We do a dramatic reading of TypeSafe founder Diogo's post about the intersection of Jev and coding agents, covering:

• The tyranny of the KV cache
• Dynamic context, routing, and smarter tool loading
• Skills, subagents, and background processing
• The case for an agentic package registry
• Open-source protocols over closed platforms
• Building a shared “System One” ecosystem for coding agents

Watch on X: https://x.com/OpenAgentsInc/status/2102110125083492458

Our conclusion:

"Here's what every author of a coding agent has done with this post in the last 12 hours, unless they're dumb. They've fed this into their agents and said, 'Make our agent do this stuff.'

Everyone launching products and updates based on Jev this week is all doing the same shit. We're just pointing our agents at Jev's docs and we're like, 'Use Jev but combine it with our thing' and then we're making a cool update.

The thing is, how do you trust that any of this stuff is actually good? My argument is that it all has to be open source. If your stuff's not open source, and you're just vibe-coding something today and launching it tomorrow, I'm just going to point my agent at your docs, at your site, and we're going to do our own open-source version of this.

So there needs to be some level of coordination I think because right now everyone's building the same thing and that's not sustainable. There's a lot of waste going on, and the more efficient way is for us to start building different pieces of the same overarching vision.

Here is an argument for building, number one, open source, number two, on open protocols.

...because when I say there's going to be a registry of billions of these packages, are you going to want some fucking YC company like 'I'm going to make the package registry for doopy-doop and it's going to be closed source and everyone's going to use my agent infrastructure.'

No one's going to use your fucking agent infrastructure if it's closed source. It's got to be open source.

And so if we're laying our cards on the table, let's also figure out how we can communicate with each other.

We're building a network we're calling OpenAgents, and you can contribute to that, or use our stuff to the extent that you want or not. But we're going to try to start building consensus across projects, unifying in this direction of System One systems, because they are hugely powerful in shifting things away from the big labs.

If we can together figure out how to apply these primitives, we're going to start running circles around these huge labs that really do not deserve the success it seems they're on the verge of."

👉 Stay tuned for tomorrow's episode on open protocols for coding agents. Sneak peak: https://github.com/OpenAgentsInc/openagents/tree/main/nips/openagents#why-use-nostr