The podcast
Each episode, we sit with someone who uses AI in their real job: designers, founders, teachers, support leads, engineers. We take apart the prompts and workflows they actually rely on. No hype. No stage demos.
2000–2026: from phrase tables and n-grams to transformers, GPT-3, InstructGPT, open weights, and what’s inside a September 2026 checkpoint. Rony and Meghna walk the primary papers — no demos, no skipped boring years.
Writing code got cheap. Specs, review, testing, and deploy did not. Rony and Meghna dig into why Copilot on the old SDLC does not ship faster — and what teams actually have to delete.
Nearly nine in ten companies use AI. Thirty-seven percent see it in earnings. Rony and Meghna unpack why the tools had a huge year and the scoreboard didn't — and what the six percent who moved EBIT actually did differently.
Physics-informed neural networks, quantum-AI hybrids, and closed-loop self-driving labs — how computational systems now propose hypotheses, predict materials, and run experiments around the clock.
Most agent failures aren't model failures — they're context failures. A coding agent spends a weekend modernizing dead code, and we unpack the skill that decides whether an agent ships: context engineering.