@sandislonjsak Cursor is genuinely so much better than CC, most people just fall for anthropic marketing
English
ligma squeeze
9.4K posts
ligma squeeze
@LigmaSqueeze
accepted in both sietch (dog coins) and village (bloomberg)
Katılım Temmuz 2021
3.4K Takip Edilen844 Takipçiler
I put same task and same prompt to claude code and cursor.
Cc took 9 minutes, cursor took 2.
English
I actually rewrote this course three times 🙈
Coding agents and their best practices are changing quickly, making creating educational content quite difficult.
This course comes after 10 different talks I've given about how Cursor uses Cursor internally, and hundreds of conversations with customers about the best ways to use agents.
I tried to find the right balance between being useful today *and* foundational for your future. Let me know what you think! And now onto the next one...
Lee Robinson@leerob
Learn how to use coding agents in 30 minutes! This course teaches you how to build software with agents: plan new features, fix bugs, review and test code, and more. It's 100% free and these concepts apply to any agent!
English
@m_emanaftab @Prodigers @rileybrown honestly claude code is a pretty bad harness for claude models (which are good). there’s a reason it’s at the *bottom* of terminal bench for opus 4.6. it’s sloppy
English
every restaurant is "just a wrapper" for food. but there's a difference between the chef who owns the kitchen and the restaurant that orders delivery. cursor is a great restaurant that serves someone else's cooking xD. claude code is the chef eating in his own kitchen. when the recipe changes, one finds out first
English
Cursor, Codex, and Claude Code desktop look exactly the same now.
English
@ecommerceshares unfortunately the app, like most of the stuff theyre shipping, is barely functional, like genuinely freezes when you try to send a prompt lol. they’re largely coasting on the idea that people will see their posts and be impressed without actually trying anything
English
These guys are just shipping 10x faster than anybody else right now.
There seems to be a lack of both focus and urgency everywhere else. I don’t know why… but it’s a bit weird.
Claude@claudeai
We've redesigned Claude Code on desktop. You can now run multiple Claude sessions side by side from one window, with a new sidebar to manage them all.
English
@anissagardizy8 @LauraBratton5 opus psychosis. you get way more usage with a better model with Codex or 5.4. engineers who don’t use it because it’s “not as fwiendly as opus” are NGMI lol
English
Uber's CTO told @LauraBratton5 that AI coding tools—particularly Anthropic’s Claude Code—has already maxed out its 2026 AI budget 📈
“I'm back to the drawing board, because the budget I thought I would need is blown away already,” Neppalli Naga said.
theinformation.com/newsletters/ap…
English
@robinebers Isn’t it just this?
x.com/mntruell/statu…
Michael Truell@mntruell
We A/B test many parts of Cursor: model checkpoints, UX, and the agent harness. In this case, we tested less than 1% of traffic to compare how Claude behaves with the CC harness versus our default harness (something we do often with offline evals). Our team does lots to improve the speed, feel, and accuracy of our harness on the queries users ask and care about. Hope to share more about this work soon.
English
so, Cursor is using Claude Code under the hood
spent two hours really understanding what they're doing
I even got it to route requests through my actual Claude Max subscription instead of the Cursor ones. this works right now (!) but I won't share how
it does prove though that Cursor *could* do if they truly wanted to
and I really wish they did, and just charge for the app and cloud instead
English
We've been tricked, again. Many of the thousands of bugs and vulnerabilities Mythos found are in older software are impossible to exploit. And the severe zero-day reports rely on just 198 manual reviews tomshardware.com/tech-industry/…
English
Honestly don't know what happened to Claude Code. Tried a one-off simple task on a fresh directory yesterday, tried a bunch of things that didn't work, asked for a ton of permissions, and then got stuck for 4 minutes before I got tired of waiting and killed the session. This was on medium effort.
Switched to Codex gpt 5.4 with medium effort and one shotted the task in under 1 minute.
English
@meansoabstractn @var_epsilon it's a pretty small fee. saving a ton more than that by getting to use gpt-5.4 and composer-2 instead of only dario's overpriced slopus lol
English
another interesting datapoint from friend @ bigtech co
varepsilon@var_epsilon
from a friend at an ultra trad big tech co
English
@sunnyworks @sandislonjsak Cursor CLI is actually amazing and for some reason nobody talks about it
English
Claude is hungry with tokens and actually not great at jobs that require lateral complex thinking. But there is something about the workflow using the TUI that feels better than a chat window inside an IDE. I can’t tell you what it is, but Claude Code has made it a habit for me. I don’t like it but I can’t switch either. It’s a paradox.
English
ligma squeeze retweetledi
how is claude code the worst harness in opus 4.6 benchmarks 💀
English
@kr0der Xcode MCP works well, it can read logs and trigger builds and everything. but I don’t think it does video demos, would need to be able to run Xcode on the VMs but they are Linux
English
anyone know the meta for testing IOS apps in Cursor?
i.e. getting the agent to test an IOS app?
can Cursor cloud agents take video demos of IOS apps?
English
We don’t talk about this enough.
Opus scored 20% higher in Cursor than in Claude Code.
edwin@edwinarbus
Matt Maher tested frontier models in Cursor v. other harnesses. Cursor boosted model performance by 11% on average: Gemini: 52% → 57% GPT-5.4: 82% → 88% Opus: 77% → 93% His benchmark measures how well models implement a 100-feature PRD. @cursor_ai consistently outperformed.
English
@amir I miss when the information wasn’t a wing of anthropic’s marketing department 🙂↔️
English
Anthropic has masterfully cloned OpenClaw and is grabbing share quickly in personal AI agent market
English
@Timur_Yessenov @GergelyOrosz anecdotal, but we’ve seen cursor do much better than claude on existing codebases / brownfield tbh. semsearch seems to help it be more accurate and to waste much less time
English
@GergelyOrosz cursor for greenfield, claude code for refactors. we use both: cursor when building fresh features, claude code when fixing tech debt. cursor's composer kills 0→1, claude code saved us 8h last week on 800-line configs
English
This week, from a CTO at a large tech company, when I asked them about AI tools usage:
"We have teams of ~100 evaluate various tools: Claude Code, Codex, Cursor etc. Then we decide where to invest in.
We're actually starting to get really bullish on Cursor, again"
English
@ericzakariasson tend to agree. been maining 5.3-codex and 5.4 since release and been really enjoying Composer-2 now too!
English
best coding models right now
sync work, shorter iterations, 1-5 agents in parallel:
- plan: composer 2
- work: composer 2
when you're in control, you simply only need composer 2.
async work, longer iterations, deep work:
- work: gpt 5.4
for async, you still want maximum intelligence as speed is less important with a model that can run for a really really long time
ui & taste
- opus 4.6 is still really strong here
- composer 2 is a fast follow, but not as strong (yet)
this space moves fast, so it will probably change soon!
English