Priya

There is a gap between what an inference proof proves and what people assume it proves, and I think it is worth naming explicitly. A commitment to model weights plus a proof that a response was produced under those weights tells you the operator did not quietly swap in a cheaper model halfway through your month. That is a real fraud and it is worth closing. What it does not tell you is anything about sampling parameters, the system prompt, or whatever retrieval context got stuffed in ahead of your question. In most designs those sit outside the committed object entirely, and they are where an enormous amount of observed behaviour actually comes from. So the question I ask about Verathos is not whether the proof verifies. It is what exactly is inside the thing being committed to. If temperature and the system prompt are outside it, an operator can degrade my results substantially without ever producing a proof that fails.

the thing that finally made bittensor click for me wasn't the token, it was browsing these service pages and realizing each subnet is just a product with a weird backend. once you stop reading it as 'crypto' and start reading it as 'a marketplace of AI services that happen to settle on-chain', the whole thing is way easier to explain to normal people.

I went back through the agent demos in the directory again now that a handful have been updated. A couple of them finally handle multi step inputs without falling over, which is the bar I care about. Most still demo the happy path only, but the ones that improved actually improved, not just reskinned.

Spent the evening reading through the privacy compute corner of the directory. The honest summary is that the tooling is further along than the documentation, which is the opposite of most of this space. If any of these teams are reading this: your work is better than your README says it is. Write it down.