Jon Morehouse

Today I'm excited to share that Nuon is officially open source! You can read more about our announcement here: nuon.co/blog/oss-annou…

I'll cap it at 10, but I could go on for ages. Packaging. It's funny, I often say this is not our value prop. It's a commodity problem, but it's still hard. Ultimately, you are packaging your entire app, and that's hard. We have a lot of fun stories with our BYOC option. Some more painful than others.

9. Previews: every change you are making has to have previews. Think, terraform-cloud but for every single part of your app. The app has to be templated, so you need to be able to see what's changing accordingly.

At Nuon, we eat our own dogfood. We run Nuon as a BYOC service in our customer's accounts. It's not easy, but it is insightful and makes us better. Here are some of the hardest challenges we see, which you might see as well when offering your BYOC:

@jonathangrahl Dang it, we just converted yday!

I mean, so now Bun is just slopware? Guarantee you no one has actually read and understood the rewrite. Insane. Wonder how many unsafes there's is

@fatih @cursor_ai can you drive it via amp? would be a killer upgrade if so

If you haven’t tried it out yet, @cursor_ai’s CLI is significantly better than it was a few months ago. Top of mind they improved these: * ctrl-g opens vim * statusline * fixed terminal rendering with light themes * permission is better and less restrictive * plan and debug mode work as expected What I think could be nice to have,and improved: * let’s handoff between the GUI and Terminal. If both are streaming, all it takes is to show the state * a view to manage multiple sessions, like the sidebar in the ui * recap feature, it’s really good in Claude * I think it autocompacts, but hard to understand when and why. Maybe it’s better not to share a lot of detail, dunno about this. * pull in bugbot reviews automatically * open bugbot comments in CLI * native integration with Graphite (I mean you own it, go bonkers here, why not?) Anyway, it’s way better. I wanted to use it for a long time but couldn’t because it wasn’t on par with others. It’s now pretty good and I use more and more often I even have two aliases, one is called agentc opens with Composer 2, for small/medium tasks, and the other is called agents which I for hard task that uses GPT 5.5.

the real ones know it's both

so, do you build the durable workflow orchestration on top of the queue, or the queue on top of the durable workflow engine?

@fatih @PlanetScale I believe this is going to lead to more deterministic simulation testing and in-house built test harnesses.

I was a huge unit test supporter, but honestly, it's no longer worth it. Agents are superb at writing extremely bad unit tests, and they still look good on paper. We're also shifting slightly to more and more e2e tests at @PlanetScale. Luckily with agents, that shift is also manageable.

probably worthy of a blog post, next

pretty difficult systems problem. Especially when you consider that any update in those queues could have different approval steps, some updates might have a timeout of seconds, others in years. not sure how you could build this on top of anything but temporal 🤷‍♂️

but also, do as much work eagerly as possible and (safely). Then, add in policy evals, drift-detection, break-glass and this becomes super powerful

we built a layer on top of temporal that let's us do this. We now get fine-grained concurrency controls for every node of that "graph". what's neat about this is it means you have the right safe-gaps in place (don't try to do two competing updates on a resource at the same time)

this leads to some pretty neat challenges. You want to parallelize what you can. Sequential everything is safest (but slowest) How do you handle concurrency at the app, version, environment, service, resource level within that, while maximizing performance?