Ryan Carson

@adamse yeah, i've been thinking about that. why hasn't someone figured all this out already? :)

@ryancarson you are going to need an enterprise MCP layer for those agents

100% of dev is going to be done in sandboxes in the cloud, controlled by kanban boards. Trust me, I love my local machine and gorgeous mac apps, but all of it is just a terrible form factor for running a team of agents effectively.

@heylegacyguy @_colemurray @modal Needed cloud sandboxes, LLM choice and less token appetite

@ryancarson @_colemurray @modal Gave up on Symphony?

Got Open Inspect setup by @_colemurray and it's showing a lot of promise. It's an OSS version of Ramp's Inspect agent. Callable from Slack. So far the @modal sandboxes are fast and fab. github.com/ColeMurray/bac…

@MatijaSosic I was there for that. My talented team at Treehouse actually created some of the first cloud-based development environments for our online school. Fast forward twelve years and we're now in an agent first world where cloud is really the only thing that makes sense.

@ryancarson I remember we had this discussion when online IDEs first appeared. Also no-code tools.

@can @modal @_colemurray This made my day

@ryancarson @modal @_colemurray our legal team did not ok this

@jessmartin @_colemurray @modal let us know how it goes

@ryancarson @_colemurray @modal I had my symphony-ts build support for remote execution ("runners", in symphony terms) just last night. looking forward to trying this out!

@modal @_colemurray it's a good pairing :)

@ryancarson @_colemurray Fast and fab 😎💚

@FieroTy @OpenAI Excititing!

Hi twitter friends, I’m joining @OpenAI to work on applied evals! A few years ago my day job was studying microbes Somehow this feels like a natural progression 😄

@_colemurray :)

OpenInspect: VSCode Code Server In-browser OpenInspect now contains a hosted VSCode instance which runs inside of the sandbox. You can now make manual changes without having to clone the repo locally

@Teknium I agree lol :)

Do you agree? Where do you do most of your dev work? I personally have a linux server machine in my house that I run hermes-agent on and do my dev work from. How bout you?

@jakubhasekk It changes so damn fast; that's the problem.

@ryancarson Are you ever planning a full overview of symphony and your setup? Would love to see!

Okay, explored the codebase more, and this is not a replacement for Symphony. It lacks the orchestration ability to shepherd issues from to-do to merged. I think that final version is something like Open SWE + Symphony.

@sumitdotml He's not joining the company; he's joining the board. Two very different things

who am I to even tell people what to do & what’s right or wrong but Mitchell Hashimoto announcing that he’s joining that company has disappointed me a lot

@ctatedev I assume you're using Vercel Sandboxes for this? Would love for you to see how you're running your setup.

~100% of my dev is done in sandboxes in the cloud Highly recommend it: - Unlimited parallel agent sessions - My local machine stays safe - Can work from anywhere - Can close laptop - Lap stays cool Interesting idea to visualize with Kanban

@michiels I disagree on both counts: The majority of work is not going to be pair programming or closely monitoring agents. It will be setting and forgetting with the occasional check-in. Therefore, speed doesn't matter. Cost is relative - it's de minimis compared to human engineers.

@ryancarson Disagree. Cloud agents are relatively slow and super expensive. I think the middle ground will be agents running on dedicated hardware in your team. The right apps will manage those agents. Semi-local and persistent machines that run sandboxes are faster.

@kubedoll That does sound awesome, but I think for building out teams there needs to be a cloud-first solution.

@ryancarson You're correct about kanban, but not about the cloud. I run mini racks of mini ITX systems, each with 32gb ram and 2t8c ryzens. they're dirty cheap, cost me a lightbulb and last forever. and when they don't? crack em open for precious 16gb dimms.

@SieversJosua I think you're trying to hold the reins too tightly.

I thought about your linear symphony setup a bit the last days. I think it totally makes sense for bigger production systems. Personally though, I found myself even disliking the workflow of local branch -> PR -> greptile review -> fixes -> another greptile rev -> merge because it was such a waiting game. Could totally be an issue with my workflows, but “bare metal vibe coding” (Claude code, codex, mostly main branch, one feature at a time, local preview) with its speed etc is the most satisfying way for me currently.

@mitchellh Love hearing this. We're built on Vercel so obv a fan :)

Excited to share that I've joined Vercel's Board of Directors. Vercel is made up of builders and tastemakers that continually ship things that deeply impact how developers work: Next.js, AI SDK, v0, etc. I can't think of a more exciting place to be. Let's fucking ship. ▲ My relationship with Vercel goes back to the earliest days. HashiCorp was an early adopter of NextJS and Vercel (~10 years ago!) and it remains my default tech stack and deployment platform to this day. Ghostty's website is all on Vercel, too! Beyond that, I've been continually impressed with the teams relentless focus on shipping meaningful software. And importantly, software that has incredible taste. Now we are in the age of agentic software development. Vercel is building agentic infrastructure that I think every app and agent will need (I certainly need it!) and I can't think of a more exciting place to be. Huge thanks to @rauchg , Jeanne, Marten, @cramforce, @tomocchino and the entire Vercel team for the warm welcome. Time to work.

@ashleyhindle :)

Coming soon 🫶

@steve_ike_ @linear @cursor_ai I hope so. This is the most obvious DX to me - can't believe it's not live yet

@ryancarson @linear This feels like the natural evolution of @cursor_ai

The company that actually builds the agent-first code factory is going to be worth hundreds of billions of dollars. No one has cracked it yet. It can't be the model labs because then you're tied to one model. I'm hoping a company like @linear will do this. I'd happily pay thousands of dollars a month for that (+ the token cost). Basically, we need SDLC 2.0 for the agent age. (Also, the right solution can't rely on gh - we need whoever does this to completely replace it as well.)