Why Two IIT Engineers Turned Down $550K Jobs To Build A Startup
The video recounts how two IIT‑trained engineers turned down a combined $550,000 salary package to found GigaML, a startup building AI‑driven customer‑support agents. After a surprising Y Combinator interview where partner HJ dismissed their ed‑tech idea, the founders pivoted to fine‑tuning large language models for enterprise support. Leveraging research from Stanford and a $4 million seed round, they built a platform that boosts call deflection from the industry‑average 10‑15 % to 60‑70 %, aiming for 90‑95 %. Their first marquee client, Zepto, led to a DoorDash contract won by an eight‑person team, proving that performance can outweigh size. The founders cite a 50 k earnings from Kaggle competitions and a rejected $550 k quant job as proof of their engineering credibility. The story underscores how early‑stage risk‑taking, YC mentorship, and a clear AI value proposition can enable tiny teams to secure Fortune‑500 deals, reshaping talent decisions for ambitious engineers.
Inside YC's AI Playbook
The episode reveals how Y Combinator has transformed from a pre‑AI organization into an AI‑native one by constructing an internal agent framework that runs on a single PostgreSQL data warehouse. Founder‑partner Pete Kumman describes the evolution from a finance‑focused...
Why Zepto's Aadit Palicha Turned Down Stanford to Deliver Groceries
Aadit Palicha’s decision to forgo a Stanford education in favor of building Zepto is the centerpiece of the talk. He and co‑founder Keville began during the pandemic by coordinating grocery deliveries through a WhatsApp group, then evolved the concept into...
How Razorpay Became India’s Largest Payments Company
The video chronicles how Razorpay, founded by Harshil Mathur, grew from a college‑side project into India’s largest payments platform, highlighting its Y Combinator entry in winter 2015 and the regulatory hurdles that shaped its trajectory. Initially the team tried to sell...
Inference Chips for Agent Workflows
The video highlights a growing mismatch between conventional AI hardware and the emerging class of agentic AI workloads. While most inference chips are optimized for a simple prompt‑in‑response‑output pattern, autonomous agents execute long, branching loops that call external tools, maintain...
AI-Native Discovery Engines
The video introduces AI‑native discovery engines, a new paradigm that moves scientific research beyond the traditional hypothesize‑experiment‑interpret loop toward fully automated, closed‑loop cycles powered by advanced foundation models. Frontier models now perform at PhD‑level on scientific reasoning benchmarks, enabling them to...
The AI Operating System for Companies
The video introduces an "AI operating system" concept that makes an entire enterprise legible to artificial intelligence. By capturing every meeting, ticket, code change, and customer interaction, companies can shift from an open‑loop decision process—where outcomes are reviewed weeks later—to...
SaaS Challengers
The video argues that generative‑AI coding is dismantling the traditional SaaS model, prompting investors to slash billions from legacy software valuations while simultaneously creating a fertile ground for new challengers. AI can shrink software development costs by more than a hundredfold,...
Software for Agents
The video argues that the next trillion users of the internet will be AI agents, not humans, and that today’s software—designed for clicks and forms—is ill‑suited for autonomous operation. It calls for a shift toward building tools that agents can...
Hardware Supply Chain
The video highlights a widening gap in hardware development speed between the United States and China, arguing that iteration time—not just raw supply‑chain capacity—is the decisive advantage for Chinese manufacturers. In Shenzhen, a team can move from concept to a physical...
Recursion Is The Next Scaling Law In AI
The Decoded episode spotlights recursion as the emerging scaling law in AI, focusing on two 2025 papers—Hierarchical Reasoning Models (HRM) and Tiny Recursive Models (TRM)—that demonstrate how repeated inference steps can boost reasoning performance without simply enlarging model size. The hosts...
Dynamic Software Interfaces
The video argues that today’s software still presents a one‑size‑fits‑all interface, even as users demand personalized experiences akin to Netflix’s content curation. It posits that advances in AI‑driven coding agents now allow end‑users to act as their own forward‑deployed engineers,...
How to Build the Future: Demis Hassabis
Demis Hassabis, DeepMind CEO, outlined the current roadmap toward artificial general intelligence, emphasizing that while large‑scale pre‑training, RL‑HF and chain‑of‑thought have propelled capabilities, core ingredients such as continual learning, long‑term reasoning and robust memory systems are still missing. He positioned...
AI-Personalized Medicine
The video outlines how intelligent agents are driving a new wave of personalized medicine by integrating diverse health data sources—from genomic scans to wearable metrics—into precise, patient‑specific recommendations. It highlights two cost‑driven revolutions: genome sequencing prices are falling faster than Moore’s...
AI for Low-Pesticide Agriculture
The video outlines a looming crisis in modern agriculture: pervasive pesticide residues in food, water and soil, coupled with evolving weeds and pests that erode the effectiveness of traditional chemicals. Farmers are trapped in a feedback loop—spraying more to combat resistance,...