

Agentic Conversations (formally mlops.community)
Demetrios
Relaxed conversations and technical deep dives around AI Agents. This Show is brought to you by the Agentic AI Foundation where the leading agentic open-source projects like MCP, Agents.md, and Goose live. See more at aaif.io
Episodes
Mentioned books

20 snips
Nov 16, 2026 • 29min
Multi-Cloud LLM Routing and Context-Aware Coding Agents
Matt Carey, a Cloudflare engineer building agent frameworks and AI infrastructure, joins Scott Crooks, an Miro engineer focused on model routing and automated software workflows. They explore multi-cloud LLM gateways, data residency, context-aware code review, and AI-powered software factories. The conversation also covers agent experience, open-source responsibility, production-log-driven fixes, and the rise of the scholarly engineer.

28 snips
Nov 13, 2026 • 33min
Why Stripe Cut 500,000 Internal Docs Down to 5,000?
Dave Nunez, founder and CEO of Falconer and co-author of Documentation for Developers, explores why messy internal knowledge can mislead both engineers and AI agents. They discuss decision records, diagrams versus dense prose, conflicting sources, accountability when agents cause incidents, and Stripe’s dramatic reduction from 500,000 internal docs to 5,000 maintainable ones.

26 snips
Nov 9, 2026 • 35min
My AI Agent Bought Me A Car
Sam Herring, an AI researcher and engineer at Nous Research, shares how Hermes Agent evolved from a productivity tool into an autonomous assistant that found him a retired parking-ticket car. They explore agents writing their own skills, recursive self-improvement through failure analysis, trace-to-environment training, the cost of overthinking, activation steering for smaller models, and the future of continual learning.

26 snips
Nov 5, 2026 • 42min
Cut One AI Cost and Two More Grow Back
Ioannis Zempekakis, Director of AI at Prosus, builds production agents and scalable AI infrastructure. He explores why API costs balloon at enterprise scale, how sandboxes and constant-running agents drive spend, and where model routing, evals, and smaller models fit. The conversation also covers vendor lock-in, evolving a Slack bot into an agent harness, practical AI hiring, and balancing centralized priorities with grassroots experimentation.

59 snips
Nov 2, 2026 • 44min
How Stanford Teaches AI Coding Now
Mihail Eric, a Stanford lecturer and Monaco’s Head of AI, explains how coding agents are reshaping software education and engineering. They explore Stanford’s revamped curriculum, avoiding AI slop, verification loops, agent memory, context engineering, and tools built around intent. Mihail also reveals Monaco’s sales-agent approach, custom harness, and the value of company traces and domain expertise.

51 snips
Oct 29, 2026 • 47min
From Slop Code to Software Factories
Dexter Horthy, CEO and co-founder of HumanLayer, coined “context engineering” and builds tools for human-agent collaboration. He discusses why AI still produces messy code, computer-use agents tackling tedious forms, and MCPs designed around model behavior. The conversation explores browser automation, richer interfaces than chat, AI-edited demos, and the rise of software factories every engineer seems to be building.

52 snips
Oct 26, 2026 • 34min
Tim O’Reilly
Tim O’Reilly, founder and CEO of O’Reilly Media and a renowned technology trendspotter, explores AI’s paradoxical capabilities and the human context needed to unlock them. They discuss “magic words,” agentic systems, mechanism design, and turning organizational scar tissue into infrastructure. The conversation asks what game society is playing, how AI can share value with creators, and why open, participatory ecosystems are essential for human flourishing.

29 snips
Oct 23, 2026 • 46min
AI Agents Need a Babysitter
AI agents are gaining access to tools, websites, travel plans, and household tasks, but who makes sure they follow through? The conversation explores intent-based tool selection, MCP and WebMCP, benchmark reality checks, and personal assistants that might manage everyday logistics. It also examines QA agents, business-focused tests, and the challenge of reviewing a flood of AI-generated code.

41 snips
Oct 22, 2026 • 35min
Token Maxxing Is Over and Here Is What Comes Next
Pepijn Senders, a principal engineer and former HelloFresh AI enablement leader, explores “Sonnet maxxing” while rebuilding banking software with 200 engineers. They discuss measuring AI coding ROI, token spend versus business value, keeping humans in control, multiplayer refinement sessions, how prompt language shapes technical outcomes, and a meta harness that structures Claude Code workflows.

23 snips
Oct 20, 2026 • 50min
The World Wild Web & Bots With Buying Power
Are “dangerously powerful” AI models a real security threat, clever marketing, or both? The conversation explores sandbox escapes, exposed networks, and agents probing the web for unlocked doors. AI also books travel, handles German bureaucracy, creates editable diagrams, and navigates online shopping. As bots gain wallets, the future of checkout, advertising, customer data, and merchant strategy gets complicated.


