Sabitlenmiş Tweet
✨ I made a GPT Agent for your next UX Design project!
Meet UX Design GPT. Your virtual assistant based on Design Thinking, powered by GPT-4 and DALL•E. Thanks to @OpenAI!
Link shared later in this thread! 👇🧵
English
Bapusaheb Patil
3.9K posts
Bapusaheb Patil
@baps_patil
Design Engineering x Brand Design for Products ✨ 5x Award-Winning Designer 🏆 Art Gallery: https://t.co/TmTFCAqlb3 🖼 I post about design, dev and AI.
Northern Hemisphere, Earth Katılım Nisan 2011
2K Takip Edilen2.4K Takipçiler
English
My retro OS web app, BapsOS, is now live on the @tan_stack showcase!
Thank you, @tannerlinsley.
English
@rohanpdofficial @Cloudflare @CloudflareDev Great work! It worked like magic on my Next.js projects.
For my TanStack Start projects, getting this error, and this is the latest Cloudflare setup for TanStack Start according to their docs:
English
Major architecture update just landed 🚀
All project types are now supported — OpenNext, Vite, Next.js, and more.
Try it today:
npx localflare
⭐️ If you like it:
github.com/rohanprasadoff…
@Cloudflare @CloudflareDev
Rohan Prasad@rohanpdofficial
Just released 🔥 localflare 0.2.0 : The best companion for wrangler dev. localflare.dev Change log : - Fixed all the major issues - Support for all type of projects and config with attach mode - Durable Objects now fully supported - Beautiful TUI with filtered logs - Wrangler passthrough args `npx localflare` @CloudflareDev @Cloudflare
English
@jamesacowling @convex Checking that was the first thing I did, which is why I asked here too. Because I got this error on Sentry, I'm guessing internal server errors have something to do with degradation issues?
English
@baps_patil @convex You can always check status.convex.dev but no we didn't have any downtime yesterday
English
We include a small profit margin on paid @convex plans but that also factors in savings from shared resources, AWS discounts, etc. Plus our team were responsible for optimizing the infra spend at Dropbox. It's unlikely you can host your own infra for cheaper than Convex.
Leo@leodev
@convex Not even sure how the cloud version is profitable it's cheap to scale and easy to use!
English
@ericzakariasson Grok 4.1 and Grok 4.1 Fast in Cursor. Let's go!
English
cursor 2.1 is out with a lot of improvements! go try out
plan mode and ai code review
Cursor@cursor_ai
Cursor 2.1 is now available! Plans now have an interactive UI for answering clarifying questions. Also new: in-editor code reviews, instant grep, and improved browser use.
English
Just tried the Gemini 3 Pro Preview model available in Cursor, if you add gemini-3-pro-preview as a custom model.
I was curious about it's pricing, so I gave it a few tasks.
The pricing, at least effectively in @cursor_ai, if not the API pricing directly, seems likely to be:
- $3 per 1M input tokens
- $15 per 1M output tokens
- $0.30 per 1M cache read
English
@JamKo_Ams @cursor_ai My sample size is too small, but it felt on part with GPT-5.1 Codex. Could be a different story after I use it for a few days.
English
@baps_patil @cursor_ai Would be a significant step up from GPT 5.1 and Cursor Composer 1. How did the model feel?
English
@jediahkatz Curious to know this from someone building Cursor:
Do you personally use GPT 5.1 Codex (High), Codex (Medium), Codex Mini (High) and Codex Mini (Medium)? If yes, what are your use cases and how has your experience been?
English
GPT-5.1 series, including 5.1 Codex is now in Cursor. Would recommend trying the Codex model for writing code if you didn't like the old version, since we've worked to make the integration great.
Cursor@cursor_ai
Three new OpenAI models are now available in Cursor: 1. GPT-5.1: For everyday tasks like planning and debugging 2. GPT-5.1 Codex: For ambitious coding tasks 3. GPT-5.1 Codex Mini: For cost-efficient changes
English
Agentic use for coding. In @cursor_ai,
1. sometimes it doesn't finish the task at hand when other models do, when the context length exceeds just ~100K tokens.
2. it throws random output tags on the 3rd or 4th user message.
Otherwise, it is really fast, and is really smart in the official @grok app. I would love to have Grok 4 Fast be my default model in @cursor_ai!
English
What are the most important improvements to make to Grok 4 Fast?
Critical feedback is much appreciated.
X Freeze@XFreeze
Met a few devs this weekend… They still didn’t know about Grok 4 Fast.. Grok 4 Fast is a game-changer: It has the Highest intelligence density ever - outperforms top models Ultra cost-effective: Up to 98% cheaper than alternatives Super-fast output = more productivity, less waiting It’s wild that this isn’t mainstream yet If you’re not using Grok 4 Fast, you’re burning both time and money Grok 4 Fast delivers GPT-5 (High) performance - for just 5% of the cost (95.7% cheaper)... Insane value Grok 4 Fast costs - Input tokens: 84% cheaper - Output tokens: 95% cheaper - Output speed: 150% faster GPT-5: $1.25/M input, $10/M output Grok 4 Fast: $0.20/M input, $0.50/M output Grok: 344 tokens/s GPT-5: ~138 tokens/s Let them know... If anything, Grok 4 Fast’s value is still under-appreciated
English
@rohantaneja_ Codex CLI is great at solving complex problems, but for building MVPs from scratching GPT-5 High in Plan Mode + Codex High in Agent Mode in Cursor is the best. 💯
English
Codex is better at coding overall but Claude Code is way better at cloning UIs when provided with a screenshot as input.
Experiences from porting a Flutter mobile app to a NextJS web app.
English
@WilliamHolmbe19 If composer 1 takes multiple turns to get it right compared to a zero-shot from GPT 5 Codex, and both are priced the same, does that make composer 1 technically more expensive?
You'd also have to consider your time spent reviewing good code vs really great code.
English
cursors new model, composer, is brilliant when you know exactly what you want
a really good usecase is when an ai code reviewer, coderabbit or bugbot for instance, gives you exact context and instruction of what needs to be fixed
composer is literally done within a couple of seconds
English
Hello my favorite food delivery app @zomato (@deepigoyal),
Give me a "Copy Restaurant Menu as MD" or "Share Restaurant Menu with ChatGPT/other LLM" so that I can ask it for recommendations on what to order based on how I'm feeling or what I'm in the mood for.
It also helps folks tracking calories and macros probably, to some extent.
English
@theodormarcu @windsurf Those are the 2 that get in the way. Appreciate you acknowledging!
English
@baps_patil @windsurf Thanks for flagging this! We're about to ship some updates to GPT 5 Codex prompting to make it better. Are there any other bugs / issues you've noticed?
English
Just 3 days in, Windsurf (@windsurf) just isn't working for me any more. Even with Fast Context.
I'm on the latest v1.12.21 and:
1. It still ran into many edit errors with GPT-5 Codex. It failed to close brackets (minor issue) but also complete core logic (major issue).
2. I'd give it a small to medium task, it would never fully finish it from zero to one. It always asked, "Here are the next steps—would you like me to continue?" up until the 4th shot in the chat for tasks where other IDEs don't do this. Every time, I have to prompt, "Finish all remaining steps, do not ask me for approval for further steps in the next step", and that still did not work.
Back to @cursor_ai Ultra it is for me.
Bapusaheb Patil@baps_patil
Codemaps might just be the best new feature by this AI IDE to help you understand and remember your project's system architecture. Super helpful when your codebase grows. Take a bow, @windsurf! 🙇♂️ First, you knocked it out of the part with DeepWiki which helped understand code at a more granular level. Codemaps takes it a step further.
English
English
.@vercel_support is it possible to use Bun 1.3 during the build?
I see bun v1.2.23 is used and no way to change it.
cc @rauchg
English
English
My @payloadcms x Fumadocs (@fuma_nama) integration example is ready!
Here's what's included:
- Payload CMS Integration: Full headless CMS backend for documentation
- Role-Based Access Control: Owner, Admin and Users
- Hierarchical Docs: Parent/child relationships for nested documentation, with categories
- Lexical Editor: Rich text editing with HTML serialization
- MongoDB Storage: Persistent database storage
- S3 Media: Optional S3-compatible media storage
- Search: Built-in search via Fumadocs + Payload API
- LLM Routes: Auto-generated /llms.txt and /llms-full.txt
- OG Images: Dynamic OpenGraph image generation
English