Petar Radošević

Things are moving so fast right now. I already read a lot during the week, but I still have a whole backlog of interesting articles that I work through during the weekend.

@kerckhove_ts Ah makes sense. I'm building something with SQLite right now, but it's read-only and performance has been OK. Thanks for the follow-up.

@wunki The server gets many concurrent writes, so this is a known-to-be-good use-case example. But the graph is also a bit misleading because we also stopped the traffic for a bit right before we changed over.

Three guesses when I switched this server over from SQLite to PostgreSQL

@twistin456 Plain ASCII since it's either part of markdown or in the conversation.

@wunki liveview makes building dev-ui breeze! what do you use for diagramming? - graphviz or mermaid?

Design is the work now. Cognitive debt is the newest drag on productivity when building with AI. petar.dev/posts/cognitiv…

@PreslavMihaylov It adds a smile to my face, at least once a day. So worth updating it. 😄

@wunki ohhh never realised you could configure those

Big upgrade to my Claude Code. Us spinner verbs inspired by the legendary movie Space Balls.

@wesoudshoorn I'm trying 😂 beta.finde.app

@wunki Not sure if I'm more scared about designers coding, or coders designing.

3rd rewrite of a new agentic system in 48 hours. Frustrating? Yeah, a little. But I'm learning so much. It's a relentless loop: spend hours on the design, execute fast, hit reality, learn, restart. Each pass I'm making fewer mistakes and building real intuition for how the system I'm building works best.

@twistin456 Any sufficiently complex system I build now comes with a bespoke dev UI just for observability and debugging. Being able to reason about the system matters. Being able to *see* it may matter even more. Same reason why I lean heavily on diagrams when working with agents.

@wunki lovely piece. can you share more on -- " But you can’t manage what you can’t see. That’s why I build observability in from day one. Not logging. Custom developer views, baked directly into the application."

@wesoudshoorn We are all everything now 😆

@wunki I keep saying to myself "I shouldn't be able to build this". I am a frontend dev and a designer. I have no business doing oAuth and database migrations 😅 but yes, locked in and loving it ❤️

@josevalim My vote goes to best of both as well.

Personally, when using Claude Code, I often kept messages in the "chat box" because I wanted the agent to complete the current work first. IMO, a workflow where messages are first queued and you can optionally click "send now" is probably the best of both worlds.

In Ramp's coding agent, messages sent while the agent is working are queued but sent only when the agent is done, not immediately after the next tool call. For Tidewave: should queued messages wait until the agent is done or until the next tool call? Reply with your rationale.

@wesoudshoorn I was already addicted to building, but I haven't been this locked in since a long time. The things that stopped me from shipping in the past are gone now.

@wunki Isn't it fun?

Let me know what you think and how we could improve them. PRs or comments are welcome. Let's share best practices so we can use LiveView and all its benefits without compromising on the user experience. #myelixirstatus

Second is working with forms, where you have to make sure that forms don't get wiped on reconnect, or when and how to use debounce to make sure you don't spam the backend. github.com/wunki/amplify/…

I created two skills which I always found to be gotcha's when working with #elixirlang and LiveView. Both can have a big impact on user experience. First, LiveView and optimistic updates, making sure that the UI stays responsive, even when there is latency. github.com/wunki/amplify/…

BEAM anyone?

I coded myself into a corner on an ambitious project. Right now, I spend hours on the design of the system and optimize for systems that are easy to reason about and have observability baked in from the start. I also make sure to have a docs directory with relevant project level documentation, also leaning on diagrams a lot.

Short musings on "cognitive debt" - I'm seeing this in my own work, where excessive unreviewed AI-generated code leads me to lose a firm mental model of what I've built, which then makes it harder to confidently make future decisions simonwillison.net/2026/Feb/15/co…

@josevalim Happy for you! We are also happy customers here 😃

Almost 5 months since launch, Tidewave has crossed 100k ARR! 🎊 The best part is that we can invest all profits back into the product, since we are not reselling/subsidizing tokens. Thanks to everyone who is using it and for all of the feedback!

@thorstenball Great post! I would add a fourth fallacy "Context Degradation". New task, fresh context.

Great prompt examples by Harry here. If we're getting to a point where agents can build anything you can specify, it pays off to be specific.