Add a toggle that overlays the release of major models kthxbai

Structurally, what is/can be offered by blockchain to some those problems though? It can be used for settlement of unrestricted inference but I actually don't forsee an issue with achieving the same offchain

Would love to prototype and see if the input values matches the form I've landed on over the years

Human ability to re-implement existing things under a new name and get excited about it undefeated

It's basically a consultant. Example from my old life: - I was a specialist in a particular - I would be embedded into a company looking to buy - I would help them design, implement and switch over to from Now do the above, but for things like Lighter. They deploy a technically savvy Lighter employee into Robinhood working groups to help design and implement Lighter (for Robinhood)

I was only able to use it for some minor contract audits before this. Sad.

Yea, you connect via Plaid and it gives the permissions

Read-only access for auditability. Single highest agent-related ROI thing for me right now is proactive awareness of all the admin shit I had to do manually. Tracking spend across credit cards / accounts, monitoring emails, proactive agenda planning and reminders. I've found that giving broader read-only access to my life has allowed for seemingly disconnected things to be connected and better maintained. Being able to read my accounts and check for fraud, inconsistency and poor habits has been one area I've enjoyed.

I don't think this proves that routers are a bad/non-useful product, although the examples are great at showcasing where the faults can lay. It just tells me that there is more opportunity in orienting an additional/adjacent product around attributable outcomes

It's a fun riff tbh. Tying that many seemingly disconnected things together is a fun thought experiment

Two problems/questions: 1. Why is bitcoin always the recipient of excess energy? We're already seeing deviation from that miners rolling off to AI inference, Pearl's PouW, etc. 2. There is a race between end of block rewards and creation of fusion, and they could very well not overlap. This would then put the only reward at block fees of a chain that is used very (very) little.

Que?

Tldr freemium model I'm generally of the same belief, and know there are whole hosts of SaaS founders who are waiting for their lunch to be eaten by labs, but an still semi-skeptical that the frontier labs can actually own and manage a huge array of user applications that nullify the competition which is solely focused on their verticals. Maybe there are a handful of obvious apps with bloated teams and outdated builds, but replacing those won't be worth trillions and I don't subscribe to labs swallowing the application world in it's entirety.

Really was great. Again think Jason is one of the better interviewers in this space, both for locating guests and extracting a quality conversation

GoogleFi. Haven't had to think about it for years and have circumnavigated the world a few times. Only becomes an issue is you're out of the country for > 90d

My counter is that despite having higher demand their moat is lesser. All incentives points towards the commodity of intelligence being distilled, cheapened and pulled internal whereas with blockspace that deletes its value prop. An private, internal blockchain is DoA because it's benefit is being shared amongst the world in a permissionless fashion. For intelligence, it's expressly the opposite. It is BETTER as an offering if it's private and internal (cheaper, fewer leaked secrets, can have params tweaked, etc)

I think it's the opposite for the "good enough" intelligence. The vast majority of tasks needing AI can be handled by OS models and/or free-tier frontier labs. Like the Google search result returns. and it'll only get more abundant / models more efficient at producing intelligence for every unit of compute. The edge blockchains have going for them is that they benefit from being public/permissionless, so there is a draw to these open/common systems whereas intelligence is expressly the opposite. Everyone benefits even more from bringing the models in-house. Cheaper, private, etc

Still very much of the belief that frontier labs are the same trade as L1s and that without building an actual product atop their offering (blockspace/intelligence) which captures fees that they're going to experience the same over-priced charts. The one key difference is that labs are sprinting towards AGI, which if achieved wipes out all "bad business" conversations because you've unlocked god mode.

It's just raw capacity. HyperEVM can't handle the literal processing throughput

They use the oracle data from hypercore. Not a true restrictor because it's ~same as using any other oracle imo but it has a good story element. It could exist feasibly on most chains