52 engineers on Hacker News describe using GPT-4 first-hand: 20 endorse it, 11 regret it, 6 are mixed.
“are all amazed at how well gpt-6.1-luna is at their jobs. They've all been using it exclusively for 2 weeks”
endorse · production · cyberpunk · 2026-10-30 · source ↗ · wrong label?“We were also surprised by the low sol score but it seems consistent with our experience in using it in the field”
regret · production · iopapa · 2026-09-04 · source ↗ · wrong label?“we run a chatbot using it at my work and when extracting data from the conversation it will occasionally make up an email address that wasn't entered into the chat”
mixed · production · sandcat_ · 2026-08-16 · source ↗ · wrong label?“I have been experimenting with this workflow across a variety of computational domains and have achieved consistent results with both Opus and GPT.”
endorse · personal · minimaxir · 2026-04-23 · source ↗ · wrong label?“PDF parsing for company headquarters address, it hallucinated an address that didn't exist in the document”
regret · production · steveBK123 · 2026-02-27 · source ↗ · wrong label?“I am continuing to have success using Cursor's Auto model, and GPT-5.1 with extremely long conversations.”
endorse · personal · elliotto · 2026-01-15 · source ↗ · wrong label?“I never really got good results with GPT5.1.”
regret · personal · djinnrutger · 2025-12-17 · source ↗ · wrong label?“Sometimes that's wildly useful. Sometimes it's nonsense. The trick is knowing which is which.”
mixed · personal · EastLondonCoder · 2025-12-12 · source ↗ · wrong label?“5.2 is very good at finding bugs”
endorse · personal · agentifysh · 2025-12-12 · source ↗ · wrong label?“I've had this same experience from GPT3.5 to the latest and greatest models.”
mixed · personal · 2 yr · torginus · 2025-04-10 · source ↗ · wrong label?“Clues are not really good, currently using gpt-4o-mini.”
regret · personal · super7ramp · 2024-12-31 · source ↗ · wrong label?“I find even 4o seems to understand the problem better.”
endorse · personal · h_tbob · 2024-12-22 · source ↗ · wrong label?“I use it in combination with GPT, which helps me create documentation.”
endorse · personal · Puchaczov · 2024-12-19 · source ↗ · wrong label?“I use Claude for writing code but GPT-4o for answering coding related questions, because I noticed that ChatGPT is much less prone to hallucinations than Claude is.”
endorse · personal · mike_hearn · 2024-12-05 · source ↗ · wrong label?“Asking gpt to first output prolog program seem to 100% fix it!”
endorse · personal · skyde · 2024-06-06 · source ↗ · wrong label?“product quality is degraded with gpt-4o.”
regret · production · ramoz · 2024-05-15 · source ↗ · wrong label?“I have tried using GPT4 & gemini extensively, and the amount of bullshit generated makes it unreliable if you don't already know the domain.”
regret · personal · huimang · 2024-03-12 · source ↗ · wrong label?“it was simply refusing to do so without me changing anything in the prompt... This lasted for a good few hours.”
regret · production · anshumankmr · 2024-02-21 · source ↗ · wrong label?“I've been using GPT4 extensively lately and I'm 60% convinced as well.”
endorse · personal · aurareturn · 2023-12-27 · source ↗ · wrong label?“I have an example of this I've been using, only gpt4 and nous capybara 34B have gotten right”
endorse · personal · ParetoOptimal · 2023-12-24 · source ↗ · wrong label?“I've been using it daily since it came out), but it seems like you know of some miracle models on HF which can?”
mixed · personal · p1esk · 2023-12-15 · source ↗ · wrong label?“it seems to fail more often or ignore parts of the context window or just doesn't really answer the question.”
regret · personal · saberience · 2023-12-10 · source ↗ · wrong label?“In my experience, `gpt-4-0314` was the peak gpt-4 model.”
endorse · personal · longnguyen · 2023-12-10 · source ↗ · wrong label?“Had a situation with 3.5 where it was refusing to show me example Ada code in response to the prompt”
regret · personal · lijok · 2023-12-09 · source ↗ · wrong label?“we switched to that and are happier for it.”
endorse · production · yinser · 2023-12-07 · source ↗ · wrong label?“GPT-3.5 is extremely powerful, and it's what we use in production.”
endorse · production · phillipcarter · 2023-12-06 · source ↗ · wrong label?“the GPT-4 code generation abilities are really incredible - but you have to be experienced at prompting.”
endorse · production · cryptoz · 2023-10-31 · source ↗ · wrong label?“am I using the latest GPTs to do my work and save a lot of time and get more things done? hell yes”
endorse · production · jostiniane · 2023-08-26 · source ↗ · wrong label?“GPT3.5 was fast enough to be usable but often inaccurate; it would ignore some information or get confused about simple mathematical comparisons”
mixed · production · 0.125 yr · jongjong · 2023-08-25 · source ↗ · wrong label?“I've been using GPT this way. ... Does a great job when prompted with a header file.”
endorse · personal · alar44 · 2022-12-23 · source ↗ · wrong label?“I had been using these amazing tools for work and wanted to make this tech available to more people”
endorse · production · solomonyardley · 2022-12-22 · source ↗ · wrong label?“While I've been using it for coding, most of it is high level "what are the trade-offs?" kind of thinking”
mixed · personal · retrac · 2022-12-20 · source ↗ · wrong label?“It's been saving a lot of time for out team - so we just packaged it for others to use”
endorse · production · 0xferruccio · 2022-12-16 · source ↗ · wrong label?“No reasonable amount of fine tuning can save GPT from performing poorly on a space insertion task”
regret · production · yunyu · 2022-12-09 · source ↗ · wrong label?“mined GPT-3 for a bunch of industry specific use cases for our platform... quite surprised by how well it worked.”
endorse · production · sidgarimella · 2022-08-16 · source ↗ · wrong label?“I was accepted into the OpenAI GPT-3 beta a few months ago; the capabilities of that model have further convinced me”
endorse · personal · imnotreallynew · 2021-09-19 · source ↗ · wrong label?“We tried GPT-3 for our email startup... prone to false positives and ultimately requires more work than doing fancy Regex.”
regret · personal · tetrisgm · 2020-09-11 · source ↗ · wrong label?“fine tuned on my home 5090 in a few minutes on 500 articles tagged by gpt 6.1”
neutral · personal · Aerialoo · 2026-10-27 · source ↗ · wrong label?“I've been using the following pattern since gpt3, the only other thing I have changed was added another parameter for schema for structured output.”
neutral · personal · 5 yr · jmogly · 2025-12-28 · source ↗ · wrong label?“With Opus4.5, Gemini 3 and gpt 5.2, it's now 20%, maybe 50% if we only talk about personal code output.”
neutral · production · orwin · 2025-12-28 · source ↗ · wrong label?“I've been using composer-1 in Cursor for a few weeks and also switching back and forth between it, Gemini Flash 3, Claude Opus 4.5, Claude Sonnet 4.5 and GPT 5.2.”
neutral · production · santoriv · 2025-12-19 · source ↗ · wrong label?“those are the daily drivers we use internally w/ Letta Code”
neutral · production · pacjam · 2025-12-17 · source ↗ · wrong label?“an earlier version of this project which actually used GPT-4 to write python scripts that Blender would then execute”
neutral · personal · gd3kr · 2024-12-12 · source ↗ · wrong label?“I personally have fine-tuned gpt models since 2022”
neutral · personal · 2 yr · bbischof · 2024-06-02 · source ↗ · wrong label?“Been using Riza for the past few months at our startup for executing code generated by GPT4.”
neutral · production · sikan_ · 2024-04-30 · source ↗ · wrong label?“Both return JSON more consistently than gpt-3.5-turbo.”
neutral · personal · bugglebeetle · 2023-12-27 · source ↗ · wrong label?“not enough to entice me to switch models in production.”
neutral · production · digitcatphd · 2023-12-06 · source ↗ · wrong label?“we used GPT-4 to extract keywords, entities, labels, and themes from the whole document.”
neutral · production · dgarnitz · 2023-12-01 · source ↗ · wrong label?“where fine-tuning even the 7B Llama-2 model outperforms GPT-4 (surprisingly)”
neutral · personal · robertnishihara · 2023-08-23 · source ↗ · wrong label?“We're currently only using GPT-3 (davinci-003) on production.”
neutral · production · nickca · 2022-12-14 · source ↗ · wrong label?“I have just been using 0 for programming tasks with text-dacinci-003”
neutral · personal · ilaksh · 2022-12-14 · source ↗ · wrong label?“We use GPT-3 to write content following best practices.”
neutral · production · jacob_rezi · 2021-12-14 · source ↗ · wrong label?