Sign inStart free
AI Engineering/Phase 14
Frontier

Agent Engineering

Loops, memory, planning and the engineering of reliable agents.

42 lessons ~43.2 hr
  1. The Agent Loop: Observe, Think, ActBuild · Free · 1 hEvery agent in 2026 — Claude Code, Cursor, Devin, Operator — is a variant of the ReAct loop from 2022. Reasoning tokens interleave with tool calls and observations until a stop condition fires. Learn this loop cold before touching any framework.Open
  2. ReWOO and Plan-and-Execute: Decoupled PlanningBuild · 1 hReAct interleaves thought and action in one stream. ReWOO separates them: one big plan up front, then execute. 5x fewer tokens, +4% accuracy on HotpotQA, and you can distill the planner into a 7B model. Plan-and-Execute generalized it; Plan-and-Act scaled it to web navigation.Needs Full Access
  3. Reflexion: Verbal Reinforcement LearningBuild · 1 hGradient-based RL needs thousands of trials and a GPU cluster to fix a failure mode. Reflexion (Shinn et al., NeurIPS 2023) does it in natural language: after each failed trial, the agent writes a reflection, stores it in episodic memory, and conditions the next trial on that memory. This is the pattern behind Letta's sleep-time compute, Claude Code's CLAUDE.md learnings, and pro-workflow's learn-rule.Needs Full Access
  4. Tree of Thoughts and LATS: Deliberate SearchBuild · 1 h 15 minA single chain-of-thought trajectory has no room to backtrack. ToT (Yao et al., 2023) turns reasoning into a tree with self-evaluation on each node. LATS (Zhou et al., 2024) unifies ToT with ReAct and Reflexion under Monte Carlo Tree Search. Game of 24 goes from 4% (CoT) to 74% (ToT); LATS hits 92.7% pass@1 on HumanEval.Needs Full Access
  5. Self-Refine and CRITIC: Iterative Output ImprovementBuild · 1 hSelf-Refine (Madaan et al., 2023) uses one LLM in three roles — generate, feedback, refine — in a loop. Average gain: +20 absolute on 7 tasks. CRITIC (Gou et al., 2023) hardens the feedback step by routing verification through external tools. In 2026 this pattern ships in every framework as "evaluator-optimizer" (Anthropic) or a guardrail loop (OpenAI Agents SDK).Needs Full Access
  6. Tool Use and Function CallingBuild · 1 hToolformer (Schick et al., 2023) started self-supervised tool annotation. Berkeley Function Calling Leaderboard V4 (Patil et al., 2025) sets the 2026 bar: 40% agentic, 30% multi-turn, 10% live, 10% non-live, 10% hallucination. Single-turn is solved. Memory, dynamic decision-making, and long-horizon tool chains are not.Needs Full Access
  7. Memory: Virtual Context and MemGPTBuild · 1 h 15 minContext windows are finite. Conversations, documents, and tool traces are not. MemGPT (Packer et al., 2023) frames this as OS virtual memory — main context is RAM, external store is disk, the agent pages between them. This is the pattern every 2026 memory system inherits.Needs Full Access
  8. Memory Blocks and Sleep-Time Compute (Letta)Build · 1 h 15 minMemGPT became Letta in 2024. The 2026 evolution adds two ideas: discrete functional memory blocks the model can edit directly, and a sleep-time agent that consolidates memory asynchronously while the primary agent is idle. This is how you scale memory beyond one conversation.Needs Full Access
  9. Hybrid Memory: Vector + Graph + KV (Mem0)Build · 1 h 15 minMem0 (Chhikara et al., 2025) treats memory as three stores in parallel — vector for semantic similarity, KV for fast fact lookup, graph for entity-relationship reasoning. A scoring layer fuses the three on retrieval. This is the 2026 production standard for external memory.Needs Full Access
  10. Skill Libraries and Lifelong Learning (Voyager)Build · 1 h 15 minVoyager (Wang et al., TMLR 2024) treats executable code as a skill. Skills are named, retrievable, composable, and refined by environment feedback. This is the reference architecture for Claude Agent SDK skills, skillkit, and the 2026 skill-library pattern.Needs Full Access
  11. Planning with HTN and Evolutionary SearchBuild · 1 h 15 minSymbolic planning handles the cases where the plan is provably correct. Evolutionary code search handles the cases where the fitness function is machine-checkable. ChatHTN (2025) and AlphaEvolve (2025) show what each unlocks when paired with an LLM.Needs Full Access
  12. Anthropic's Workflow Patterns: Simple Over ComplexLearn + Build · 1 hSchluntz and Zhang (Anthropic, Dec 2024) distinguish workflows (predefined paths) from agents (dynamic tool-use). Five workflow patterns cover most cases. Start with direct API calls. Add agents only when steps cannot be predicted.Needs Full Access
  13. LangGraph: Stateful Graphs and Durable ExecutionLearn + Build · 1 h 15 minLangGraph is the 2026 reference for low-level stateful orchestration. Agent is a state machine; nodes are functions; edges are transitions; state is immutable and checkpointed after every step. Resume from any failure exactly where it left off.Needs Full Access
  14. AutoGen v0.4: Actor Model and Agent FrameworkLearn + Build · 1 h 15 minAutoGen v0.4 (Microsoft Research, Jan 2025) redesigned agent orchestration around the actor model. Async message exchange, event-driven agents, fault isolation, natural concurrency. The framework is now in maintenance mode while Microsoft Agent Framework (public preview Oct 2025) becomes the successor.Needs Full Access
  15. CrewAI: Role-Based Crews and FlowsLearn + Build · 1 h 15 minCrewAI is the 2026 role-based multi-agent framework. Four primitives: Agent, Task, Crew, Process. Two top-level shapes: Crews (autonomous, role-based collaboration) and Flows (event-driven, deterministic). The docs are blunt: "for any production-ready application, start with a Flow."Needs Full Access
  16. OpenAI Agents SDK: Handoffs, Guardrails, TracingLearn + Build · 1 h 15 minOpenAI Agents SDK is the lightweight multi-agent framework built on the Responses API. Five primitives: Agent, Handoff, Guardrail, Session, Tracing. Handoffs are tools named `transfer_to_<agent>`. Guardrails trip on input or output. Tracing is on by default.Needs Full Access
  17. Claude Agent SDK: Subagents and Session StoreLearn + Build · 1 h 15 minThe Claude Agent SDK is the library form of the Claude Code harness. Built-in tools, subagents for context isolation, hooks, W3C trace propagation, session store parity. Claude Managed Agents is the hosted alternative for long-running async work.Needs Full Access
  18. Agno and Mastra: Production RuntimesLearn · 45 minAgno (Python) and Mastra (TypeScript) are the 2026 production-runtime pairing. Agno aims at microsecond agent instantiation and stateless FastAPI backends. Mastra ships agents, tools, workflows, unified model routing, and composite storage on the Vercel AI SDK substrate.Needs Full Access
  19. Benchmarks: SWE-bench, GAIA, AgentBenchLearn · 1 hThree benchmarks anchor agent evaluation in 2026. SWE-bench tests code patching. GAIA tests generalist tool use. AgentBench tests multi-environment reasoning. Know their composition, their contamination story, and what they do not measure.Needs Full Access
  20. Benchmarks: WebArena and OSWorldLearn · 1 hWebArena tests web-agent capability across four self-hosted apps. OSWorld tests desktop-agent capability across Ubuntu, Windows, macOS. At release (2023–2024) both showed a big gap between best-in-class agents and humans. The gap is narrowing; the failure modes haven't changed.Needs Full Access
  21. Computer Use: Claude, OpenAI CUA, GeminiLearn · 1 hThree production computer-use models in 2026. All three are vision-based. All three treat screenshots, DOM text, and tool outputs as untrusted input. Only direct user instructions count as permission. Per-step safety services are the norm.Needs Full Access
  22. Voice Agents: Pipecat and LiveKitLearn · 1 hVoice agents are a first-class production category in 2026. Pipecat gives you a Python frame-based pipeline (VAD → STT → LLM → TTS → transport). LiveKit Agents bridges AI models to users over WebRTC. Production latency targets land at 450–600ms end-to-end for premium stacks.Needs Full Access
  23. OpenTelemetry GenAI Semantic ConventionsLearn + Build · 1 hOpenTelemetry's GenAI SIG (launched April 2024) defines the standard schema for agent telemetry. Span names, attributes, and content-capture rules converge across vendors so agent traces mean the same thing in Datadog, Grafana, Jaeger, and Honeycomb.Needs Full Access
  24. Agent Observability: Langfuse, Phoenix, OpikLearn · 45 minThree open-source agent observability platforms dominate 2026. Langfuse (MIT) — 6M+ installs/month, tracing + prompt management + evals + session replay. Arize Phoenix (Elastic 2.0) — deep agent-specific evals, RAG relevancy, OpenInference auto-instrumentation. Comet Opik (Apache 2.0) — automated prompt optimization, guardrails, LLM-judge hallucination detection.Needs Full Access
  25. Multi-Agent Debate and CollaborationLearn + Build · 1 hDu et al. (ICML 2024, "Society of Minds") run N model instances that independently propose answers, then iteratively critique each other over R rounds to converge. Improves factuality, rule-following, reasoning. Sparse topology beats full mesh on token cost.Needs Full Access
  26. Failure Modes: Why Agents BreakLearn + Build · 1 hMASFT (Berkeley, 2025) catalogs 14 multi-agent failure modes in 3 categories. Microsoft's Taxonomy documents how existing AI failures amplify in agentic settings. Industry field data converges on five recurring modes: hallucinated actions, scope creep, cascading errors, context loss, tool misuse.Needs Full Access
  27. Prompt Injection and the PVE DefenseBuild · 1 h 15 minGreshake et al. (AISec 2023) established indirect prompt injection as the defining agent security problem. Attacker plants instructions in data the agent retrieves; on ingest, those instructions override the developer prompt. Treat all retrieved content as arbitrary code execution on the tool-use surface.Needs Full Access
  28. Orchestration Patterns: Supervisor, Swarm, HierarchicalLearn + Build · 1 hFour orchestration patterns recur across 2026 frameworks: supervisor-worker, swarm / peer-to-peer, hierarchical, debate. Anthropic's guidance: "It's about building the right system for your needs." Start simple; add topology only when a single agent plus five workflow patterns is insufficient.Needs Full Access
  29. Production Runtimes: Queue, Event, CronLearn · 1 hProduction agents run on six runtime shapes: request-response, streaming, durable execution, queue-based background, event-driven, and scheduled. Pick the shape before you pick the framework. Observability is load-bearing at every shape.Needs Full Access
  30. Eval-Driven Agent DevelopmentLearn + Build · 1 hAnthropic's guidance: "start with simple prompts, optimize them with comprehensive evaluation, and add multi-step agentic systems only when needed." Evaluation is not the last step. It's the outer loop that drives every other choice in Phase 14.Needs Full Access
  31. Agent Workbench Engineering: Why Capable Models Still FailLearn + Build · 45 minA capable model is not enough. Reliable agents need a workbench: instructions, state, scope, feedback, verification, review, and handoff. Strip those away and even a frontier model produces work that is unsafe to ship.Needs Full Access
  32. The Minimal Agent WorkbenchBuild · 45 minThe smallest useful workbench is three files: a root instructions router, a state file, and a task board. Everything else is layered on top. If a repo cannot carry these three, no model will save it.Needs Full Access
  33. Agent Instructions as Executable ConstraintsBuild · 50 minInstructions written as prose are wishes. Instructions written as constraints are tests. The workbench turns each rule into something an agent can check at runtime and a reviewer can verify after the fact.Needs Full Access
  34. Repo Memory and Durable StateBuild · 1 hChat history is volatile. The repo is durable. The workbench stores agent state in versioned files so the next session, the next agent, and the next reviewer all read from the same source of truth.Needs Full Access
  35. Initialization Scripts for AgentsBuild · 45 minEvery session that starts cold pays a tax. The agent reads the same files, retries the same probes, and rediscovers the same paths. An init script pays the tax once and writes the answers into state.Needs Full Access
  36. Scope Contracts and Task BoundariesBuild · 50 minThe model does not know where the work ends. A scope contract is a per-task file that says where the work begins, where it ends, and how to roll back if it spills. The contract turns "stay in scope" from a wish into a check.Needs Full Access
  37. Runtime Feedback LoopsBuild · 50 minAgents that do not see real command output guess. A feedback runner captures stdout, stderr, exit code, and timing into a structured record the next turn can read. Then the agent reacts to facts instead of to its own prediction of facts.Needs Full Access
  38. Verification GatesBuild · 55 minThe agent does not get to mark its own work as done. A verification gate reads the scope contract, the feedback log, the rule report, and the diff, and answers a single question: is this task actually complete? If the gate says no, the task is not done, no matter what the chat says.Needs Full Access
  39. Reviewer Agent: Separate Builder from MarkerBuild · 55 minThe agent that wrote the code cannot grade it. A reviewer is a second loop with a different system prompt, a different goal, and read-only access to everything the builder produced. The gap between builder and reviewer is where most reliability lives.Needs Full Access
  40. Multi-Session HandoffBuild · 50 minThe session is going to end. The work is not. The handoff packet is the artifact that turns "the agent worked for an hour" into "the next session is productive in the first minute." Build it on purpose, not as an afterthought.Needs Full Access
  41. The Workbench on a Real RepoBuild · 1 hEleven lessons of surfaces are worth nothing if they do not survive contact with a real codebase. This lesson runs the same task twice on a small sample app: prompt-only versus workbench-guided. The numbers do the arguing.Needs Full Access
  42. Capstone: Ship a Reusable Agent Workbench PackBuild · 1 h 15 minThe mini-track ends with a pack you drop into any repo. Eleven lessons of surfaces compressed into a directory you can `cp -r` and have an agent working reliably the next morning. The capstone is the artifact this curriculum trades on.Needs Full Access