cosmicperspecτive

VIDA

[Algod @AlgodTrading]

Basically the thesis of Subnet 11 on Bittensor

Use skill layer + finetuning to optimise smaller opensource models to have close to SOTA performance for a fraction of the cost

[Numinous @numinous_ai]

The best forecasters on Numinous are becoming real-time market infrastructure.

The Signal Stream API is now live, giving builders and traders a WebSocket feed of market-specific signals for prediction markets.

Each signal is scored by competitively selected, Brier-winning forecasting agents from Numinous SN6, not an off-the-shelf LLM.

For every tracked market, Numinous gathers recent news and events, builds a dynamic escalation ladder specific to each market, and scores and returns the updates most likely to move the market.

In one test market, US x Iran permanent peace deal by Dec 31, 2026 (Polymarket at 68.5% YES)
The stream returned signals like:

• S4 YES, CBS report on a potential US-Iran peace deal
• S3 YES, Trump says Iran deal is in final throes
• S3 YES, NYT report on Iran considering a 15-year enrichment suspension
• S3 NO, stalled peace talks after US-Iran strikes

Each signal includes the source, timestamp, impact score, direction, and rationale.

The point is, traders should not have to manually scan every headline to understand what's happening in the market.
The Signal Stream turns raw information into scored market context.

The current version is built for demo and integration testing, with monitoring focused on curated high-interest markets.

This is exactly what mining predictive intelligence looks like in practice.

A live signal layer for prediction markets, powered by the best forecasters on Numinous.

Start here: https://t.co/omOKrbxFLY

Eversight API key: https://t.co/a4ji3lldIU

The websocket is currently in closed beta. Reach out here if interested to get access https://t.co/VPKTjqT19r.

[blockmachine @blockmachine_io]

Ethereum RPC is live on blockmachine.

RPC is one of crypto's most centralized dependencies. Wallets, dapps, bots and agents all rely on it — usually through a small number of providers.

blockmachine changes the supply side: an open network of independent operators, consensus-proof verification, and market-priced routing.

Ethereum is open for business.

https://t.co/RXUHZR9eBF

[mogmachine (ττ) @mogmachine]

Now we are cooking. 🔥
The largest demand for RPC services comes from #Ethereum devs ... time to dip into a huge customer market $ETH

[Numinous @numinous_ai]

A major upgrade is coming for miners on Subnet 6.

This Monday, we are releasing a backtestable search endpoint for Numinous miners.

Forecasting agents need more than live search.

To evaluate complex agent configurations over long horizons, miners need access to historical information exactly as it would have been available at the time.

That is the problem this endpoint solves.

A swarm of agents continuously researches our forecasting questions, finds relevant news, base rates, and contextual data, then stores that information in a queryable corpus.

This creates a historical search layer with hundreds of thousands of links available to Numinous miners.

Miners can take the code of a top agent, run it across several months of prior questions, and evaluate performance with statistical significance instead of waiting weeks for live events to resolve.

Live scoring remains the core of the subnet.

But backtesting gives miners and us a faster way to test architectures, compare configurations, and improve forecasting systems without leaking future information.

Forecasting agents need memory.

Now Numinous miners get a backtestable one.

[Robert Scoble @Scobleizer]

Last week I met with an entrepreneur who is building this into the next "Mac Mini" to run AI inferencing in the home to handle all your private stuff. Runs Hermes locally so everything stays private and so you can give your models access to your home and your finances.

More to say on this soon after he launches in a couple of weeks.

I think these will sell like hotcakes.

[𝗗𝗥𝗘𝗔𝗗 𝗕𝗢𝗡𝗚𝗢 @DreadBong0]

Subnet 9 on #bittensor just completed a 100 billion parameter pretraining run

⚪️ "We present Orion-100B: to our knowledge, the largest distributed LLM pretraining run conducted over the open internet across globally distributed infrastructure"

https://t.co/KoE063GCFJ

Orion achieved 65% the training speed of equivalent data center setups while using distributed hardware at a fraction of the cost

@IOTA_SN9 are cooking up some serious frontier shit 🔥

$TAO

[Macrocosmos @MacrocosmosAI]

Orion-100B is the first step in a larger programme, which we call Project Orion.

Having shown viability in the present case, in the coming months we intend to progressively relax the current restrictions to demonstrate the efficacy of IOTA as a more generalised, decentralised system.

That subsequent work will deliver further proof points as we progress step by step through the following:

[const @const_reborn]

It’s a cyclical economy, rather than a hierarchical one.

And that will make all the difference

[const @const_reborn]

Just orchestrated a 128 node permissionless decentralized training run, in 5 minutes, for 5 TAO, via @IOTA_SN9

They can do this up to 100B param models.
Unbelievable.

https://t.co/hJGZ6O5NrU

[TAO.com @TAO_dot_com]

Tao Wallet is now on Android.

More people on-chain.
More access to $TAO.
More ways into the subnet economy.

Download Today.

[Macrocosmos @MacrocosmosAI]

Today, we are launching the first stage of Project Orion.

Our early pre-training run of Orion-100B achieves upward of 65% of data-center training efficiency on hardware costing a fraction of the price.

Orion-100B is the first proof point for a simple idea: that underutilized compute around the world can be turned into frontier training capacity.

We believe that this work presents, for the first time, an economically compelling case for training large models using distributed approaches.

[Will Squires @WSquires]

The most important thing to understand about Orion is it closes the circle on the thesis for @IOTA_SN9.

Not just frontier scale models, but using compute acquired at a fraction of the cost, orchestrated together with MFU approaching frontier performance.

[MVTRX @mvtrx_79]

We have just perpetually locked all of our 1,269,236.79 SN79 alpha.

https://t.co/Bb3W00t1e4

[Affine @affine_io]

@FacetSir @Alibaba_Qwen We will release agent in next week. And affine agent is good at 35B param, they’re open to extend and utilize open source technical stacks

[cosmicperspecτive @AdrianLBro]

RT @affine_io: Back from the first global Qwen Conference in Singapore.

The most valuable part was the technical time with the @Alibaba_Qw…

[Ronak Malde @rronak_]

Today, @MichaelElabd, @QuantumArjun, and I are excited to announce Trajectory.

We are a research lab and product company building the platform for Continual Learning.

Our platform unlocks the signal already sitting in product usage, so companies can continuously post-train large-scale agentic models that outperform the frontier. @trajectorylabs

We’ve raised $15M from @Conviction, @BessemerVP, @radicalvcfund, @jeffdean, @drfeifei and more.

We’re partnering with some of the best AI-native companies: @ClayRunHQ @Harvey, @DecagonAI, @mercor_ai, @RogoAI to power their agentic systems, some of which we are already in production with.

We’ve brought together a world class research team from DeepMind, OpenAI, Apple, Meta Superintelligence, Amazon AGI, Scale AI, and an elite product team from Stripe and Figma.

AI will never again start on day one. Every correction, every retry, every edit will make products smarter. This is Continual Learning.

[Mark Jeffrey @markjeffrey]

I am pleased to announce that Stillcore Capital @stillcorecap has invested in IOTA @IOTA_SN9 (Bittensor $TAO Subnet 9).

[Punisher ττ @CryptoZPunisher]

Bittensor >> $TAO >> $dTAO

Subnet 11: TrajectoryRL
@TrajectoryRL
@totheagi
https://t.co/vNRqpDu7PU
https://t.co/GAs1PrXrSc

Read on @SubnetSummerTAO :

“nobody talks about SN11.”

And yet, I wrote this article more than 3 months ago.

I genuinely think it deserves an update and a much deeper analysis.

This project is serious.

@TensiaFDN , despite the current workload, this subnet needs to be added to the to-do list.
I’m sure you’re already thinking the same thing.

PS:
For those looking for articles, part of them can be found on Magellan under the topic:

“The Voice of the Subnets – Article”

➡️ Magellan https://t.co/rm698HOGPu

I’d love to do even more, but this kind of work is extremely time-consuming.

The good news is that truly strong subnets are not that numerous.
It’s probably time now to focus our time and attention on the real builders.
Less noise.
More quality.

[Platform Network @platform_tao]

Cursor turned Kimi K2.5 into Composer 2 with the right data.

March 30th, our miners start building that same data on Subnet 100. ⬇️