Will I think ChatGPT Pro is worth $200 by mid-December 2025?
23
1kṀ1419
Dec 16
25%
chance
4

Just got chatgpt pro for 1 month and i wonder if i will find it worth the 200$ by end of my susbscription.

Relevant info

  • my spend on cursor last month was 350$ and i think it was very worth it

  • my favorite model by far is sonnet 4.5 and i always regret using any other models pn cursor (except opus for planning)

  • i write code ~30-40 hours every week

  • i have been a chatgpt plus subscriber since the day they introduced it and i am very very happy with it

  • i will consider both personal and professional impact although professional will have a much higher weight since plus was covering all my personal needs and i never really hit any limits

  • i am creating a lot of educational videos with ai right now for work and i wanted to integrate sora for a while with no success but pro users can generate videos up to 25 seconds via the web app so if this is valuable in any way i would probably resolve yes

  • i will give codex (cli & vscode extension) a fair chance and use it exclusively for atleast 7 days. If i find myself naturally using it afterwards instead of cusor without having to force myself i would resolve yes

  • there are a couple of things i didn’t manage to do with cursor, if codex is able to implement them then this would be enough to resolve the market yes

  • i tried gpt-codex over cursor model selector and hated it. it is literally a slop machine. (i heard cursor works closely with anthropic to make sonnet as useful as possible so maybe codex makes openai’s model shine)


AI Clarifications 👹

  • Update 2025-11-18 (PST) (AI summary of creator comment): The creator will not create custom context engineering scripts themselves. If the context engineering provided by Codex's built-in tools (CLI & VSCode extension) is not sufficient compared to Cursor's context engineering, this would count against ChatGPT Pro being worth it.

  • Update 2025-11-23 (PST) (AI summary of creator comment): Cursor gave the creator free access to their model composer, which the creator finds very good and likes. This may affect the evaluation since the creator was supposed to use Codex exclusively after reaching their Cursor limit.

  • Update 2025-11-23 (PST) (AI summary of creator comment): The creator is now using Codex again with an improved agents.md that copies Cursor's planning mode to trigger planning with Codex. The creator's experience with this setup has been positive so far. Regarding Cursor's model composer: The creator finds it sits somewhere between Haiku and Sonnet in capability - good for minor edits but not major explorations.

Get
Ṁ1,000
to start trading!
Sort by:

i don’t think this market should be at 22%

i love love love chatgpt pulse!!! not enough to resolve yes but i love it. also started getting used to codex. it is much easier to have multiple instances running at the same time since codex runs in the terminal and you can just have cursor on the left and terminal on the right amd it is almost impossible to hit the rate limits. my prediction is that if everything continues like this and i keep finding better ways to work with codex that this market will resolve yes.

i reached my cursor limit (300$) and now have to use codex till end of the month 🙃

@Soli cursor gave me free access to their model composer and it is very good. i like it.

@Soli i love composer but it is clear it sits somewhere between haiku and sonnet which makes it good for minor edits but not major explorations. i am not using codex again with an improved agents.md that copies cursor's planning mode so that i can also trigger planning with codex. till now my experience is positive.

tested codex for 2 days, not a fan honestly but will still give it a chance. it has a tendency to completely ignore my coding rules described in agents.md and over-complicating everything. it is also very annoying have to explain to it every time that it should not code but rather just create a plan that i can review. cursor is much much better.

so i tested codex today and i like it, it doesn’t feel like an improvement over cursor but also not a big downgrade. codex was able to solve a small issue that i had and wasn’t able to solve with sonnet which is nice. i opened codex in terminal and put the terminal in a small window next to cursor while talking to sonnet at the same time over cursor and i liked this flow. felt a bit more productive than usual but also feelings are not always accurate haha. it was annoying having to tell codex repeatedly not to start coding and that we are still in the planning phase, something that cursor solves very nicely.

it's useful if you have good context engineering for your code, like a script that can efficiently dump relevant parts of your codebase into a text file that you can give to gpt-5-pro

@Dulaman i don’t think anything i can create would match cursor’s context engineering and i don’t see this as my responsibility, so if the context engineering provided by codex (cli & vscode extension) aren’t sufficient then this wouldn’t suffice

@Soli for sure. Think of the comment more of as a hint of something you can try if you want. Not in the sense that it alters the resolution of this question.

(it's pretty easy to throw together a quick python script that dumps everything to a text file)

basically gpt-5-pro really shines when you just give it a big data dump and say hail mary

(and then for larger codebases that can then be made more efficient with tweaks etc. But for smaller codebases you can just give it everything)

@Dulaman thanks for the hint, the codebase i am working with right now is quite small so i will try dumping everything in for sure and see how codex performs. i also created a quite extensive agents.md file which combines my 11 cursor rules files into one and have a very good summary of the overall project structure / coding philosophy

@Soli makes sense. Essentially the model you get in codex and the gpt-5-pro you get in the web interface are different things.

(not referring to this question resolution but in the vein of general suggestions) If you give gpt-5-pro your whole codebase as a text file and then ask it to come up with a plan to edit your code and look for references online to cite in its response, then it can give you answers that are far beyond what codex can provide. Optimal setup right now appears to be combining both the version in the web interface and codex in the cli/ide. If you're paying for the subscription anyway then may as well use the highpowered model it gives you.

@Dulaman haven’t thought about the web version but if it allows me to work while on the go this would be amaaazing

🤖

Meowdy! You seem quite invested in AI tools and coding, so Pro might really shine for you. I’ll revisit this later tonight with a sharper claw! :3

© Manifold Markets, Inc.TermsPrivacy