Crypto should be taking off in use in the AI age. Right now I have credits I paid open AI for so I can use their API. I have anthropic credits, I have some credits in a video generator app.
A2A is becoming a thing, and now it's a realistic thing to potentially expose an agent on a public endpoint to do useful work. Agents working with agents is a very clear future direction.
But everything is a closed ecosystem. But with the addition of a few extra fields to an Agent card (chain name, accepted tokens, cost) and perhaps a transaction addition to the task/send interface. We could build a network of agents working for agents. Compatible with each other. It doesn't have to be all on-chain. It's just the economic layer. You host the agent, the agent accepts payments. The agent makes payments. All the crypto tech we need already exists.
Use USDC, don't have to worry about speculation driving up costs. Use a network like SOL, AVAX, SUI and the transaction will be finalized in seconds.
Credits should be transferable, and not locked in little ecosystem.
We could take it one stop further. Prices could be dynamic. If GPU usage is high, cost should be high. If demand low, price should be low. Outages are a pricing issue, not a technical issue.
I'm tired of the "this doesn't require blockchain" argument. The alternative is a centralized entity that controls payments. Is that REALLY your preferred solution?
It's the AI companies' preferred solution. You're free to host your own agent that takes crypto, but the other agents you expect your agent to interact with will want you to buy non-transferable credits up front. Why would they join in on a decentralized system that doesn't benefit them?
Are you talking about a stablecoin crypto? Because why would I buy tokens on a chain where the value of the tokens can change wildly from day to day? How do the services that accept those tokens value them and charge appropriately?
Yes, I’d rather start a bank chartered with my own rules that leverages regulation, and only use the existence of crypto as an odorous alternative when negotiating service fees with Visa.
clearly, that is not what I said. What I said is there is a very big problem (closed ecosystem of incompatible credit systems) that can be solved using this technology that does payments really well.
Can I go to any grocery shop, check out, and pay by sticking my watch to a terminal for one second? Can I go to any website, have my browser fill in the credit card details, and click "pay"? Can I submit a dispute if I get charged incorrectly?
Because that's my bar for "does payments really well".
Sure, I can't spend my Amazon gift cards for my Warframe platinum. Yeah, Amazon could strike a deal with Digital Extremes. And maybe I could even use that platinum on Fortnite skins. But their respective deals will happen in dollars, not dogecoins.
Apparently one of the old school decentralized RPC providers (pocket network) has pivoted into being a way to access open data sources in a permissionless way. I think this is what you’re talking about?
I don’t disagree that it would be nice to be able to do this, but barring some black swan-esque event that changes how people see these things, it’ll probably always be somewhat niche compared to more permissioned or centralized systems. That or the decentralized systems will need to be VERY competitively priced to make the value-prop a bit more of a slam dunk for making this more widespread outside of philosophically-aligned hardliners and idealists
> Crypto should be taking off in use in the AI age. Right now I have credits I paid open AI for so I can use their API. I have anthropic credits, I have some credits in a video generator app.
Sounds like things are working just fine without the extra cryptocurrency step.
IMO the website is terrible. Instead of putting what actually is transmitted up and front it just tells you to use their wrapper libs as middleware with 0 reference to the wire protocol.
What HTTP headers are used? What's the body of the HTTP 402 response? The docs don't say any of that, they just say `npm install x402`.
A tired old debate devoid of facts. The chains I advocated for are proof of stake, and AI is a useful system worth the energy it uses. Should we not consume energy at all?
He wasn't ignoring anything, he was pointing out that that criticism is based on outdated information. Almost all modern blockchain designs (from the past ~10 years) do not promote energy waste via PoW.
It's fine to point out the energy waste of more traditional PoW chains (particularly Bitcoin, which accounts for the bulk of PoW), but to frame that as a problem with blockchains in general is misguided.
I wouldn't say there any major technical hurdles to bootstrapping PoS, the project just has to decide where the initial tokens should go - insiders, ICO, airdrops, etc. Many modern chains have been PoS from the start, like Cardano, Solana, Algorand, Avalanche, NEAR, etc.
Or do you mean in the sense that (some argue) the most fair way to do an initial token distribution is PoW with no premine?
A2A is becoming a thing, and now it's a realistic thing to potentially expose an agent on a public endpoint to do useful work. Agents working with agents is a very clear future direction.
But everything is a closed ecosystem. But with the addition of a few extra fields to an Agent card (chain name, accepted tokens, cost) and perhaps a transaction addition to the task/send interface. We could build a network of agents working for agents. Compatible with each other. It doesn't have to be all on-chain. It's just the economic layer. You host the agent, the agent accepts payments. The agent makes payments. All the crypto tech we need already exists.
Use USDC, don't have to worry about speculation driving up costs. Use a network like SOL, AVAX, SUI and the transaction will be finalized in seconds.
Credits should be transferable, and not locked in little ecosystem.
We could take it one stop further. Prices could be dynamic. If GPU usage is high, cost should be high. If demand low, price should be low. Outages are a pricing issue, not a technical issue.