Live data from Hacker News

Ask HN: How are you using GPT to be productive?

news.ycombinator.com

31–40 of 752 posts

Re: Ask HN: How are you using GPT to be productive?

#31
I've asked GPT-4 about the human brain and specific metabolic pathways every day so far. If you ask it for answers in jargon like written in a textbook or a journal it will provide an obscene level of detail that you can then verify.

As a "domain specific words" generator for building up a glossary, it can't be beat. From a prompt perspective I have to tell it that we are both world class neuroscientists but it knows more about this specific question than i do. Then I tell it to reply using correct jargon for the domain like that written in a textbook

Re: Ask HN: How are you using GPT to be productive?

#32
I've been a paying ChatGPT subscriber for about three or four weeks now. One day I had a few thorny work issues, asked the free version, and it was so good I got a subscription.

Since then I've asked some about general knowledge, history, religion, geography, politics, other topics of interest. Mostly in English, but some in Portuguese and a little in Spanish. It's extremely good in all three languages.

Mostly though I've been asking about random work topics that come up every day. We use a lot of lots of systems, tools at work, and I need to write software to handle diverse areas. ChatGPT cuts right to what I need as far as: (a) general knowledge of tools and what their purpose is; (b) surveys of categories of tools, comparisons between competing offerings; (c) specifics on how to use, configure, program against various tools, query data, change things; (d) questions on best practices and pitfalls. This is mostly in context of macOS, Linux, AWS, kubernetes, observability tools, and APIs for lots of DevOps-related systems. I do lots of coding in Python, I also do a lot of ad-hoc diagnosis of situations. (We have a great DevOps team that manages infrastructure with standard DevOps tools -- my job is to build what those tools don't address so well, and also to help build out future data-engineering efforts.)

I'd say my use of Google search to find relevant articles / pages has gone down 70%. One small example today, I wanted to use `jq` to process some `docker ... --format json` output to pull out some data. I don't want to learn the ins and outs of `jq`, I described my problem and it gave me a good template I could adapt.

Any time now in scrum or other meetings, if there's any question about something, we often just consult ChatGPT during our Zoom/screen-share sessions. I think generally I have a better sense on how to structure questions and question progressions to get quick answers than some others.

I've also found ChatGPT makes up stuff sometimes ... but it's usually close enough.

One comfort I have is that, at least for now, ChatGPT can't direct the overall organization of code for the many situations I need to address, so I'll have a job for a while. It does though fill in the knowledge gap at the edges, I don't waste near as much time searching for and reading documentation and examples. ChatGPT usually has good ready-made low-level examples when I need them, and high-level descriptions of tradeoffs and best practices.

I'm "committed" to Jetbrains tools, been using them for a long time. Today I began wondering what I might be missing from CoPilot, downloaded the CoPilot plugin for PyCharm (would also work for IDEA, DataGrip, etc.). I couldn't get the CoPilot plugin to log into Github and saw that others have had similar problems in the past ... so I can't use CoPilot yet. Maybe in a week or two I'll have a basis for comparison. (I don't want to switch to VSCode.)

EDIT: word choice, minor clarifications

Re: Ask HN: How are you using GPT to be productive?

#33
Lately I’ve been doing a lot of very deep discussions with GPT-4 about design issues I’ve had, and brainstormed ideas on how to improve them. In these discussions, GPT-4 has been able to really dig into the details and grok the problem space, and has surprised me with the quality and brilliance of its ideas and suggestions. Some of these I have tried, things that I would not have thought of, and it turns out the suggestions really helped solve the challenges I was facing. To me it’s like working with a brilliant person who can only communicate via text, can’t look anything up on the internet, and can’t remember more than a few pages of text at a time, but aside from those limitations is a genius.

Lots of people talk about generating code with Chat GPT, but to me its real value is in having deep detailed discussions about design problems.

It’s been so successful at this that recently I gave GPT-4 the full interview design skill assessment that I give to engineers when I interview them at Microsoft. GPT 3 wouldn’t be able to handle this, but what GPT-4 did here astonished me. My assessment is that this is a principal level performance. It didn’t have to do other things that normal candidates have to do but for this raw skills assessment of design skills it was spot on. It would have impressed me even more if it figured its final solution out from from the beginning, but that’s what it gets for blurting things out before thinking about them, which people do, too.

The important thing I’m communicating here is not that I am impressed because it’s amazing that a computer can do this stuff; I’m impressed at what it has done here compared to almost every human I’ve ever walked through this question with. The approach I used here is the same I use when asking candidates this question, because in addition to testing their ability to code, I’m looking for how well I can understand the candidate’s ideas, and how well they can understand and then apply my own ideas when I ask them to take a different approach on certain things than they were thinking of themselves. This is one area where many great coders struggle; they can code like the wind when it’s their own idea but struggle to work collaboratively. This kind of mental flexibility, ability to think of things in a different sequence or consider other ways to solve the problem after thinking of their own solution is also a required skill often in real-life meetings and other collaborative settings. I’d rate this candidate as an outstanding, top-notch collaborator.

My point here is that using it only to write boilerplate code is a waste of its best value. My suggestion is to get GPT-4 with its larger token limit, and talk strategy with it. Tell it all about your biggest challenges at a level of detail that would exhaust a normal human, and talk through ideas of how to improve your world.

Talk to it about your people problems too. It’s an astonishingly wise counselor who has a wealth of positive insights and suggestions. It’s also great for elegantly wordsmithing things.

Don’t miss out on the chance to collaborate with this endlessly creative and endlessly patient collaborator.

Re: Ask HN: How are you using GPT to be productive?

#34
post #7

A while back, I mentioned in a thread that I have found ChatGPT to be quite useful for correcting grammar and spelling errors ( https://news.ycombinator.com/item?id=34781888 ). Later, when they released their API, I developed a CLI tool for this purpose ( https://github.com/pncnmnp/grammatical ). Note that it is not flawless, but it works well. It has improved my writing productivity, both for blogging and emails.

I like your prompts for your script, I may integrate it into my workflow too. I have found that when I am editing my work, it is helpful to have the original sentence and the suggested sentence one after another. While this takes away the paragraph form, it clearly helps you tease out the "improvements" to either accept or reject. I also have GPT number the sentences. And then when I am done, I say - something like - "please assemble back into paragraph form, all revised sentences are accepted except numbers 5, 12." And then it reassembles the paragraph(s) with the revised corrections. I use ChatGPT (4) for really long stuff though... might not work with Turbo3.5.

Re: Ask HN: How are you using GPT to be productive?

#35
post #4

Mostly just asking stuff directly on https://chat.openai.com/chat . Last 8 requests were (all successful): - Asked it to improve a HN comment I wrote. - Asked about an idiom I couldn't remember, by saying it in other words. - Asked it to dumb down some things about options (finance) I didn't understand. - Asked it if I could use the eBay API to list my purchase history (you can, and it knew how). - Asked it to genera…

Asked it to generate pretty standard Terms of Service for an app I'm working on.

So you're using it generate a legally actionable document. Is this a good idea?

Re: Ask HN: How are you using GPT to be productive?

#36
post #4

Mostly just asking stuff directly on https://chat.openai.com/chat . Last 8 requests were (all successful): - Asked it to improve a HN comment I wrote. - Asked about an idiom I couldn't remember, by saying it in other words. - Asked it to dumb down some things about options (finance) I didn't understand. - Asked it if I could use the eBay API to list my purchase history (you can, and it knew how). - Asked it to genera…

Asked it to generate pretty standard Terms of Service for an app I'm working on. So you're using it generate a legally actionable document. Is this a good idea?

Yes. I got a cheap AI lawyer if someone sues me, all good. All jokes aside, the alternative was to not have a "Terms of Service", so fairly sure it's better than nothing.

Re: Ask HN: How are you using GPT to be productive?

#37
I recently had a relative staying over who was asking for help with some arduino code. He needed his chicken incubator to read from a temp sensor and activate a heat source when it was out of range. I thought it was a good opportunity to show him ChatGPT.

Description of the problem, some specifics on equipment, it bangs out a solid first pass to the problem that might require some tweaks. Then it occurred to me that it had not only written the code, it had set the temperature threshold to a sensible default without my having to separately look up that information.

Re: Ask HN: How are you using GPT to be productive?

#38

I’ve been using it to generate bash scripts because I don’t know bash scripting, and also have it generate regexes for me to search for code references in a programming language where the “find all references” functionality don’ doesn’t work quite well.

I didn't know bash scripting and then I wrote some scripts and then I knew bash scripting and use it for lots of stuff. Has ChatGPT resulted in learning by doing for you?

Re: Ask HN: How are you using GPT to be productive?

#39

I recently had a relative staying over who was asking for help with some arduino code. He needed his chicken incubator to read from a temp sensor and activate a heat source when it was out of range. I thought it was a good opportunity to show him ChatGPT. Description of the problem, some specifics on equipment, it bangs out a solid first pass to the problem that might require some tweaks. Then it occurred to me that…

please send your input, stories are whatever

Re: Ask HN: How are you using GPT to be productive?

#40
For coding, I've been using it like Stack Overflow. It really decreases my barrier to doing work because I can ask lazy follow-up questions. For example, I might start out by asking it a question about a problem with Pandas like "How do I select rows of a dataframe where a column of lists of strings contains a string?". After that, GPT realizes I'm talking about Pandas, and I'm allowed to ask lazy prompts like "how delete column" and still get replies about Pandas.

I also use it for creative tasks - for example I asked it for pros and cons of my cover letter and iterated to improve it. I also used it to come up with ideas for lesson plans, draft emails, and overcome writer's block.

GPT has drastically lowered the emotional-resistance barrier to doing creative tasks and improved the quality of my output by giving me creative ideas to work with.

Post reply on HN