I use it for first drafts of anything I need to write. Once I get something I like I then edit it to my liking.
I also use it as a quick project organizer that I can customize.
Also, it's very good as an idea generator.
How do you use it?
For example, we are now troubleshooting an old Openstack, and the online doc is a mess. ChatGPT can give examples, deep explanations and suggestions.
80% of the answers are good, there are still 20% that I need to do (fill it more info, change the questions, etc).
I am actually looking forward to be able to pay it, so I can get answers faster and it can be stable, really helpful when I am in a production incident.
That aside, my one try at using it for something productive was asking it, erm I mean her, to write some mod_rewrite rules for my Apache configuration for something I was working on. What it gave me wasn't perfect, but then again I didn't really pose the question correctly in hindsight. But it was a little surprising that it gave me something pretty close to what I needed.
I used it occasionally for other things like different coding question or framework, etc but the verdict for me is still will use the tool longer term.
I’m not comfortable sharing my queries with OpenAI for a lot of the potential use cases I would have. I would be much more interested if that $42 pro plan came with a contractual privacy guarantee. But that’s unlikely.
It's a great little trick if you're feeling stuck or procrastinating.
The most benefit though, I get from it communicating with others (note they verbosity of code is also for the benefit/communication with others!). Generating reports with explanations why code is bad/insecure, broken down task/fix lists, more elaborate communications to the team than I would write etc. All kinds of situations where, theoretically, the receiver could’ve used my input and could’ve pulled it through chatgpt themselves, but didn’t and instead comes back with questions which are answered/encoded in my input and chatgpt can drag out for them.
It’s well worth the $42/mo.
Before ChatGPT, I learn those by searching examples and real repos which is more efficiency that follow tutorials or searching in docs. But sometimes it's hard to find examples you need. But with ChatGPT, it generates exactly sample codes for me to learn stuffs, and I even do not need to spend lots of time searching for resources.
So yeah, tending to agree it's not really up to $42/mo value so far...
Id always been weary of using postgre as a DB (which it suggested I use) due to past trauma, but helped walk me through installing and using it.
I've tried using it for solving errors, but it seems to have trouble here. Perhaps the next version will be better. E.g. it tries suggesting things that it had previously said could be the source of my error. It's the lack of logic which is its weak point in solving problems that can't be solved quickly.
For instance I had a long sound file I needed to cut to pieces based on silences and rename the pieces in a certain way, and it gave me a working script using a tool I wasn’t aware of.
Otherwise I would have just done it by hand as it would have taken only 15 minutes, and wouldn’t have been worth writing a script by hand, but this way it took 5 minutes instead.
I'm slowing down my Google search usage so fast, this other day I was "chatting" with chatgpt about a niche topic (creating Maori Tui cloaks) and although it doesn't have images it's way better than seeing a bunch of ads and shitty links.
ChatGPT writes the draft and I rewrite it in my style. It improves the efficiency significantly.
I use it in place of a search engine if I get bad results from Google. Farily often the result from ChatGPT is better at this point.
It's good for book summaries if it knows about the book, which it often does.
It's great for learning new concepts. I have a enthusiast level interest in quantum physics and quantum computing. I always have a bunch of naive questions after I read book related to the subject. ChatGPT's explanations are limited but fascinating and have introduced me to concepts I was previously unfamiliar with such as quantum teleportation and quantum communication.
It's also good for learning new programming languages. I think I will start with ChatGPT from now on with any language I learn. Paticularly the ability to type in 'I know how to do X with python, How would I do it with Rust? This part doesn't make sense to me - can you explain why it does that? Could you show what the code would look like in Go?' This is just endlessly useful. You still want to incorporate the official docs and other sources (I think?) but it's just such a delightful path to learn. It's worth noting that with learning being much easier this way you could also probably learn more langugages per year now than might have otherwise been possible if that was your goal.
Occasionally I use it for creative stuff like making a silly poem or rap about something.
I haven't used it much this way yet but I've seen others comment on using to generate better prompts for Midjourney or other Generative models to great effect.
I love to use it as a linux machine if you know that trick[0]. There are other interesting things it will simulate for you too though. It can pretend to be a cisco router on a small subnet, a kuberentes cluster, a cloud platform. These are all interesting and strange paths. With these you can also include meta characters in your prompt that allow you to ask it to do things within the simulated OS in natural language too.
I still feel like my understanding of the overall model is quite limited and I also like to try to explore what it might be able to do. I like to try to push at the boundaries and edges of what it's capable of, particularly connecting it's generative capabilities with batteries included frameworks I know of to build prototypes absurdly fast.
It takes some proding and knowledge of the underlying tools (though you could ask ChatGPT to explain this to you) you can create a whole stack. Describing in natural language what you want the terraform to look like to deploy into a kubernetes cluster in eks and accompanying nix to configure the docker containers. You can then use tools like django or flask that have lots of things preconfigured and in 15 or 20 minutes have an entire infra ready to be deployed in one command. In many environments this would take months of work. Though, Replit seems to be working on making approaches like these less of a hack.
I think thats what ChatGPT mostly feels like to me. One big hack. I see this criticism of it even sometimes, that it's sort of hacky and not actually that advanced. But especially if you do with what you're supposed to do with any other hack, chain it together with dozens of other hacks - you get a final result of crazy capability. And ChatGPT can chain with a whoooole lot of other stuff. Particularly through code.
I'm pretty excited it's only $42 a month. I don't know what I would pay for it but probably a lot. I just have so much fun goofing around with it.
[0] https://www.engraved.blog/building-a-virtual-machine-inside/