For example, assuming an API behaves a certain way, misunderstanding an existing utility/dependency, or getting a business rule wrong.
How do you currently catch these assumptions before the agent makes changes?
I'm specifically interested in the cases where the agent sounds completely confident but is actually wrong.
For example, assuming an API behaves a certain way, misunderstanding an existing utility/dependency, or getting a business rule wrong.
How do you currently catch these assumptions before the agent makes changes?
I'm specifically interested in the cases where the agent sounds completely confident but is actually wrong.
For example, assuming an API behaves a certain way, misunderstanding an existing utility/dependency, or getting a business rule wrong.
How do you currently catch these assumptions before the agent makes changes?
I'm specifically interested in the cases where the agent sounds completely confident but is actually wrong.
For example, assuming an API behaves a certain way, misunderstanding an existing utility/dependency, or getting a business rule wrong.
How do you currently catch these assumptions before the agent makes changes?
I'm specifically interested in the cases where the agent sounds completely confident but is actually wrong.
Cursor, Claude Code, Codex, Copilot, etc. can handle a lot now, but I'm curious what developers still find unnecessarily manual.
Could be anything—context, debugging, reviewing generated code, managing prompts/rules, testing, documentation, switching between tools, whatever.
What's something you find yourself doing repeatedly that you wish the AI/tool handled better?
When I’m debugging or building something from a webpage, I sometimes end up jumping between the browser, DevTools, screenshots, console errors, source code, and my AI coding assistant.
I’m curious how other developers handle this.
Do you have a workflow or extension that makes it easy to send the right context from a webpage to Cursor, Claude Code, Copilot, ChatGPT, etc.?
What still feels unnecessarily manual or annoying?
When I’m debugging or building something from a webpage, I sometimes end up jumping between the browser, DevTools, screenshots, console errors, source code, and my AI coding assistant.
I’m curious how other developers handle this.
Do you have a workflow or extension that makes it easy to send the right context from a webpage to Cursor, Claude Code, Copilot, ChatGPT, etc.?
What still feels unnecessarily manual or annoying?
When you’re debugging a website, how do you get things like the selected element, DOM/CSS, console errors, screenshots, network errors, etc. into your AI coding workflow?
Do you use an extension/tool for this, or do you still copy/paste everything manually?
If you’ve tried a tool, what do you wish it did better?
When I’m debugging or building something from a webpage, I sometimes end up jumping between the browser, DevTools, screenshots, console errors, source code, and my AI coding assistant.
I’m curious how other developers handle this.
Do you have a workflow or extension that makes it easy to send the right context from a webpage to Cursor, Claude Code, Copilot, ChatGPT, etc.?
What still feels unnecessarily manual or annoying?
I’m curious how people handle this when the task isn’t something the AI can solve from one or two files.
Do you manually gather relevant files, GitHub issues/PRs, docs, Slack discussions, logs, etc., and then give that context to the agent?
Or have you found a better workflow?
I’m particularly interested in what you do when the project is large and the AI keeps missing important context.
I’m curious how people handle this when the task isn’t something the AI can solve from one or two files.
Do you manually gather relevant files, GitHub issues/PRs, docs, Slack discussions, logs, etc., and then give that context to the agent?
Or have you found a better workflow?
I’m particularly interested in what you do when the project is large and the AI keeps missing important context.
I’m about to start building my first SaaS and want to avoid the common beginner mistakes.
I’ve spent time researching problems, talking to people, and learning what makes a product worth paying for, but I know there’s a lot I don’t know yet.
If you’ve built a SaaS (successful or not), I’d love to hear: