Zortos

@LexnLin damn bro got 10k credits how?

been iterating A LOT in cursor using gpt 5.5/opus 4.7/composer 2.5 It's actually really expensive to run these to create a skill 😭

@RedBuildsThings @cursor_ai @OpenAI It’s set up in Codex too, but Codex didn’t use it.

@Zortosdev @cursor_ai @OpenAI you configured context7 in cursor? did codex use it?

I tested @cursor_ai Composer 2.5 vs @OpenAI GPT-5.5 Medium Fast in a head-to-head coding speed test. Same prompt. Same task. Results: Composer 2.5: 43 sec GPT-5.5: 2 min 3 sec Composer finished faster, while GPT-5.5 took longer and made more tool calls.

That’s fair, but in this clip it wasn’t really “verification.” Codex first produced broken code for a simple task, then used extra tool calls to fix its own mistake. Composer 2.5 got it right in one shot. So the point isn’t “more tool calls = bad.” The point is Composer was faster while also needing less recovery work.

@Zortosdev @cursor_ai @OpenAI ngl framing 'more tool calls' as a negative is the tell faster isn't better if it's skipping verification steps

@pedroshakoor @cursor_ai @OpenAI Generally it doesn't hallucinate haven't encountered so far

@Zortosdev @cursor_ai @OpenAI Is composer hallucinates a lot ?

@KellehEyad @cursor_ai @OpenAI Composer 2.5 hits different. It completely changed the flow. It’s so fast I don’t even have time to think about the next message before it’s already done.

@Zortosdev @cursor_ai @OpenAI Composer 2.5 hits differently, it has changed the flow

@wachig_wiwossa @cursor_ai @OpenAI In this test, yes. Output quality was comparable, but Composer 2.5 one-shotted it while Codex generated broken code first and had to fix itself. So Composer performed better here overall.

@Zortosdev @cursor_ai @OpenAI Is it actually comparable in performance though?

@maingocanh1702 @cursor_ai @OpenAI In this test, the quality was pretty much the same, if not better. Codex generated broken code first and had to fix itself, while Composer 2.5 one-shotted the task. So Composer was faster and more efficient here.

@Zortosdev @cursor_ai @OpenAI The TPS of Composer 2.5 is wild. But how does the output quality compare to GPT-5.5?

@its_wandering_p Nearly triple the speed of GPT-5.5 on Fast Mode.

@Zortosdev how fast is it?

@MattDabit Most of the time, when people complain that a model is bad, they’re just prompting it badly.

Hard agree. No matter how good your LLM is, you own 100% of the output. "It doesn't matter if an AI assisted you or wrote it all — by the time it evolves into a pull request, no one should be able to know." AI doesn't make engineering easier. It forces you to go deeper and take real ownership. This piece nails it: rootfn.com/writing/10k-ho… @grok @xai @EricHasegawa1

@daradoescode Congrats!

We hit 4k 😢

@cursor_ai @OpenAI Subscriptions used: Cursor 20$ plan Codex Pro 200$ plan

Raw Prompt: # Stripe Discord Webhook Implementation Plan **Goal:** Build a Node.js Stripe webhook service that listens for successful payment events and sends Discord embed notifications with payment details. **Architecture:** Use Express with a raw JSON body on the Stripe webhook route so Stripe signatures can be verified safely. Keep payment event normalization, Discord embed construction, and server routing in separate modules so each piece can be tested directly. **Tech Stack:** Node.js, Express, Stripe SDK, Node built-in `fetch`, Jest, Supertest, dotenv.

@ericzakariasson Add computer use to Cursor, and with Composer 2.5 it will be unstoppable because of its speed.

what do you think of composer 2.5 so far? how can we make the next model even better? want to hear your feedback on behavior, speed, quality, whatever!

@piliponful try to use composer 2.5

bro..... claude is so fucking bad at frontend I can't

@daradoescode Oh hell yeah!

is this what I think it is 👀

@KitsuneKode Yes, I have used sub-agents, and I haven’t had any problems even with heavy usage. I always give very detailed prompts explaining exactly what I want them to do.

@Zortosdev Hey have you used subagents? I got to like 40% usage after using subagents to complete multi task workflows

@leodev Yeah, but there’s no API endpoint you can integrate with directly without relying on their SDK wrapper. That makes it unusable for some application use cases.

@Zortosdev For the downside don't they offer it through the cursor sdk where you still pay api prices, but for the first week it's off 90% or something like that. I do agree with you that Composer 2.5 sucks at long/big tasks, I'll try your workflow and see if it works.