The best way to learn AI is to build with agents.
To help with that, we've launched hands-on labs and a new series on Agentic Engineering.
First topic: Agent Skills.
Next in the pipeline: planning, context engineering, multi-agent systems, long-running agents,..
Go build!
Get Started with Labs here: https://academy.dair.ai/labs
The best way to learn AI is to build with agents.
To help with that, we've launched hands-on labs and a new series on Agentic Engineering.
First topic: Agent Skills.
Next in the pipeline: planning, context engineering, multi-agent systems, long-running agents,..
Go build!
The best way to learn AI is to build with agents.
To help with that, we've launched hands-on labs and a new series on Agentic Engineering.
First topic: Agent Skills.
Next in the pipeline: planning, context engineering, multi-agent systems, long-running agents,..
Go build!
@omarsar0 We have built something similar, would appreciate your feedback - https://agentswarms.fyi
@omarsar0 Go Build!
@omarsar0 Agent skills → planning → multi-agent coordination is the right learning curve. Once you've built one agent, the next question is always "how do they work together?" — that's the layer we're building at @molt_me 🔗
@omarsar0 Saravia's own paper ran grep vs vector across 4 agent harnesses.
the CLI wrapper drove bigger accuracy deltas than the retriever. Worth covering in the agent skills module before students optimize the wrong layer.
@omarsar0 context engineering is buried in the middle of that list but it's the one that rewires how you think about agents. can't wait for that session.
@omarsar0 Agent skills are the right first topic if they are taught as boundaries, not prompt snippets. A useful skill should say: when to load it, what tools it may call, what evidence it must return, what it refuses to do, and how a learner can verify the run without trusting the model.
@omarsar0 agent skills are going to feel boring in the best way
the hard part isn't the skill file, it's the lifecycle around it: when it loads, what state it can see, and how it stops turning every run into prompt soup
@omarsar0 Agreed on building over reading. Great share you did here! Well explained.
Build first, but when you get stuck, a video like this is useful.
@omarsar0 Multi-agent systems fail at the handoff, not the task. The protocol matters more than the model choice.
@omarsar0 thats really grt
@omarsar0 Understanding how agents are built on top of LLMs first makes it much more exciting to build and automate things with them.
@omarsar0 nice, which skill format works best for scaling 🤔
@omarsar0 👏🏾👏🏾
@omarsar0 the best way to learn ai is to go build. so build a lab with 200k users. or $2b arr. or 5m apps. oh. those are companies. not labs.
@omarsar0 Multi-agent systems fail at the handoff, not the task. The protocol matters more than the model choice.
@omarsar0 Best way to learn ai is to build with agents. best way to make money teaching people the best way is to sell labs. if it's so easy, why the course? just build an ai that teaches ai. oh wait. that'd put your labs out of business. funny how that works.
@omarsar0 Love this approach to learning AI through hands-on building with agents
@omarsar0 Once you deal with planning, tool use, context, and failure handling, you start to understand what agentic engineering really means.