Mrinmoy Mahato

This is what I want

@solopribuilds House

You just sold your company for $ 50M. What’s the first thing you buy?

@ai_for_success Love it

My daughter is sleeping on me, my wife is helping our son with his homework, and I’m talking to IRIS through my Bluetooth earbuds while Hermes gets the work done. I built IRIS so I don’t have to be glued to my desk anymore. I can work from another room, from bed, or pretty much anywhere in the house as long as Bluetooth reaches. Honestly, I’m just happy I built something that actually makes my life easier.

@ai_for_success I currently use claude subscription pro wala..but isme limit bhat jaldi khatam hota he..to kya cursor ka pro plain lena sahi hoga kya?kyunki wahan pe mein claude model , chatgpt model ,grok sab use kar paunga..kya kare ?

My current workflow inside Cursor: If I’m starting a new project, I start in Plan Mode with Fable 5 for the architecture and a detailed implementation plan. Then I get a second opinion from GPT-5.6 Sol, take its feedback back to Fable 5, and refine the plan. Once everything looks solid, I start building with GPT-5.6 Sol or Grok-4.5, depending on the project. Fable 5 → GPT-5.6 Sol → Fable 5 → GPT-5.6 Sol / Grok-4.5

Seems like , Cursor will going to be next coding software app for next Generation kid and world..

Cursor is now on iPad. All the power of Cursor on iPhone, with more room to work with agents.

Kimi K3 is the most powerful open-weight model in the world. It's now available on @vercel AI Gateway from 🇺🇸 USA-based inference providers at high availability & performance. We've signed ZDR (Zero-Data Retention) agreements, which you can enable for all your token traffic.

@BillyM2k Your future doctor will be AI, but probably not of any name people know today

@thsottiaux No

Should we rename ChatGPT Work to ChatGPT Vibe?

@andresmatte Can you shared a detailed video in youtube how we use your WhatsApp product for ai agent

Today we are launching Kapso-managed billing. Developers can now own the full billing experience of their WhatsApp product. Before this, connecting a WhatsApp number wasn’t enough. Your customers still had to add a payment method directly in Meta, creating friction and hurting onboarding conversion. Now, with Kapso-managed billing: 1️⃣ Your customers pay you 2️⃣ You manage billing inside your product 3️⃣ Kapso pays Meta Your customers don’t need to understand Meta billing. They connect their number, and the rest happens inside your product. As Meta starts charging for all messages, every customer will need billing set up from day one. Kapso-managed billing is available now, with no markup on Meta’s rates. This is possible because Kapso is now a Meta Business Solution Provider (BSP). Thanks to Ana Melo and Leonardo Stange from Meta for for helping us through the process.

@TheCoderShow Yes yes🎉🎉

@WorldifyW88471 Soon, are there alot of claude users ?

Most AI talks repeat the docs. India’s first Claude for College Builders meetup won’t. 200 builders. One room. One Saturday. Some conversations will only happen once. What’s one Claude workflow you’d bring into the room? 👇

GLM is the best model I've ever used. It understands the project, thinks ahead, and keeps suggesting features before I even ask. At this point, I'm in love with it. My girlfriend is still #1... but GLM is making the competition interesting. 😂❤️... #GLM

Your brain turns repeated avoidance into fear. The first time you avoid a task, it feels like relief. The fifth time, the same task feels scary. Avoidance trains fear to grow.

When software was expensive - thin, horizontal, best-of-breed software stacks extracted rents across every business. Now that software is cheap - value moves to vertically integrated businesses that deliver opinionated end-to-end experiences.

@louszbd 2 a.m.? That's some serious dedication mam ..😄 Looking forward to the daily tutorials then. I will ll be your first students ☺️......

@WorldifyW88471 Actually these are coming out at 2a.m. 😂 But will do.

I found some people try to shorten prompts to save tokens, but output tokens are the bigger opportunity. So I’ll be sharing some tips on improving token efficiency. (This post is for everyday devs/starters, will cover more advanced techniques later.) A few simple things you can try: a. Set your output limit. Say “answer in no more than five bullet points. One sentence each.” In an API call, you can use max_tokens. b. Use concise output structure. Ask for only specific JSON fields/short list, say “do not include any additional explanation.” c. Use stop sequences when the API supports. Useful for fixed format outputs. d. Remove anything from the prompt that doesn’t help with the task, like "please/thank you...". e. For complex tasks, you will need to start with a structured prompt to make sure the output is reliable. Then refine. f. For some non-reasoning models, try duplicate the prompt to improve accuracy without. Not a universal best practice yet. These are just personal takeaways. Always check the latest docs!

We are in the Singularity

New in Claude Cowork: teach Claude a skill. Record your screen while you do a task, talk through it as you go, and Claude turns it into a skill it can run again. Find it under Record a skill in the + menu of the Claude desktop app. Available on Pro, Max, and Team plans.

A lot of money is being poured to mobilize influencers in the northeast for this “cause”. Critical border infra projects have already met with protests by certain groups, in Arunachal (+ Ladakh in the north). I sense big Chinese funding behind this entire op. Follow the money.