Anthropic engineer: “90% of our engineers were using self‑improving loops. Now everyone shifted to building agentic Graphs" "No more prompting” In 10 minutes she shows her full Claude Code setup and workflow live, from a blank terminal. Worth more than a $500 agentic course. Watch this video, then save... the article below on how to become a graph architect👇show more

This is not a video on graphs This is literally a video of the original release of Claude

original video link:

We can just lie on X now? This video has nothing to do with graph engineering

Nice bait, she doesn't mention "graph" once in the video

moving from prompting to graph architecture changes the entire developer role

its the new thing

thank you for consistently sharing things at this level — the curation is as valuable as the content itself

yeah was digging for that video for a while

Dude, this is 1y old video

now we can create bullshit with more speed

Prompts to loops to graphs to transcendental meditation circuit. We don’t even use computers anymore.

@grok is this guy an archeologist ? where did he find old video like this?

this video is from a year ago when they first launched claude code. it has nothing to do with building agentic Graphs

alpha video. I'm pretty sure that this gotta be secret or smth

yeah it was deleted from anthropic channel

good to see you saved it for us

Sounds good, bro Thx for sharing

appreciate you so much! hope you find smth new out

saved

many thanks!

Where are the graphs?

This is literally a video of the original release of Claude

this shift to agentic graphs shows ai maturing into workflow layers

this is an open source graph harness for doing just that. It works with cursor. It works with grok. It works with clod. It works with codex. It takes spec-driven development and graph engineering and marries the two with harness and context engineering, building a layer in a container for your agents to call and reference back to, tracking against the original PRD.

your curation is legit as valuable as the content. appreciate you consistently sharing stuff this good.

@zodchiii that's wild. makes me wanna dive into agentic Graphs now. seems like i've been missing out

From a special group of people doesn't even understand the basic mathematical requirements for reinforcement learning, and stupidly believes non deterministic and stateless LLM can self improve. You really have to wonder what mushrooms they ate.

the setup is the only real part, got the link?

Nice share brother

apreciate you aviddd

Are people still doing graph engineering? That’s embarrassing. I’ve moved on onto network engineering now

This is so old ?

inviting you to telegram with daily AI alpha:

@grok - how many times in this demo did she prompt Claude?

The move from prompting to building agentic workflows feels like a real phase change. Curious how far along people actually are — we’re collecting that signal at

Graphs are old.. the edge is agentic tesseracts.

Instead of deciding more and more how ai produces more, lets just first execute and make useful products. Most loops work fine

Agentic graphs make control flow explicit, but what proves a node output is safe to advance rather than merely well-formed? Every handoff needs an acceptance check and action boundary. Practical checklist:

absence of promts is a good thing

it’s the new leverage

Anthropic is teaching people but dont know why people dont want to learn

Most graphs are paying full frontier rates to remember things nobody will ever ask about.