Nalin Senthamil

the last few weeks feels like a mini breakthrough for AI apps hiding in plain sight - partial autonomy, high agency models, and context eng vs prompt eng. are all incredibly important concepts a lot of these principles were apparent but were just recently articulated (mostly by Karpathy - legend!!). one is the concept of partial autonomy apps - autonomy sliders, app specific UI, and “keeping the AI on a leash” two is the idea of small, high agency models that know how to solve problems vs knowing answers via tool-use / reasoning / multimodality (“maximally sacrifices encyclopedic knowledge for capability”) three is context engineering vs prompt engineering - putting the right stuff into working memory, building around the “jagged intelligence” of models etc these should probably be the starting point for founders thinking through AI native apps Refs: Partial autonomy apps: [Karpathy @YC startup school] High agency mini models: https://lnkd.in/gthCXtMe Context Eng: https://lnkd.in/gha2i7BX

101

10 Comments

The future of voice AI is unfolding on two fronts: 1) Speech-to-speech models that capture not just what we say, but how we feel 2) Developer-first “Voice in a Box” APIs that make conversational apps work As voice becomes a dominant interface layer, the infrastructure behind it needs to evolve fast. The demand is already here: sales agents, schedulers, care providers. Every founder building in this space is building toward a much bigger shift. “Voice AI in a box” platforms will make conversational apps and agents more personal and make developers' lives easier, like getting phone numbers, scaling calls, translation, custom-branded voices, voice cloning, and more. This area is poised to grow fast both now and well into the future as interfaces like avatars become mainstream.   If you’re building here, I’d love to meet! https://lnkd.in/dD9U3jpy

110

21 Comments

Announcing the Krutrim AI lab today! https://lnkd.in/gkrNjfi2 While we’ve been working on AI for a year, today we’re releasing our work to the open source community and also publishing a bunch of technical reports. Our focus is on developing AI for India - to make AI better on Indian languages, data scarcity, cultural context etc. Here’s a list of models we’re releasing: - Krutrim 2 and Krutrim 1 LLMs: While Krutrim 1 (India’s first LLM) was launched in Jan 24, it was a basic 7B model. We’re launching Krutrim 2 today as a much improved model. More here: https://lnkd.in/gEC-rzc2 - Chitrarth 1: India’s first Vision Language Model built on top of Krutrim 1 capable of understanding images and documents. More here: https://lnkd.in/gaaqpVKF - Dhwani 1: India’s first Speech Language Model built on top of Krutrim 1 capable of tasks like Speech translations. More here: https://lnkd.in/guj8GAFQ - Vyakhyarth 1: State of the art Indic Embedding model for use cases like Search and RAG. More here: https://lnkd.in/gxNbmVgA - Krutrim Translate 1: State of the art text to text translation. More here: https://lnkd.in/g32RkkNM In addition, since there was no global benchmark for Indic performance, we’ve developed “BharatBench” and the technical report is here: https://lnkd.in/gA6YAxpz We’ve also published a bunch of technical reports and papers here: https://lnkd.in/g4nSbXvX Also announcing India’s first GB200 deployment in partnership with NVIDIA! Will be live by March and we will make it the largest supercomputer in India by end of year. We’re nowhere close to global benchmarks yet but have made good progress in 1 year. And by open sourcing our models, we hope the entire Indian AI community collaborates to create a world class Indian AI ecosystem. We’re still learning to walk before we can run, hopefully within this year! All our open source work here: Web: https://lnkd.in/gm4iKJHa GitHub: https://lnkd.in/gP8gQi46 Huggingface: https://lnkd.in/gJfNiffb Also, announcing an investment of ₹2,000 Cr today into Krutrim and a commitment of ₹10,000 Cr by next year!

2,210

164 Comments

Introducing Agentic Culture — an open-source experiment in collective memory and in-context cultural accumulation for multi-agent systems. A hot topic from the recent Andrej karpathy // dwarkesh podcast is that LLMs don't yet have culture. >>> So we built the scaffolding for them to develop one. Agno Agents can now tap into a shared cultural memory — knowledge that persists beyond individual sessions, users or memories. Culture is collective, evolving context that shapes agent behavior over time. Now they can create, read, explore and learn from their culture. You can use the `CultureManager` class to create and manage `CulturalKnowledge` objects. These are stored in your database of choice (SQLite for now) and automatically retrieved as in-context cultural grounding for your agents. To give your agents access to shared culture, just set `add_culture_to_context=True`. That's it. Agents now learn from collective experience. Explore the code and README: https://lnkd.in/e7E3AcJX Agno is open source — come build with us: https://agno.link/gh

95

3 Comments

Some common observations on the situation of AI coding with Tech Leads/CTOs -- YMMV as the tech leads/CTOs I talked with are mostly from mid-scale startups where the software products have certain complexity... 1. Claude and GPT-5 agents are the best. Everyone thinks they're very useful. 2. AI agents magnify the ability of individual developers -- if you were a 10x developer, with AI agents you're 100x better at code. If you were 10x worse, with AI agents you produce 100x worse code. 3. Yet, I asked if it helps deliver faster from a PM POV: everyone says no - maybe 10%. 4. But most still think it's useful because it helps achieve ~100% test case coverage and makes repeated refactoring possible. 5. Everyone now feels like they would avoid hiring junior developers as much as they can -- one described to me "The quality of code is no better than AI but they require so much management and emotional support from me." Everyone agrees it's also very useful for quick prototypes or proof-of-concept code as well. We observe the same at Oursky/FormX/Authgear, but that kept me wondering -- while quality of code is important, how does it impact business value if it can't make us deliver faster? What's your experience? Share your thoughts below.

33

11 Comments