“Over the last two days, though, my usage has suddenly gone crazy. After only around 10 relatively normal prompts, I'm already at about 70% of my usage.”
“From the outside, Cursor shouldn't have worked. They compete against the most well-funded and fastest-growing companies of all time, and *sit on top* of those same competitors to power their product.”
“Generating the code with Cursor or Claude is only half the battle. Getting it off localhost so a client or teammate can actually click around is always the annoying part.”
“Cursor basically has grok 4.6 and they will let you use it to the ground. Technically the best limits since you get 2x on grok and composer. Buy rw, you end up taking much longer on any cimplicated task, and the limits run dry when you reiterate.”
“I was thinking of getting Claude code, but is Claude pro still worth it? because there have been a lot of talk about the latest ChatGPT model and cursor being good so I just want to know, like which AI coding tool are you guys using?”
“I have it drive (literally control) Cursor on auto mode. Sort of best of both worlds, I don't run out credits on either service this way, and they share the cognitive load.”
“I've been trying Cursor alternatives for months, dozens of them (GUIs, I don't really like TUIs). I've recently started using OpenChamber and I'm loving it so far.”
“Cursor is a mature environment where features for agent-based development emerge much earlier due to the team's experience and the vast amount of data. Zcode is a relatively new harness that is still far from Cursor's quality.”
“I was happy with Grok 4.5 and Composer 2.5, but this led me to cancel - twice, I was willing to give it a second shot. Too many hoops to jump through.”
“Wondering if anyone noticed that the browser helper that Cursor Cloud agent uses is hard-wired to Claude Sonnet. It does **not** use your parent model, and there’s no setting to change it. So you can think you’re on a cheap model and still burn a ton of tokens under “Other Models”.”
“My pattern is I fire off a Cursor agent task, tab over to Slack or a browser, and then just, forget to check back. Lose 5, sometimes 10 minutes before I remember it finished. Ended up building something that talks the result back to me out loud instead, so I'm not stuck babysitting a tab waiting for it to finish.”
“It does not take from your cursor usage BUT, a big BUT, it can at times launch a cursor agent in the background so that can use your limits so they are saying...It hasn't happened to me yet but most likely, if you use it to code a lot it probably does use your cursor account.”
“Using cursor since begining it's the best value and works great. Now with grok and grok bot there's no better offer. You can supplement yourself with Claude code to work in conjunction on same repos for planning only.”
“A workflow can move from Cursor to Claude Code to Pi to Codex by copy-pasting prompts and outputs. Each harness logs differently, which leaves a hole when you try to debug, audit, or trust the run.👇”
“Got me wondering - do you think we could eventually see a similar level of performance from future Cursor/SpaceXAI models, especially for coding and software-engineering tasks?”
“I made the jump to Freebuff a few weeks ago after constantly running into Cursor’s monthly request caps on larger builds. I’ve been running Claude 3.5 Sonnet through it to build out full-stack Next.js dashboards.”
“I started with cursor, then eventually tested out Hermes agent out of curiosity. Then I set it up so that hermes agent delegates coding tasks to cursor.”
“The results are in! 57% of you voted for Code A (Fable 5.1 Max in Cursor) and the other 43% voted for Code B (GPT Astra Extra High on Codex). The conclusion I’ve made, having used both over the last few days, is that Astra and Fable are effectively equivalent for Laravel work.”
“cursor is an ide with agent mode bolted on, zcode is a harness built specifically around glm. same model can score differently depending on the harness wrapping it, so 'which is better' depends more on your workflow than on raw model output..”
“I find that I am not actually mining the code myself, so the IDE is not required, and the agent window could be replaced with a system that works exactly the way you do, while still using cursor agents (via CLI)”
“Cursor IS the official harness for Composer 2.5 (and supposedly coming Composer 3.0), and has seemingly become the official harness for Grok. I think the era of harness exclusivity is upon us.”