Posts

Are Small Language Models the Future of Agentic AI?

Image
 The AI world has placed its bets on large language models (LLMs) as the default brain for AI agents. This bet has become so entrenched that entire infrastructures, venture bets, and enterprise rollouts have been built around this paradigm. But what if this default isn’t the only option? During a recent weekend reading session, I stumbled upon a paper from NVIDIA Research and the Georgia Institute of Technology: “ Small Language Models are the Future of Agentic AI .” The paper argues that for many tasks, the future is not LLM-only, and after interrogating its claims, I believe it’s not SLM-only either. The future is hybrid. At its core, the paper suggests that small language models (SLMs) are powerful enough to run on consumer hardware with fewer than 10B parameters. It makes a compelling case based on three key points: SLMs are already powerful enough for most agentic tasks. They hallucinate less and are easier to constrain, making them operationally more reliable. They are relati...

Statistical ML in the Age of LLMs - A Real-World Playbook

Image
  Introduction Does statistical machine learning still make sense in the age of LLMs? This isn’t an academic question, it shows up in product meetings, technical brainstorming, and post-mortems. Someone proposes, “Can we explore GPT/Claude/LLaMA or an open-source model or latest paper,” and the room tilts toward the shiny option. The clout around LLMs is real: they can feel like the easiest answer to every problem because the rapid pace of technical content and the flood of new tools creates FOMO. That pressure seeps into our technical decisions. I want to ground this in personal experience . In my product meetings, the proposal to use LLMs comes up all the time. Even when we have proven, homegrown models, the allure of LLMs often overshadows rational thinking. This is where you have to really dig in, understand the nitty-gritty, and validate the problem and solution from both a macro and micro perspective. That paragraph isn’t a lament, it’s the setup for this post. This isn’t n...