I have recently seen platforms, including some CRM systems, that operate almost entirely through chat and MCP connections rather than traditional menus, dashboards and settings.

It made me question how software platforms should be built going forward.

Hey Hunters, I know that user feedback is very important when building a product, especially for startups. The problem is, I m still figuring out the best way to collect it effectively.

Right now, I m debating between:

Email: Sending out direct emails to users. But how should I structure them? Short & casual or more structured surveys?
In-App: Using a widget or pop-up inside the app. But where and when is the best moment to ask?
Feedback Tools: There are so many. Which ones are actually worth it?

Since I develop alone, all feature requests come to me with equal weight. In the past week, I received three: dark mode, CSV export, and Slack integration.

So far, my way has been to develop whatever the nicest user wanted. Bad system, obviously. These days, I use frequency of the request stated in other words as my measure rather than its volume.

I was dealing with an issue where automated scripts kept trying to "optimize" small stuff instead of just finishing the main task. You ask for a simple feature update, and suddenly it's changing file structures and refactoring background logic nobody asked for.

It feels like the more context you give them, the easier it is for them to get distracted by non essential tasks.

For people building with cursor or agent setups daily how do you look down scope so it stays focused on a single file or task?

I ve been testing AI coding agents on React projects, especially for issues involving state management, broken API calls, and unexpected UI behavior. They can identify the problem quickly, but I still find myself checking whether the fix actually addresses the root cause. How reliable have AI agents been for you when debugging React or Next.js projects?

I explicitly asked an agent to write a lightweight helper function for formatting dates. Instead of using built-in JS utilities or our existing setup, it tried to npm install a obscure library just to parse a string.

It feels like agents often default to importing heavy external package for trivial logic rather than writing 5 lines of clean native code.

What s your strategy for enforcing a zero new dependencies policy without micro-managing every single prompt?

@Paste just launched their latest version, and tbh I'm considering switching over but I'm curious to hear what other people use?
I know @Raycast has a pretty powerful and integrated one that I'm considering using. I currently use @ Pasta which I love.
Do folks have any they recommend? What's the consensus?

Yesterday, @zaczuo shared an idea about delivering packages from space. To me, that seems quite sci-fi and financially demanding to actually pull off. Today, I m reading that Bezos predicts millions of people could be living in space by 2050.

How realistic is this scenario, given that the last time we set foot on the Moon was more than 50 years ago? (And above all, I feel like we still can t solve basic problems on Earth, let alone expand into space.)