Stanford CS329A: Self-Improving AI Agents

LinkAakanksha Chowdhery, Azalia MirhoseiniResource

Stanford University's CS329A is a graduate course dedicated to self-improving AI agents, taught by Aakanksha Chowdhery and Azalia Mirhoseini. The curriculum covers foundational and modern techniques enabling LLMs to iteratively improve via environment interaction and internal feedback.

Key points

  1. Explores self-improvement techniques including test-time compute scaling, robust verifiers, train-time reinforcement learning, and search integration.
  2. Covers tool integration, code grounding, memory architectures, and multimodal orchestrations for agentic workflows.
  3. Addresses complex planning, reasoning trees, and challenges in establishing rigorous evaluation benchmarks for long-horizon tasks.
  4. Features industry and academic guest lectures from Google DeepMind, Reflection AI, Physical Intelligence, and UChicago.

Why it matters

The syllabus consolidates key technical paradigms driving modern agent research, particularly the shift towards inference-time compute scaling, automated verification, and autonomous feedback loops.

Analysis

By pairing recent literature (such as STaR, MemGPT, and AlphaCode) with practical project work, the course outlines how frontier AI engineering is advancing beyond static prompting toward autonomous, self-evaluating systems.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Video

    Video:Stanford Online

    Stanford CS329A: Introduction to Self-Improving AI Agents and Scaling Trends

    In the inaugural lecture of Stanford's CS329A course, Aakanksha Chowdhery and Azalia Mirhoseini outline the evolution of large language models towards autonomous, self-improving agents. The session explores foundational scaling laws, the emergence of chain-of-thought reasoning, post-training alignment via RLHF, inference-time scaling techniques, and the architectural principles behind agentic workflows in software engineering and research.

    AI & LLMs· Talk

  • Link:Diyi Yang, Michael Ryan, John Yang

    CS 329Z: Stanford Course on Engineering AI Agents

    Stanford University is running CS 329Z (Engineering AI Agents) for Fall 2026. The course explores the transition from monolithic language models to compound AI systems and autonomous agents. All lecture materials, syllabus details, and reading lists are made publicly available online.

    AI & LLMs· Resource

  • X post:Mark Phelps

    Stanford Launches CS329Z: Engineering AI Agents

    Stanford University has released the curriculum for CS329Z ('Engineering AI Agents') for Autumn 2026. The course covers compound AI systems, building RAG and tool use from scratch, agent frameworks like DSPy, and systematic agent evaluation.

    2.3kbookmarks66.1kviews

    AI & LLMs· Resource

  • Link:Harrison Chase, Rotem Weiss

    AI Agents in LangGraph: New Short Course on DeepLearning.AI

    DeepLearning.AI offers a 1h42m course on LangGraph taught by LangChain CEO Harrison Chase and Tavily CEO Rotem Weiss. It covers building controllable Python-based AI agents, state persistence, agentic search, and human-in-the-loop patterns.

    AI & LLMs· Resource

  • Article:Siru Ouyang, Jun Yan, I-Hung Hsu, et al.

    ReasoningBank: Scaling Agent Self-Evolution with Reasoning Memory

    ReasoningBank is a memory framework for LLM agents that distils generalisable reasoning strategies from self-judged successful and failed experiences, enabling continuous self-evolution over time.

    AI & LLMs· Research

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.