Lesson 1546 of 1596
Agents Demystified: What They Are and Are Not
Cut through the hype to see what an AI agent actually is — a loop, not magic.
Creators · AI Foundations · ~7 min read
The premise
An AI agent is a model in a loop with tools and a goal. The hype obscures how simple the mechanism is — and how the simplicity is also the source of the failure modes.
What AI does well here
- Running a model in a loop where it plans, acts, observes, and re-plans
- Letting the model decompose vague goals into concrete tool calls
- Recovering from many small errors that simple chains would propagate
- Operating asynchronously over long time horizons
What AI cannot do
- Avoid compounding errors — small mistakes early derail later steps
- Stay on task without good guardrails and time/step budgets
- Replace human oversight in any high-stakes domain
Key terms in this lesson
End-of-lesson quiz
Check what stuck
10 questions · Score saves to your progress.
Tutor
Curious about “Agents Demystified: What They Are and Are Not”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Creators · 11 min
Attention deep dive: queries, keys, values, and why it works
Understand attention as a content-addressable lookup over a sequence — and where the analogy breaks.
Creators · 11 min
Tokenization economics: why your bill depends on the tokenizer
Tokenization decisions ripple into cost, latency, and capability — for languages, code, and rare strings.
Creators · 11 min
RLHF vs DPO: aligning models without breaking them
Compare reinforcement learning from human feedback and direct preference optimization at the level of intuition, not equations.