PWV Portfolio Feed

The diagrams below show the "network" of nodes/edges/channels involved to make the Chrome extension work for LiveStore.

#ZedEditor

Lastly, you can now drag and drop files onto the project panel when working in remote sessions to copy files from your local machine to the remote server.

Vim users can now customize the cursor shape for normal, insert, replace, and visual modes in Zed.

{
  "vim_mode": true,
  "vim": {
    "cursor_shape": {
      "normal": "hollow",
      "insert": "bar",
      "replace": "block",
      "visual": "underline"
    }
  }
}

We added support for Qwen3 in assistant text threads.

The Git branch picker in Zed now supports remote branches.

🚀 Zed v0.184 is out!

We're working to improve the accuracy of our autocompletions. Starting with today's release, you should start to see more relevant suggestions prioritized near the top, and we will continue to improve our completion matching going forward.

We've seen how products like GitHub, Vercel and Supabase have changed how we build apps.  

Now, we're doing the same for AI. 

Making it accessible to every builder, every developer, not just ML experts.

3. Why now?  

The foundation models are ready. 

The tools are maturing. 

But the glue that connects everything is still missing.  

LLMs are evolving fast—like, literally every week. New standards pop up (looking at you, MCP), and APIs change faster than you can keep track.

The result?  

Most AI agent projects never make it to production. 

Those that do require constant maintenance and lack the reliability users expect.  (new LLMs every week)

AI agents should be composable, reliable, and production-ready from day one. (minute one?)

2. What problem are we solving?  

Every company is trying to build AI agents, but the infrastructure is missing. Takes too long. It's too expensive. Hire MLEs?

Teams are reinventing the wheel, building brittle systems that break in production + can't scale with their needs.

Chai removes all the complexity of: 

• LLM orchestration (unified API)
• Tool integration (deployment and scale)
• Memory (long term memory Auto RAG)
• Deploy agents, and scale to millions
• Langbase processes 100M+/mo agent runs

With Chai, what used to take days/weeks — now only takes minutes.

From idea → agent → deployed!! 

We wanted to make it super easy for everyone to building their AI agents.

People spend weeks wiring together LLMs, tools, and memory systems just to get something basic working.

1. Why are we doing this?  

Everyone is building an AI agent today, but it's hard.

Cursor, perplexity, v0, chai, lovable, bolt — what do they all have in common? 

They weren’t built on AI frameworks—they're built using primitives optimized for speed, scale, and flexibility.

Chai is built on AI primitives by Langbase: 

• Agents - for reasoning and planning 
• Tools - for taking actions in the world 
• Memory - for human-like context & learning 
• Workflows - for orchestrating complex tasks  

All with type safety and production reliability.

Chai vibe coded a mini Perplexity for me 🤯

Unbelievably good. Can't wait to see what y'all will build. Our beta testers had multiple 🤯🤯🤯 moments.

Let's go! 👊

Vibe coded a Receipt OCR Agent

Give this agent a receipt image and it can analyze whatever you want in it. 

Woohoo! 🥳🥳🥳