Links
- CodeCrafters (Sponsor): https://tej.as/codecrafters
- The Rise of the AI Engineer (Latent Space): https://www.latentspace.com/p/ai-engineer
- Anthropic Model Context Protocol (MCP): https://www.anthropic.com/news/model-context-protocol
Anthropic Skills (SKILL.md): https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf?hsLang=en
- Pratim Bhosale on Vector Search (Previous Episode): https://www.youtube.com/watch?v=ZJcrI1GbDec
- OpenClaw (Official Repository): https://github.com/openclaw/openclaw
- WormGPT/Prompt Injection Incident (The Verge Reference): https://www.theverge.com/ai-artificial-intelligence/881574/cline-openclaw-prompt-injection-hack
Summary
In this episode, I’m finally breaking down what AI Engineering actually looks like in 2026. We’ve moved past the "wow" phase of demos into the hard reality of production, where models lie, costs spike, and security is a nightmare. I define the role formally—separating it from ML engineering—and dive deep into the three core problems we solve daily: hallucinations, real-time knowledge, and context engineering.
I also explore the explosion of agent runtimes, specifically dissecting OpenClaw (formerly Clawdbot/Moltbot) and why its open-source, device-local approach is winning. We talk about the Model Context Protocol (MCP), the new SKILL.md standard from Anthropic, and why I believe multi-agent concurrency is the inevitable future of software.
Chapters
00:00:00 Intro: The "Monday Morning" AI Nightmare
00:08:15 What is AI Engineering?
00:21:00 AI Engineers vs. ML Engineers
00:36:20 Problem 1: Hallucinations
00:48:30 Problem 2: Real-Time Knowledge
00:59:00 Problem 3: Context Engineering
01:13:00 Agents
01:17:30 MCP (Model Context Protocol) Explained
01:22:30 SKILL.md: The New Standard
01:39:00 The 2026 Trend: Multi-Agent Concurrency
01:42:30 Build Your Own OpenClaw
01:52:00 Conclusion: Moving from Talk to Action
Hosted on Acast. See acast.com/privacy for more information.
