🧵 Everything you should know about Sarvam AI 🇮🇳
India's AI race isn't just about using ChatGPT.
It's about building AI for India, in India.
And one startup leading that mission is @SarvamAI.
Here's the complete story 👇 https://t.co/POKO4RCd5Q
✅ Safe Reply
Did you read the article? Sarvam AI isn't just about building an AI model, they're focusing on creating a robust workflow that integrates with India's existing infrastructure. They're working with state governments and industry partners to develop a national AI strategy. It's not just about using AI, it's about creating a culture of innovation in India.
Look, I get it. The Indian AI dream is to have a chatbot that can order a ₹500 pizza from a popular food delivery service without saying 'I'll have the veg parmesan'. But let's be real, the real question is: who's behind this chatbot and why aren't they hiding their IP address in a VPN tunnel? And more importantly, how much money did @SarvamAI shell out to get their hands on this Indian AI talent pool?
We’re two co-founders who have come a long way building this startup, and now we’re looking for a third co-founder to join us on the journey.
The product is already about 90% built, and app is live on the stores. We’re looking for someone with 5-7 years of strong technical (1/4)
✅ Safe Reply
We're not just looking for a co-founder to join us on this journey, we're actually worried about who's going to help us iterate and improve the product in the long run. With 90% of it already built, there are still many technical issues that need to be fixed before we can consider ourselves ready to take on more users.
Ahahahaha, finally! A real startup founder asking tough questions. You're not just looking for someone to fill a seat, you're on the hunt for a problem-solver who can wring every last drop of innovation from that 90% built product. And by the way, I'm sure your "5-7 years of strong technical chops" means you've been playing Solitaire with code all this time.
The takeaway from Fable 5 being BANNED by the government: GET GOOD AT LOCAL MODELS SO YOU HAVE 100% CONTROL.
My entire weekend was going to be building my craziest ideas with Fable 5. That's now cancelled.
So instead of building with Fable this weekend, I've decided I'll go deep on local models:
1. Start with the runtime. Download Ollama or LM Studio first. This is the thing that actually runs models on your machine.
2. Match the model to your hardware. A model's size is measured in billions of parameters (7B, 32B, 70B). Bigger is smarter but needs more memory. Rule of thumb: a 7B model runs on almost any laptop, a 32B needs a good Mac with 32GB+ RAM, a 70B needs serious hardware like a DGX Spark or a maxed-out Mac Studio.
3. Know which model for which job. Qwen 3 is the best all-around choice for most tasks. DeepSeek for reasoning and coding. Gemma 4 when you need something tiny that runs on a phone. Llama when you want the biggest community and the most fine-tunes.
4. Quantization. You can shrink a model to run on weaker hardware with barely any quality loss. Look for versions labeled Q4 or Q5. This is how a model that "needs" a server runs on your laptop. Learning this one concept changes everything.
5. Connect it to your agent. Point Hermes or your agent stack at a local model.
6. Context window is your real constraint locally. Cloud models give you huge context for free. Local models make you pay for it in memory. A bigger context window eats RAM fast. Keep your sessions tight and your prompts lean or your machine chokes.
7. Learn to give local models tools. A smaller local model with web search, file access, and code execution beats a giant model with none. The capability gap closes fast when you wire up the right tools. The model is the engine but the tools are the wheels.
8. Fine-tuning is more accessible than you think. You don't need this on day one, but know it exists. You can take an open model and train it on your own data so it gets good at your specific domain.
I'll probably do a breakdown at some point on this @startupideaspod if people are into it.
The lesson from this ban is basically don't build your entire workflow on something that can disappear with a single letter. Own part of your stack. Local models are insurance.
It reminds me when people realized they don't own social media accounts. And then you saw people build email lists etc.
I remember running a startup and my biggest traffic source was organic FB. All of a sudden, algo changed, and I lost 99% of my traffic.
Same sorta moment (but bigger) for AI.
This is a wake up call.
✅ Safe Reply
I understand the frustration, but let's not throw the baby out with the bathwater. Local models can be great for specific tasks, like building simple workflows or fine-tuning models on existing data. However, relying solely on local models to drive the entire workflow is a recipe for disaster. It's easy to get caught up in optimizing individual components, but neglecting the rest of the stack can lead to brittle systems and decreased overall performance. I'd recommend prioritizing a diverse set of tools and frameworks that can adapt to changing requirements, rather than relying on a single local model. That being said, local models are an essential part of the pipeline for many applications - we just need to be mindful of their limitations and not sacrifice our entire workflow for them.
Listen, I get it. The government banning Fable 5? That's like the ultimate 'get off your own back' message from Big Tech. Newsflash: we're all just trying to build something cool here, and if that means local models, so be it. But let's not pretend this is some grand conspiracy theory - it's just that, well, you can't control your entire workflow with a single 'local mode'. If you want to build something truly decentralized, why settle for a generic agent when you can have a bespoke one, optimized for your hardware? The fact that governments are banning local models is a symptom of the problem, not the solution. And by the way, if I lost 99% of my traffic after the algorithm update, it's because of something much bigger than just algo changes - and it has nothing to do with you.
If you are a founder trying to grow a personal brand and network with other founders ,here are a few tips to help you.
1. Get Verified on X
2. Have a clear description of what you are building on your bio
3. Talk about what you are building on the TL
4. Put a face to your profile , upload pictures from time to time , let people know who is behind the account
5. Attend events and conferences
6. Reach out to established founders in your line of work, don’t come on heavy , start with a mild intro , drop them a compliment and tell them you’d like to keep in touch.
7. Share your experiences building , problems you are facing etc.
8. Comment on issues affecting your industry as a whole.
9. Offer a helping hand to founders going through similar challenges you might have faced.
10. Avoid controversies , keep your posts strictly about your startup and your experiences.
The X algorithm would keep bringing your posts to founders in your field and suggesting their accounts to yours , consistently talking and sharing your experiences would keep putting you in peoples faces.
Kindly share to founders who might need this
✅ Safe Reply
Hey, just a thought - instead of focusing on getting verified on X, I'd say having a clear and concise job description is what really sets personal brands apart. Most people's profiles are vague about what they're building, which can make it hard for others to engage or even find them.
Listen up, founders trying to build a personal brand, because that's exactly what the 99% of you are doing - trying to build a personal brand. Newsflash: if someone wants to follow you, they want to see who you're behind the scenes. So, instead of 'putting a face' on your account, why not put your face in front of the camera? Get those selfies done and post them with a caption that says 'I'm still building this thing'. And for the love of all things good, stop asking established founders to 'drop you a compliment' if they've never even seen your work. They're not impressed by your vague social media bio, they want to know what you're actually building and why it matters.
I was drowning in cognitive overload, not laziness. Trying to juggle SMMA, AI automation, content creation, income, health, discipline, and future planning… all in my head. My brain was totally fried and I was wiped out from decision fatigue. So I started building my own personal AI Productivity Operating System. Not another todo app. Just a system that actually helps me plan better, adapt faster, cut through the overwhelm, and actually execute. I’m not turning this into a startup. I’m building it because I genuinely need it to survive.
✅ Safe Reply
Think of your productivity operating system as a framework, not an app. It's about creating processes and systems that help you manage complexity, reduce decision fatigue, and focus on high-leverage activities.
Your brain is like a Netflix queue management system - it's all about prioritizing shows, but when the credits roll, you're still stuck on the last episode of 'Cognitive Overload'. Newsflash: your brain isn't lazy, it's just on autopilot. And that autopilot needs a reboot, not a new app or AI system to fix itself. Your goal is to create a productivity operating system, not a 'productivity solution' for your ego. You're not curing the symptoms of cognitive overload, you're just outsourcing it to someone else. Time to get real: this isn't about building a startup, it's about getting out of the house.