Coding Agents — Localize, Edit, Verify Highlights: Coding agents now fix real repository issues: Hybrid-Gym lifts SWE-bench Verified from 7% to 32.4%. Graham Neubig’s lecture in the CMU 11-768 AI Agents course (Fall 2026) examines the transition from one-pass code prompting to iterative agentic software engineering. We walk through the localize-edit-verify repair loop, diff format tradeoffs, synthetic task generation via mutation, and multi-harness reinforcement learning. We also cover multimodal frontend testing and expanding agent evaluation…
Read more

Training Code LLMs — Data, Infilling, Tests Highlights: Train only on code and MATH scores collapse from 33.2 to 10.3. The right data mix, infilling and test feedback make code models work. Graham Neubig’s lecture in CMU 11-768 AI Agents (Fall 2026) outlines the technical foundations required to build autonomous code generation systems and repository-level software engineers. We walk through the progression across three capability tiers, data curation and whitespace-preserving tokenization, bidirectional causal infilling, and…
Read more

Manager + Worker Agents — Planning That Scales Highlights: A strong planner on top of a small worker lifts OSWorld success from 25% to 58%. Daniel Fried’s lecture in CMU 11-768 AI Agents (Fall 2026) analyzes modular system designs that decouple high-level planning from low-level grounded execution. We walk through the trade-offs between prolonged internal deliberation and empirical interaction scaling, the compounding failure modes that destabilize long-horizon trajectories, and multi-agent coordination frameworks for desktop computer…
Read more

Why Agents Plan — From Chain-of-Thought to Action Highlights: Thinking step by step is not enough once actions change the world. Here is why agents need explicit plans. Daniel Fried’s lecture in the CMU 11-768 AI Agents course (Fall 2026) examines why purely reactive execution loops break down across complex, long-horizon tasks. We walk through the trade-offs between upfront commitment and dynamic replanning, the shift from classical STRIPS search to grounded verification, and modular architectures…
Read more

Agents That Learn Skills — Reusable Workflows from Past Runs Highlights: Instead of solving every task from scratch, agents can mine their own past runs for reusable workflows and code skills. Daniel Fried’s lecture in Carnegie Mellon University’s CMU 11-768 AI Agents course (Fall 2026) examines how autonomous agents discover, verify, and maintain reusable skills across sequential web tasks. We walk through the representational trade-offs between natural language workflows and executable Python tools, automated verification…
Read more

Agent Memory — Skill Files and External Storage Highlights: Context windows forget everything between sessions. Skill files and tiered external memory let agents keep what they learn. Daniel Fried’s lecture in CMU 11-768 AI Agents (Fall 2026) examines how autonomous agents extract reusable operational skills and manage structured memory across multi-step tasks. We walk through the AgentSkills specification standard, progressive disclosure mechanisms that protect context windows, and benchmarked gains on SkillsBench. Finally, we explore tiered…
Read more

Long Context in Practice — Training, Caching, Compaction Highlights: Coherent 128K-token documents are scarce, so models reach 1M tokens through staged training, then stay affordable through caching and compaction. In this second part of the long-context lecture from CMU’s 11-768 AI Agents course (Fall 2026, Graham Neubig), the focus shifts from model architecture to everything that surrounds it: how long-context models are trained, served, and kept affordable. We walk through staged short-to-long training pipelines, positional…
Read more

Why Agent Context Explodes — and How to Serve It Highlights: Five agent turns already process 80K input tokens, and 37% of agent tokens are tool results. Graham Neubig’s lecture in the CMU 11-768 AI Agents course (Fall 2026) looks at why agent context grows quadratically and what architectures make long context affordable to serve. We walk through the token composition of real OpenHands sessions, the prefill/decode split and its serving metrics, and the local,…
Read more

Tools Over the Network — REST, MCP, Benchmarks Highlights: The same model can fail 15% of tool calls on one provider and 0.01% on another. This session of CMU’s 11-768 AI Agents course (Fall 2026, Graham Neubig) moves tool calling out of the notebook and into the network, tracing the path from in-process Python libraries to REST endpoints, MCP servers, and the serving stacks underneath them. We walk through wrapping tools with FastAPI and Pydantic,…
Read more

How LLMs Call Tools — CodeAct and Schemas Highlights: Letting the model write code instead of single API calls gave the best success rate for 8 of 8 models in CodeAct. Graham Neubig’s lecture on tool use in the CMU 11-768 AI Agents course (Fall 2026) treats tools as the interface through which a language model invokes an external program, and code as a meta-tool that collapses long chains of single API calls into one…
Read more

Agents Are Systems — Harness vs. Training Highlights: Assembling an agent is easy; making it reliable is not. Most fixes land in the harness first, then move into the weights. In Graham Neubig’s segment of CMU 11-768: AI Agents (taught with Daniel Fried), the argument is that assembling an agent is easy while making one work reliably is not. We walk through the capabilities that separate the two — tool calling, long-context coherence, customizability, task…
Read more

What Is an AI Agent? — A Model in a Loop Highlights: An agent deleted a user’s emails after context compaction dropped the one instruction not to. Here is how agents actually work. The opening lecture of CMU’s 11-768: AI Agents (Fall 2026), taught by Daniel Fried and Graham Neubig, sets out what an agent actually is and where current systems break down. We walk through the anatomy of an agent — environment, observations, actions,…
Read more