Anthropic senior engineer just released a 1-hour course on building agentic systems with Claude from scratch: • 00:37 - how Claude Code actually works • 02:29 - building CLAUDE.md and Plan for agents • 25:38 - creating /skills for Claude agents • 45:39 - Claude agents team design patterns... • 52:10 - Agents SDK: subagents, loops, graphs this 1-hour workshop will replace 10 paid courses on agentic engineering watch it today, then read how to build self-learning agentic systems with loops & graphs in the article belowshow more

One hour to learn what most people will still take a decade to misuse.

yeah, this 1-hour watch will change the way you use Claude Code forever !

skills section looks worth saving

all the course is worth your time mate, this 1-hour watch will change the way you use Claude forever in my opinion

@0xMovez nice! Diving into agency models changed how I handle task automation. Curious about your thoughts on scaling Claude agents effectively?

legendary workshop, this is the kind of teaching that actually sticks with me

not just a workshop, but a full course on Claude Code amigo !

awo

this is the kind of agentic systems course that actually moves the needle, definitely worth watching ngl

The self-learning part is interesting — most loop/graph setups reset per session, re-deriving from scratch each run. What makes it actually persistent is recording what a human already vetoed, not just what succeeded — otherwise it re-proposes the same rejected idea next time.

Anthropic’s burn is ~$3B/yr. The 1-hour course saves you $40k in Claude API fees during the trial phase before it gets too good to need a syllabus. Ship first.

Anthropic senior engineers always cook !

amazin master-class about Claude Code bro !

yeah bro, amazing article as always !

benchmark olympics again huh

Great Maam thanks a lot 🙏

What’s this have to do with the article you linked?

The sequence matters here. Defining context in CLAUDE.md and planning the work before adding skills, subagents, and graphs should prevent many brittle builds. Too many people start with orchestration.

this workshop is pure alpha

gmgm 25:38 on skills is the only timestamp that matters here. did you test the skill limit yourself or take it from the video?

at 52:10, are the subagents actually sharing state cleanly, or is the course hiding the coordination complexity behind a neat diagram?

the /skills section alone is worth the hour if you're wiring agents

the CLAUDE.md + /skills combo is the part most people skip. once repo conventions live in the project file, agent teams stop inventing APIs every session.

When the engineers building the SDK hand you the exact blueprint for state management, agent loops, and context control for free, buying third-party prompt guides becomes an instant intelligence test.

The CLAUDE.md and planning pieces are a great fit for agent workflows. Manna Code keeps the whole session verbatim too, so the context is still inspectable when you pick work back up.

empty terminal to full system in an hour goes crazy