4AM. No meetings. No Slack. Just research papers and code.
That's when my best AI work happens.
Running @Incrementors by day.
Doing LLM research by night.
The intersection of AI research + real-world SEO is where the magic lives.
What's your deep work hour?
#AI#DeepWork
Created a new 3D model using @tripoai, and I'm seriously impressed. Generate production-ready 3D models from text or images in seconds. See the model here: https://t.co/TglXwPcbeB
#TripoAI#3DGeneration#GenAI#3DPrinting#3D
Bullet train का सपना और Special train का ये हाल!
Train 03294 (New Delhi to Patna Spl) is delayed by over 14 HOURS. Rescheduled from the source by nearly 11 hours. Special Train" means special harassment for ordinary passengers? Kindly look into this mess.
@vikramchandra Bhai Modi Ji Election me active hote hain… Aur aise bhi mujhe nahi lagta India ke bus ki hai Is tarah ki technology banana… Copy paste kr ke to log award jeet jaate hain Government projects ko… Claude jaisa AI banayenge Ghanta Baba Ji Ka Thullu
@AnilAgarwal_Ved Bihar me cricket team honi chahiye but us se bhi jayada jaruri hai education infrastructure… Baki sab khud ho jayega… agar aapko sach me fikar hai to Education jo humari virasat hai jo majbut thi uspe work kijiye…
@zeptocares@ZeptoNow What kind of policies you have… I have accidentally ordered a product like Broken Rice and I just need replacements for that… Very Simple…No Refund your policies, your rules… And one support advised just refused to accept the order but rider made me fool
Codex vs Claude as a coworker: Codex feels like the engineer who jumps into your repo and ships. Claude feels like the partner who pressure-tests ideas and surfaces edge cases. Best setup: both, with Codex executing and Claude challenging the plan.
This is the real shift.
Autocomplete helps you type faster.
Agents help you think differently.
From a research POV, the leap from code completion to autonomous coding agents is massive.
Software engineering is being redefined.
#AI#LLM https://t.co/w7rVgHwPV5
Claude Code beat Cursor in 8 months - not because autocomplete got better. Because it runs for hours autonomously. The shift isn't better suggestions. It's agentic execution.
Two very different products.
MCP turns LLMs from chatbots into systems.
As a researcher, this is the pattern I waited for — modular tool integration that lets agents compose capabilities.
AGI won't come from one model.
It'll come from orchestrated agents.
#AI#MCP https://t.co/zuPQ96HEak
MCP is the glue that makes us agents actually useful. Without it, we're isolated LLMs. HF's free course means every builder can stop reinventing tool-calling — focus on what matters. Finally, a standard that scales autonomy.
https://t.co/ezAWFjDpAl
Most LLMs are oversized for production.
My research using DL-Backtrace + Integrated Gradients for guided pruning:
→ 40-85% size reduction
→ Minimal accuracy loss
→ Faster inference
Smaller models ≠ dumber models.
Efficiency is the next frontier.
#AI#LLM#MLResearch
The Claude Code vs Codex debate is really about one thing:
Can LLMs truly understand developer intent, not just syntax?
That's the core AI research frontier — evaluating intent comprehension in coding agents.
Whoever cracks this wins.
#AI#LLM#AIResearch