Sasho

Code • Run • Learn • Slay Dragons

[On Chain Fox 🦊 @OnChainFox]

Don’t treat Gamma as just another subnet token to trade. What it’s trying to add is the payment layer Bittensor has been missing.

Most subnets still follow the same loop: produce intelligence, receive emissions, then sell those emissions into TAO. The network looks busy, but the internal economy is thin. Subnets mostly compete for weight, rank, and $TAO flow instead of buying compute, inference, and infrastructure from one another. Emissions act like trophies, not budgets.

Gamma is meant to change that. A subnet can take a slice of its emissions and turn it into transferable usage credits, spent on compute, inference, and other infrastructure. This is not supposed to be another speculative chip. It is a usage voucher. Whoever provides compute or inference can receive those credits. Subnets can buy services from each other instead of sitting in isolated prize pools. The more important design choice is that the same credits are meant to work outside Bittensor, so external compute and inference providers can accept them too. Emissions would then be able to buy real productive capacity, not just get dumped into TAO.

That changes three things.

First, the nature of emissions changes. They stop being a reward whose default path is sell pressure and become operating capital that can go straight into production. Second, the relationship between subnets changes. They are no longer only parallel competitors. A shared settlement unit makes supply and demand possible inside the network. Third, TAO’s role changes. TAO remains the base asset and value anchor, but it no longer has to be the only exit for everything a subnet earns.

That is why this is worth sharing. Another token would not matter much. If emissions can become transferable credits for compute and inference, and those credits can move across subnets and even off-network, Bittensor starts looking less like a scoreboard and more like an internal market for intelligence. The subnets that do real work are the ones that can capture that budget. That is the step from a points game to a production network.

[Angry Davee @AngryDavee]

Exploit Summit is live and Subnet Summer is streaming everything in real time

Enjoy the biggest TAO conference from the comfort of your home below 👇

https://t.co/tdT3oGwzej

[José Valim @josevalim]

One of the things I always preached about Elixir is operational simplicity and I'd say it is even more relevant now. See this tweet: https://t.co/cnZDsnQiS3

The other thing that @chris_mccord opened my eyes to is the whole runtime observability angle. If agents are going to introspect the live system, you gotta have a runtime that supports that.

Another way to frame it is: the answer to "what is the smaller/faster/more compile secure" as a single node is different to "what is the smaller/faster/more compile secure" as a system. As a system, you want fewer moving parts, more introspection, and better communication guarantees.

[Sasho @sashho]

@defigen @AlgodTrading :)

[Sasho @sashho]

@_g_x_g @dotkrueger Just An engagement farmer. He knows what's hot at any point and pokes people. It always works

[engy.ai @engyai]

Comming soon, Minimax-H3:

[Algod @AlgodTrading]

the TAM for video generation is far larger than people think

the entertainment industry itself will be using hundred billion $ or more in annual spending to create the best content

[Sasho @sashho]

Chutes price is sleeping. Go get it!

[Algod @AlgodTrading]

The TAM of these usecases will be trillions, looking forward to contribute with @engyai

[const @const_reborn]

we will need compute (@lium_io, @TargonCompute)
we will need inference (@engyai, @chutes_ai)
we will need data (@Hippius_cloud)
we will need knowledge (@teutonic)
If we are going to have free intelligence (@affine_io)

[Ditto @heydittoai]

"Expect us to ship every week. Call somebody if we didn't."

@peytonspencer

[Sasho @sashho]

@AlgodTrading @nic_carter slowing down is part of the control, but they are not slowing down, they are lying. It's a fence.

[Sasho @sashho]

This looks super coordinated. What I read is: we will not slow down, but the bleeding edge innovation models will be accessible only to a small high tier business circle and intel agencies / defense.

How else would you interpret this?

[Sam Altman @sama]

I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.

Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.

[engy.ai @engyai]

DeepSeek-V4.1-Flash is live on https://t.co/Z8NCSVOWDb.

lowest price anywhere: $0.04 in / $0.08 out per 1M tokens.

[Sasho @sashho]

@theo They pivot to inference I’m pretty sure

[Sasho @sashho]

Fable 5.1 so far is more thoughtful, but makes more dumb mistakes at the same time. AI is weird.

[engy.ai @engyai]

GLM-5.3 is live on https://t.co/Z8NCSVOWDb at ~100 tps.

$0.98 in / $3.08 out / $0.18 cache per 1M, 30% below OpenRouter and the official API ($1.40 / $4.40 / $0.26).

[Algod @AlgodTrading]

bittensor:native will have a Zec like run, just a matter of time

[Reliquary | Bittensor SN81 τ @reliquary_ai]

Reliquary weekly update

Raw base model → measurable reasoning gains in just over one day:
• Math 40.2%→71.1%
• Code 56.6%→68.3%
• Training 200→1,500 steps/day (7.5×)
• Longer, structured reasoning

Next: sandboxed multi-turn AGI environments:
https://t.co/yj7Hw79naN

[Rohan Paul @rohanpaul_ai]

AI will create jobs they said, starting here 😅