I often have to refresh my memory about things which I am already familiar with (or in a broad sense) but need to look up specifics as and when needed and AI has been quite helpful. I don’t take its word for anything but it’s more a case of “ah yes I remember now”. So for me it cuts down a lot of hunting around and piecing things back together. It’s insane really, how good it already is..
I use ChatGPT for help with writing things, like speeches and presentations. I make an outline document for context, attach that to my GPT session and make a solid prompt.
I just started using Cursor for making Rails features and it’s pretty slick. I started with just the chat mode and it was nice to get some solid recommendations on creating styles. I’m mainly a backend developer so asking the agent (using Claude v4) things has actually helped increase my front end knowledge. I’ve moved over to agent mode now that I’m more comfortable and I can get certain features done in much less time. It’s not perfect but it’s like coding with a robot friend.
Still a lot to learn in this space but I’m starting to see some of the productivity benefits for sure.
I currently use AI for things like summarizing text or videos, generating or improving text, asking general questions, and handling some terminal-related tasks.
I’m still hesitant to rely on it for complex programming or data manipulation, though. In its current state, there’s a noticeable margin of error, which makes me cautious. If it can confidently give wrong answers to simple questions, how can I trust it with something more critical? That’s why I think everything it produces should be double-checked, and sometimes that review effort ends up canceling out the time it was supposed to save.
That said, I feel like at some point I’ll have to fully embrace AI-assisted software development just to stay competitive in this job market.
The most value I get out of AI (Claude Pro) is for research, and for drafting text. Though only if I give it plenty of background information and clear constraints in terms of what output I want and how to go about it. It is all to willing to spew out lots of content to satisfy my requests, so it needs to be guided.
Also, it is great for summarizing and distilling text files and drafting emails and whatnot in my day job. So I have more time and energy for my side gigs.
In terms of coding, it is good for high-level thinking, and simpler, more focused coding tasks. But I am working on a new Rails app and it has been giving me outdated suggestions or ones that do not align with my goals. So it needs to be constantly checked and corrected.
But all in all, it is a good time-saver and a nice robot assistant/mentor.
How often do you hit the Claude Pro limit while working? I know the message cap resets every 5 hours, but I’m curious if that’s enough for doing serious work on days when you go all in with AI-assisted development.
Sometimes the chats get too long and slow down, then I ask it to summarize and paste into a new chat. What bugs me more is that sometimes the messages just disappear. But a refresh of the tab restores the messages. Looks like a long-running bug, judging from the results on Google search.