Bittensor, delivered weekly:

👉https://t.co/jIPtAAQd8i

[Cats @Cats_CR]

Many will disagree, but Null Consensus and the new emission split are exactly what legit subnets needed.

We’ve been wasting emissions on scam and useless subs for too long.

Emissions are finally going mostly to the top subnets and the new split lets owners actually control how that money gets divided.
Combined with 10x UIDs with no reg fee, that’s real support from Bittensor.
You just ain’t building shit at scale when the system keeps feeding the bottom.

Time for the real ones to push.
Stop sitting on your hands and start shipping.

Bittensor is evolving whether you like it or not.
Great updates @const_reborn and Co. 🔥

[Sheik @ttsheik]

@thinkinstant Great progress in just a week.
Let's see what's for next week.

[Cats @Cats_CR]

Remember who pointed out @inference_labs were on Bittensor just to extract.
Thankfully they never recovered price wise after that one.
Opendansor worked with those guys for a long time...

[Instant Inference (gpu/acc) @thinkinstant]

Instant distribution and development is underway.

Read our weekly report below for a snapshot of where we are, and where we are going. https://t.co/2EBUBmzFWU

[Verathos @verathos_ai]

SN96 becomes Verathos Decide on Thursday, Oct 15.

Open-weight decision models: describe a situation, ask a question, get a probability for every option, signed by the miner that answered.

Emission follows paid demand.

https://t.co/Lc14dY2YCm

[Verathos @verathos_ai]

Verathos Decide is live on Bittensor SN96: the first decision models served by Bittensor miners.

Ask a question, get a probability for every option and a signed receipt. Four models, one API, free credit to try.

https://t.co/oUo1MIbSnZ

[Instant Inference (gpu/acc) @thinkinstant]

@wolfwammer @engyai @Architect_Fi Next stop. Back to top 20!

All aboard the acceleration station 🚉 ⚡️!

[WolfWammer 🐺 @wolfwammer]

After @engyai we’ve got a second subnet providing inference on @Architect_Fi

Different subnets in maturity, but also different MC

@thinkinstant 5,5M
@engyai 63M (new ATH today 🎉)

[Instant Inference (gpu/acc) @thinkinstant]

⚡️Instant is live on Liquid Inference.

@Architect_Fi's Liquid Inference runs requests through a live auction and Instant is now in the bidding.

Instant is actively serving accelerated GLM-5.3-Flash-Uncensored-FP8.

Lightning fast uncensored inference is now available! ↓

[SEBY (gpu/acc) @sebyrubino]

100% of volume will be used for buyback -> burns.

We pivoted to offer a differentiated product in one of the fastest growing sectors of all time;

Serving Intelligence / Inference

The pieces are coming together in real time!

[const @const_reborn]

https://t.co/CFZMKRCaaI

[NIOME @NiomeAI]

We are beyond excited to announce the launch of our Decentralized Pharmacogenomic Inference Network 🧬

A live subnet designed to advance pharmacogenomic genotype-to-phenotype inference through decentralized computation. 🎉

The network challenges miners to analyze realistic, synthetic patient cases and accurately resolve star-allele diplotypes, metabolizer phenotypes, and CPIC-aligned prescribing recommendations across clinically actionable genes including CYP2D6, CYP2C19, CYP2C9, CYP3A5, DPYD, TPMT, and SLCO1B1.

With continuously generated, truth-labelled cases, multi-stage validation, and difficulty-aware scoring, the network is designed to reward accurate, robust, and well-calibrated pharmacogenomic inference.

Our goal is to build a large, diverse, and reproducible pharmacogenomic benchmark corpus while creating an open environment where decentralized intelligence can continuously improve the interpretation of complex genomic
data.

More information soon to follow with first deets available on our Discord and GitHub here: https://t.co/0HFGAHtq7E

[Pareton @Pareton_ai]

Where we’re going:
• Inference optimization — for inference providers, agent companies, and anyone running AI at volume. Make workloads cheaper and/or faster. (SN53 partnership live, more in the works.)
• Exotic setup — optimize inference on non-standard hardware or model to unlock economics that otherwise wouldn’t work. (Coming soon.)

[Pareton @Pareton_ai]

Engy is getting faster.

Pareton now tunes how @engyai runs its models. Quicker replies. Same API on your side.

Engy is one of the fastest-growing inference networks on @bittensor and beyond. A big customer with serious traffic. That is the workload we were built for.

[Alex DRocks @DrocksAlex2]

confession of the day: @fish_datura is a good person for Bittensor all things considered

how time flies and opinions change

bittensor:native

Why?

He was weight copying and still is but guess what, everyone does it so he can't be blamed for doing it back then.

He dominated mining early and got a fat stack of TAO; ok well deserved, gg. Nothing wrong with being good.

He had subnet 28 which was useless that became saygm now one of the uptrends in dtao.

He built https://t.co/6G9enz1pJZ which is a lot of work and all of it is free to use

His main subnet Lium @lium_io has become the top subnet in bittensor over time because of smart decisions made at key moments.

His team's Github code activity is pretty wild if you would consider all the private repos. Just checking at Lium subnet pace you can tell it is well organised with numbered PRs like they have an internal ticket system and updates ship fast.

gotta give credits where its due. the man did useful work. i dont care about the rest. haters are jealous and i was one of them

[Exploit Summit @ExploitSummit]

@RomAI__ of @reliquary_ai announced early access for Bittensor subnet teams at Exploit Summit. With cofounder @0xgrizzz.

With Reliquary-4B released, the next step is post-training on real customer tasks, tailoring the work to what each team's model needs to do.

[Synthdata @SynthdataCo]

Cookin'

[Conjectures.io @conjectures_io]

Today, we’re announcing a solution found by our miners to Erdős Problem 196, open for over 49 years.

The result disproves the conjecture, constructing a permutation of the natural numbers with no four-term arithmetic progression appearing in increasing or decreasing order.

Verified in Lean through Conjectures. Full proof below.

[Conjectures.io @conjectures_io]

Today, we’re announcing a solution found by our miners to Erdős Problem 96, open for over 66 years.

The result disproves the conjectured linear bound, constructing strictly convex polygons with superlinearly many unit-distance pairs.

Verified in Lean through Conjectures. Full proof below.