Ex-Google engineer just compressed the shift from AI agents to "graphs" and "loops" into one 2h47m lecture: • 00:00 - understanding the different layers of AI memory • 30:00 - moving from simple LLM calls to AI agents • 50:00 - why agent systems are becoming a graph engineering... problem • 1:12:00 - connecting context, tools and agent state • 1:24:00 - building a minimal agent harness from scratch • 1:46:00 - breaking down the architecture behind Hermes Agent • 2:08:00 - turning agent concepts into working systems 167-minute deep dive, and one of the clearest ways to understand how AI engineering is moving beyond prompts The progression: Prompt → Context → Memory → Agent → Loop → Graph → Agent Harness Watch it today, then read the full roadmap belowshow more

from prompts to graphs is the shift everyone needs to understand

yeah

graphs sounds useful. Is there a link or repo people can try directly?

it's looking so good. full lecture about graphs

corporations haven't cracked this yet either, but at least one guy's making progress on the how-to part

The progression is useful. I'm building Ovandor around the gap on the human side too: retrieved context still has to be valid for the decision in front of you. A graph can show relationships, but it does not make stale assumptions current.

2h47m to explain that everything is just a graph now. could've been a meme.

The 1:12 bit, tools plus agent state, is where mine falls over. It keeps calling a tool with a value the user corrected an hour earlier, because the correction lived in chat and the state never learned it. Does the lecture show where a correction gets written down?

The prompt to graph jump is where it gets serious :)

graphs and loops explain it memory layers first

2h47m is a lot for the graph vs loop distinction. does he show it breaking in prod or just the clean run

both graphs and loops in 2hours, best guide

Topology and loops are the easy part to draw. The hard part is holding one agent's focus across a long run and checking what it produced. No clean graph diagram hands you either. That is where most agents quietly come apart.

Saturday you wire three agents into a loop because one call kept dropping the plan. Monday the usage tab is 8x. Nobody budgets the context tax: each hop re-reads the same 6-12k of memory before a tool even runs.

do you also this video on YouTube?

167 minutes, that's a real commitment 😅

ahaha, yeah

Graphs deserve that kind of patience

I've been thinking about agents more as systems than prompts lately.

The shift to graphs and loops is exactly what separates "demo agents" from "production agents." In robotics, this is the difference between a robot that does a task once and one that can recover from a physical slip in real-time.