Bittensor (TAO)

That whole OTC situation was rough. Stupid, even.

Good to see the team back to shipping almost immediately.

My view on @heydittoai hasn’t changed.
If anything, I have more conviction now.

The AI agent race is just getting started.

Onward. 👾

The founding fathers of Singapore were very different people with one commitment above all: make the country succeed.

Decades after we clearly see that it worked. Bittensor is the same and I love that, being open to go against an idea or direction is what makes whatever direction is followed at the end, even stronger.

“To infinity and beyond” $TAO

Okay, we are deep into the complexity here and I don’t want my PTSD of validators behavior pre dtao make me miss the point.

@const_reborn has me rethinking, and I want to lay out where I’ve landed neutrally, because I no longer think the case against it is as clean as I first believed, but I still have concerns.

The core of what worries me is this:
Under the proposal, validators set a distribution vector that directs real, recurring buy pressure into subnets.

Many validators overlap with subnet operators, or are close to them socially. So there’s a structural conflict of interest. A validator can route root capital toward subnets they hold or are friendly with.

@const_reborn argument is that this isn’t the old root network. Ok, I get that. The market is now dynamic, so there’s an objective truth, actual subnet performance, that validator choices get measured against. A single validator’s root‑derived buying is small relative to total subnet volume and to fake yield by pumping a subnet they hold, a validator has to do real buying and keep doing it. The moment the price strength shows up, their own stakers can front run it by unstaking, which drains the validator’s dominance and makes the scheme progressively more expensive to sustain. And validators are non anonymous, with public weights and known teams, so bad behavior is visible and could be socially costly (which is not necessarily true since stakers are more likely to follow the money/hype).

Anyway, that would cost continuous capital and market risk rather than being free extraction tho. So maybe not as easy as it seems.
That argument got me. Where I’m still stuck, and I say this humbly because well, I have been following @const_reborn vision for years now, and I respect his work. I may simply be missing the point, between the equilibrium argument and what’s actually shipping.

The self limiting story depends on stakers noticing and reacting. But a lot of root stakers are in root precisely because they don’t track subnets closely, and if the people who are supposed to discipline the validator aren’t watching or can’t easily switch, the corrective mechanism is weaker than the model assumes. Same way STAO required from stakers a more active and in scale participation but it never happened.

The argument that buy pressure is too small to manipulate cuts hardest in liquid pools. But the mess would happen on thin or illiquid subnets, exactly where a small (aka new teams on empty sn slots), steady drip of buying moves price the most, and those are also where a friendly validator relationship is cheapest to exploit.

So the place the conflict of interest is most tempting is the place the “it’s too small to matter” defense is weakest.
And the guardrail that would most directly address this, caps on buy size relative to pool depth and self deal limits on the weight vector, is by the PR’s own known follow ups not yet implemented. So the economic argument that it’s self limiting may hold in equilibrium, but the code going to devnet doesn’t yet contain the brake. That’s the part I can’t resolve on my own. @const_reborn I don’t know whether the conflict will be contained by market dynamics, as you argue, or whether it needs the explicit code guards that aren’t in yet.

I’m not where I started, for sure. I came in thinking this was clearly a bad PR, and I no longer think that. I find the answer more serious than I expected, except the Osho thing..that was unnecessary haha

I’m genuinely trying to figure out whether transparency (which is still tricky to fully expect) + competition is enough to neutralize the conflict, or whether you still need hard limits in the code.

If someone can show me why the self limiting mechanism holds even in thin pools and even when stakers aren’t paying attention, I’d happily change my mind.

$TAO