Spent some time going through the SpaceX S-1 and built an interactive breakdown of the 3 business verticals i.e Space, Starlink, and AI (xAI + X).
Lot of interesting numbers in there if you love data → https://t.co/XNEs8CmEFm
@cursor_ai released their Developer Habits Report and a few numbers stood out to me:
- Lines of code per dev doubled YoY. PRs are 2.5x bigger. 36% of AI changes now go to commit with zero manual review.
- But the most interesting stat, the top 1% of devs produce 46x more than the average dev. That gap is only getting wider.
The tool is the same for everyone. But workflow are widely different.
@cursor_ai has cooked composer 2.5
Some observations:
It creates good UI mockups. So my workflow has been using it for concepts or changes by first creating mockups using html then turning it into code. With composer the workflow has sped up.
Genuinely impressed!
Love using it and the speed is insane.
@thsottiaux
Thoughts on codex and its comparison with CC and cursor.
My workflow:
1. Ideate with AI and highlevel write what needs to happen for users from product perspective.
2. write spec doc using superpower plugin. Roughly every feature that I develop has a spec doc.
2. Then implement them and review using superpower and manually check.
What codex does right:
1. Model trained well to catch how it should work in codebase.
2. Limits are awesome!
3. It follows instructions the best.
What codex needs to improve:
1. Model lacks curiosity. What do I mean by this ? I doesn't ask questions and doesn't interact with questions and asking clarification as much as I'd like. This might come down to personality of the model. I found that for now claude does this better especially for spec doc. Please improve this 🤞
Tried playing around with @GeminiApp and making it's response easier to read and more useful. Here is the prompt that has helped me. The outputs that I've been getting are a lot better!
For some reason, @claudeai code and @Google Gemini CLI give better output for the same SQL query questions than their web app. Gave the same prompt and question to both. CLI versions answered better.🧐
@AnthropicAI's Claude is agreeing a lot more with me similar to @chatgpt reverted release of 4o personality. Coincidence for me, or is anyone else feeling the same?
I've come to realize that action provides clarity. Not the other way around. It's still an idea I need to wrap my head around, so I wrote about it. Taking the inspiration from @brian_armstrong's words- action provides information.
Read on here👇
https://t.co/rJNNMFRhU4