if you are using and paying to cursor i think you are living in 2024. reasons: - bad usage limits. - poor model reasoning output - bad roi get a claude code or codex subscription they have really good usage limits, great roi and better reasoning. ps: i paid for cursor subscription for whole 1 year in 2024.

@code_kartik subsidised tokens aren't forever

@code_kartik You can also use claude code & codex side-by-side I built an IDE for this. It's open source. github.com/stablyai/orca

CEO Charles Liang Advances Data Center Building Block Solutions Don't Miss the Keynote @ Supermicro Innovate!/COMPUTEX

the 'good usage limits' part keeps shifting. claude pro hits 62% weekly by monday on one refactor now, and the 5-hour rolling window kicks in mid-session. still beats cursor's caps but it's no longer the unlimited buffer most people quote. we built claude-meter to track that weekly burn and the 5-hour window in real time so you see the cap before it hits, claude-meter.com/r/9unpyb4c

claude code's usage limits aren't 'really good' on opus, you hit the rolling 5h wall mid-refactor and the weekly bar's at 60%+ by tuesday. cursor pricing is worse but the limit framing is the wrong axis. we built claude-meter to track that exact 5h rolling window and weekly burn so you see the wall coming, claude-meter.com/r/zm2tf5a8 written with ai

@code_kartik Cursor is objectively a better harness, but so is Droid and it can be used with your Codex sub. Either way, you should exploit the token subsidies from OAI/Ant while they are here.

@code_kartik For individual Claude code and codex is good for company and business however cursor is better Claude specially have tons of issue with rules and stuff One bad prompt whole org is gone And another things is extra usage cost too extra

@code_kartik But they gave me 14 months free pro subscription 😞

@code_kartik We have been using cursor for our whole team and it works like a charm

@code_kartik Still I think overall cursor feels better atleast from usability pov, it has a lot of features that makes agentic coding and reviewing easy..

@code_kartik did you try cursor in 2026?

Chat with the most truthful AI on Earth. Try Grok free today.

@code_kartik If you are using claude code then you are still living in 2025 Use open code with OS models

@code_kartik hear everyone's but do yours I have used cursor pro for a year and it does my job pretty well

@code_kartik Agree with the limits but I heard using claude model in cursor increase the performance of output a little more than normal claude code

@code_kartik so tru

@code_kartik just locally run qwen3.5 or smth on top of claude code instead

@code_kartik also use hermes and tools like graphify they will help you save your tokens and produce better op

@code_kartik skill issue

@code_kartik get a claude code subscription they have a really good usage limits 🫩🫩 agree with the codex though

@code_kartik The reasoning gap really shows on multi-file work. Cursor's autocomplete feels polished but once you're delegating a whole feature, claude code's context handling pulls ahead fast.

@code_kartik Great 2025 advice, 2026 looks like @opencode and @OpenRouter for specialist stuff

@code_kartik Claude code is best than cursor

@code_kartik is cursor composer limits also bad?

@code_kartik @cursor_ai deserve some thanks

@code_kartik Le me building saas with antigravity (free google pro plan) 🥲

@code_kartik i got pro+ for a year