Great to see Eve hit v1.0 , congrats to the @eve team!!!
We have been running eve in production since v0.3, moving Dagny from our legacy agent loop to Eve back in early August - primarily for durable sessions and because most of our infra was already on Vercel.
Since then things have moved pretty fast, and almost everything in our production now runs on Eve -
> durable sessions, turns & resumable streaming
> multi-tenant agents, workspace with tools
>human in the loop ( both on our web and slack version of the product )
> mcp connection, integrations
ofcourse,vercel ecosystem comes along - schedules, agent traces, sandboxes, deployments
Great shipping, Eve team!
Also shoutout to @shardara for all the quick help and fixes in those early days , lets gooo
eve is generally available with v1.0.
• Build agents on your own harness
• Ship multi-tenant, multiplayer agents
• Govern & trace sessions with Agent Runs
• Render agent UIs via Next.js, Nuxt & more
Read the blog → vercel.com/blog/eve-is-now-g…
Avra retweeted
This is how agents will ship all code.
→ Agent writes feature-gated code
→ Verification gates on push (proof checking, security audits)
→ Agent monitors metrics
You can live the future today with Vercel CLI: 𝚟𝚎𝚛𝚌𝚎𝚕 𝚏𝚕𝚊𝚐𝚜, 𝚟𝚎𝚛𝚌𝚎𝚕 𝚖𝚎𝚝𝚛𝚒𝚌𝚜, 𝚟𝚎𝚛𝚌𝚎𝚕 𝚍𝚎𝚙𝚕𝚘𝚢.
I haven't touched localhost in month.
I ask AI to make the feature only visible to me in production, so I can assess and iterate. If it's good, I release it to users, and ask the agent to set hourly/daily reminders to check logs and database to fix any issues. My AGENTS·md also has rules to force the agent to test everything.
Now AI is so good, everything goes straight from prompt to prod.
The blank chat box is not the future of AI.
Introducing Dagny: The Next UI for AI.
A super agent with a company brain that knows your goals and generates apps, widgets, and interfaces on the fly.
Moving from the CLI era to the Macintosh era of AI.
Try it for free at bydagny.com
Since shipping Dagny a few months ago, we've had:
>200+ users using it
>60% month-over-month growth
And most importantly, a ton of learnings from our customers
Biggest takeaway?
A blank chat box or an MCP works for technical people. But for everyone else, AI needs a more intuitive interface.
It's time to move from the CLI era to the Macintosh era of AI.
We're building the next UI for AI. For the 99%.
We are killing ChatGPT
AI has a UI-problem. we're stuck in the CLI-era
The information density in chat is terrible. That’s why we're going ALL IN on generative user interfaces
Today we are launching the new Dagny
the next UI for AI
3
8
307
stop planning. start building
if you’re a future thinker like me, it’s easy to get trapped chasing the “perfect” plan for weeks, months, or even years ahead
but that’s a slippery slope, planning time keeps expanding instead of failing fast, iterating, and learning
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Why we stopped using RAG in our product.
We tried both RAG and OKF.
So far, here's what we learned:
Not every AI agent needs a vector database.
Some just need a filesystem.
1/ Models have changed
> RAG made sense when context windows were small and tool use was limited
> today, give an agent a VM, a filesystem and a few basic tools
> it can grep, read entire files and follow references on its own
2/ Treat company knowledge like a codebase
> markdown files, one concept per fileKeep sources, tags and links
> let the agent search and decide what it needs to read
> don't overcomplicate retrieval
3/ Most businesses dont have millions of documents
> for a few hundred or even thousands, a filesystem can work surprisingly well
> no separate vector index to maintain
> vector search still has its place at larger scales
4/ Make knowledge useful beyond retrieval
> humans and agents can read the same files
> agents can maintain notes and playbooks
> changes can be validated before theyre saved
Most of these ^ learnings came from building Dagny as a self-serve product for smaller service businesses.
Over the last few weeks, we have started bringing the same product into larger businesses, deploying it in their own environments and working closely with their teams, while keeping the core agentic architecture intact.
Dagny started as a general-purpose AI agent to run our company.
But over the last few months, we've realized - some of its most useful work happens without us even asking.
We use it to keep track of what's happening across our company, flag things we might have missed and tell us what needs attention.
And sometimes, it catches things before any of us even think to check.















