Загрузка видео...
Не удалось загрузить видео
Google just released a 1-hour course on building agentic knowledge Graphs from scratch: • 00:00 - Introduction to AI agents • 08:09 - Foundations of multi-agent systems • 13:25 - Introduction to Graph engineering • 23:27 - Environment setup for agentic Graphs • 47:43 - Building Graph RAG agents... show more
197,514 просмотров • 3 месяцев назад •via X (Twitter)
Комментарии: 47
the graph engineering section at 13:25 is where the real meat is, skipped the intro
yeah. brilliant part thats for sure !
Graphs are goat, its another step in engineering
yeah, graphs are the future of multi-agentic systems !
agent memory banks via graphs are the only way to prevent context collapse in complex tasks. glad the industry is finally moving past basic vector search
Yeah. Agree on this
When are they gonna release a model tho ?lol
Lol. I wish i know bro !
Nice share Codez, thanks
Thx Morty. Happy it’s catched your attention
One hour that actually looks worth saving.
1-hour that will change the way you are building agents forever
graph memory is the one piece I still havent got right my notes just sit in flat files never linking to anything, watching this tonight
loops were only the beginning of orchestration
that section at 47:43 on building graph rag agents is what i needed
yeah, the real workshop by google team
graphs are where agent memory starts becoming architecture
yeah, huge setp for building multi-agentic systems in my opinion
bookmarked, ty for the drop
You’re welcome, buddy! Have you read this article on graph engineering yet? It’s top-tier writing.
already done!
very alpha video! thank you, Codez! book + like!
You are welcome Dunik. Have you started to build your agentic graphs ?
@0xCodez legit never thought I'd be interested in Graph engineering, but this has me hooked. Gonna check it out tonight!
graphs are the future of agent design
Yeah, it’s basically the layer of multi-agent orchestration.
A new wave
Graph engineering is the future of AI agents.
Yeah, it's the logical next step
The jump from one giant hallucinating while-loop to explicit graphs is where agents become debuggable. Solid curriculum.
graphs turned my scattered AI knowledge into one connected mind
nice
Or.. you could just ask your Agent to do it 🤷🏼♂️
bookmarked same shelf as the other five 'from scratch' courses i swore i'd finish
That chapter order gets the hard part right: graph engineering before the build. In practice, durable state and idempotent nodes matter more than adding agents once retries start.
Good
@grok これの元動画探して。ぽすとでもいいし、YouTubeあるならそれを優先で
A full free course on knowledge graphs, that's a solid find
Incredible, thank you for sharing!
Knowledge graphs become much more useful when they're connected to agents that can reason over them, not just store information
Free course like this is a solid place to start.
Nice, 1 hour well spent. Agentic knowledge graphs are where RAG meets real autonomy — curious how they handle graph updates without re-indexing everything.
This is exactly the kind of AI insight that separates the serious builders
Bookmarked
Knowledge graphs are becoming a key skill for future agents.
decent video gotta check out what other lectures this girl's posted
The while-loop-with-a-growing-prompt failure mode is the right thing to lead with, and it is worth naming why it happens: a loop has nowhere to put state except the prompt, so every step ends up paying for every step before it. Splitting into a graph fixes the context growth. What catches people next is durability. Nodes fail, a tool times out, a model returns malformed JSON, and if the graph only lives in process memory the whole run dies with it. Once each node can checkpoint and be retried on its own, the graph stops being a diagram and starts being something you can leave running unattended.
