coming in planning

building production agents

design cursor on a whiteboard. agent loops, tool use, durable runtimes, sandboxing, and agent-specific evals.

13 lessons|4 modules|~7 hours

what you’ll learn

design coding agents, deep research agents, and code review bots from loop topology to memory
pick the right multi-agent shape, knowing when supervisor patterns help and when they introduce bugs
run agents durably across crashes, retries, cancellations, and rainbow deployments
build sandboxes and capability gating that let agents act without blowing up production
design eval methodology for stateful agents that mutate state across many turns

curriculum

planning sketch

this is a rough curriculum we’re still planning. modules and lessons are likely to shift before any lesson is recorded. want to shape it? mail@karnstack.com.

module one

agent patterns

~90 min3 lessons

01what is an agent: the loop, react, plan-execute, reflexioncoming soon30m

02tool use, function calling, mcpcoming soon30m

03memory architectures and context engineeringcoming soon30m

module two

agent runtime

~95 min3 lessons

04multi-agent topologies: supervisor, sub-agents, swarm, debatecoming soon30m

05runtime concerns: concurrency, idempotency, cancellation, durable executioncoming soon35m

06sandboxing, capability gating, human-in-the-loopcoming soon30m

module three

evals for agents

~60 min2 lessons

07online evals, shadow runs, a/b for prompts and modelscoming soon30m

08end-state evals for stateful agentscoming soon30m

module four

agent case studies

~165 min5 lessons

09cursor and cursor composer: a coding agent dissectioncoming soon35m

10claude code: agentic coding internalscoming soon35m

11anthropic research and chatgpt deep research: long-horizon agentscoming soon35m

12code review bot: greptile and coderabbitcoming soon30m

13customer support agent: sierra and intercom fincoming soon30m

frequently asked

when does this launch?: in planning, sequenced after ai-foundations. the curriculum on this page is a sketch. modules and lessons are likely to shift before any lesson is recorded.
do i need to do ai-foundations first?: ideally yes. this course assumes you've already internalised inference, retrieval, and the eval foundations. you can come in with equivalent shipping experience instead.

how this course is made

the curriculum is curated by karnstack and reviewed by senior engineers in the industry before it ships. narration is an ai voice (elevenlabs) reading human-written, human-reviewed scripts. read how courses are made.