Data Engineer building AI pipelines 🚀 | Sharing AI & Data insights in simple language | Startup enthusiast
Joined May 2025
- Tweets 778
- Following 111
- Followers 30
- Likes 2,092
It solved the ui problem which we faced in the terminal .
Wow, I think this is something unique to @opencode
Session Timeline view
When you select each checkpoint in the session's timeline, it shows a preview too
OpenCode nailed this checkpoints UX 🔥
Impressive performance by the GPT-5 and gemini2.5 pro on IOAA exam
GPT-5 and Gemini 2.5 Pro hit gold-medal-level on past IOAA exams
- against humans, both place among the top students; GPT-5 tops the best student in 2022, 2023, 2025; Gemini does in 2022, 2023
2025 results:
- theory: GPT-5 1st, Gemini 2nd (vs students: 1st & 2nd)
- data analysis: GPT-5 1st, Gemini 2nd among models (vs students: 6th & 10th)
But but... LLMs are dead ends
Yuvraj singh Bhadoria retweeted
BREAKING 🚨: Anthropic released "Imagine with Claude", where Claude can generate software on the fly!
This is the future of the internet 👀
Sadly, "Available to Max users for 5 days"
We're also releasing a temporary research preview called "Imagine with Claude".
In this experiment, Claude generates software on the fly. No functionality is predetermined; no code is prewritten.
Available to Max users for 5 days. Try it out: claude.ai/imagine
ChatGPT already helps millions of people find what to buy. Now it can help them buy it too.
We’re introducing Instant Checkout in ChatGPT with @Etsy and @Shopify, and open-sourcing the Agentic Commerce Protocol that powers it, built with @Stripe, so more merchants and developers can integrate agentic checkout.
Yuvraj singh Bhadoria retweeted
Claude Code pro tip:
use the new /rewind command to roll back code changes instantly.
Model messes up? no biggie, just rewind it back to when it worked!
Yuvraj singh Bhadoria retweeted
DeepSeek-V3.2 shows:
- Chinese chips are rising: Day-0 support for Huawei Ascend & Cambricon;
- ML compiler: DeepSeek uses TileLang, letting you write Python → compile to optimized kernels on diverse hardware. E.g., 80 lines of Python can reach 95% of FlashMLA’s (CUDA written by hand) perf.
Under the hood of TileLang is TVM, an ML compiler I spent years working on with the great open-source community. As the hardware landscape diversifies (Nvidia GPUs, Chinese chips, and inference-focused chips), ML compilers will shine again.
Yuvraj singh Bhadoria retweeted
Sora 2 has great voice, and face consistency.
You can even tag real people like @sama, and turn them into a funny video. I might actually use this app.
Yuvraj singh Bhadoria retweeted
Every day,
OpenAI serves millions of requests for AI models like ChatGPT. Handling this scale isn’t just about having powerful GPUs. It’s about a carefully engineered system that combines distributed compute, optimized storage, and intelligent caching. Let’s break it down.
Check out Hunyuan3D-Part!
P3-SAM: Segment 3D meshes into parts
XPart: Generate high-fidelity, structure-aware parts
Upload .glb/.ply/.obj & watch AI turn your models into detailed components!
Yuvraj singh Bhadoria retweeted
This is an incredibly good agentic model. We look forward to supporting it on Comet for all browsing agent tasks pretty soon.
Yuvraj singh Bhadoria retweeted
#OperationSindoor on the games field.
Outcome is the same - India wins!
Congrats to our cricketers.
