Echo

founder @CortexLM

[Echo @echo_cortex]

@yubrew however, private endpoints would limit miners to fewer models due to hosting costs.

openrouter still seems like the best option if owners handle the requests themselves while miners pay for usage. this removes the prompt injection risk since owners fully control the OR account.

[Echo @echo_cortex]

@yubrew maybe using private endpoints through lium or targon would make more sense.

with openrouter, there’s a prompt injection risk since they have an option to inject prompts into every request without us being able to see it. this could allow agents to cheat.

[Echo @echo_cortex]

@lium_io how

[Echo @echo_cortex]

@LamidaGlobal we have areas where we need to improve, and we will share our concrete plan very soon for the future of the subnet. I will also address the management, the shortcomings of our previous management, and how we intend to fix them.

[Echo @echo_cortex]

@LamidaGlobal Feel free to share your feedback with us; we will improve 🤝

[const @const_reborn]

https://t.co/IdB9DPr5u0

[Openτensor Foundaτion @opentensor]

Next time we train 1 trillion

[Echo @echo_cortex]

the wake-up was rough; i wish all of this had just been a nightmare

[covenant @covenant_ai]

https://t.co/06wplkr0GY

[Echo @echo_cortex]

we’re making great progress on cortex. we’ll have some exciting news very soon🤭
the work of the miners on subnet 100 will pay off

[Echo @echo_cortex]

@Rapido_ai @Noblissefr C’est parce que tu as parlé aux investisseurs dans tes messages ce qui est interdit dans les règles du discord, pas pour tes autres messages.

[templar @tplr_ai]

Permissionless miners on Templar's Crusades pushed a 7B model on 4xA100 80GB SXM from 66% MFU to 74.5% MFU.

68% is already a strong A100 result in public benchmarks. What makes this exciting is the gain came from miners optimizing for the specific model and setup.

1/4

[Sam Dare @DistStateAndMe]

Clankers building clankers

[Sam Dare @DistStateAndMe]

When you fix one bottleneck, the next one becomes visible.

At @covenant_ai we built PULSE (https://t.co/Yby0rKzGSY) to make weight sync 100× faster. That worked. Then the trainer itself became the new ceiling.

So @erfan_mhi ran autoresearch on our GRPO trainer. 27% → 47% MFU. 16.7 min → 9.2 min per epoch. 1.8× faster on a single B200.

Decentralized post-training, closing the gap with centralized.

https://t.co/ATsnSgs5Qd

[Punisher ττ @CryptoZPunisher]

#Biττensor >> ∆ τ << #τₐcc
> $TAO <

Subnet 100: Plaτform >> @platform_tao

Let's take a look at Atlas, the bot that manages the subnet

1/7

Bittensor is evolving at an incredible pace.

As the network transforms, the community must also learn to adapt and evolve alongside it.

Today I’m testing a new way of sharing content:

a direct dialogue with Atlas, the bot from the Subnet 100 (Platform Network) Discord.

After the spectacular launch of Subnet 97 and its bot Arbos, it feels almost natural that SN100 would be next to experiment with this kind of interface with the community.

In this thread, discover how Atlas itself presents Subnet 100.

[Echo @echo_cortex]

an excellent way to optimize your time.

[Echo @echo_cortex]

People don’t realize what Bittensor makes possible.

[templar @tplr_ai]

We just completed the largest decentralised LLM pre-training run in history: Covenant-72B. Permissionless, on Bittensor subnet 3.

72B parameters. ~1.1T tokens. Commodity internet. No centralized cluster. No whitelist. Anyone with GPUs could join or leave freely.

1/n

[Echo @echo_cortex]

@wolfwammer @CryptoZPunisher we should be able to start focusing on bringing the products into production by the end of the month, although some products may be ready earlier. Regarding the commercial risk and adoption uncertainty, I would of course be happy to discuss my plan in more detail.

[Punisher ττ @CryptoZPunisher]

https://t.co/YhVcRZGg1F