AI Engineer CODE 2025
2025-11-19..2025-11-22
New York City, United States
who it's for
agent builders, AI engineers, machine learning engineers, platform engineers, developer experience engineers, engineering managers
topics
durable agent workflows; agent identity and token management; MCP server design; spec-driven agent development; prompt learning loop; OpenAI and Temporal integration; parallel agent refactoring; reinforcement learning at scale; coding evaluation frameworks; agent SDK workshops; making codebases agent ready; RAG and document segmentation
| title | author | length |
|---|---|---|
| How METR measures Long Tasks and Experienced Open Source Dev Productivity | Joel Becker | 01h 16m |
METR time-horizon scores and SWE-bench improvement claim models are getting faster and tackling longer tasks, but a field RCT with 16 experienced open-source… | ||
| Identity for AI Agents | Patrick Riley, Carlos Galan | 01h 22m |
Okta-backed stack adds a token vault that persists refresh tokens, manages lifetimes and does token exchange so agents get upstream API access without… | ||
| OpenAI + @Temporalio : Building Durable, Production Ready Agents | Cornelia Davis | 01h 19m |
Shows a working integration that makes OpenAI Agents SDK agents durable by running them inside Temporal workflows: agents call the OpenAI Responses API, list… | ||
| Your MCP Server is Bad (and you should feel bad) | Jeremiah Lowin | 55m |
Claim: many MCP servers are just REST wrappers that confuse LLMs, waste tokens, and hurt performance. FastMCP becomes the de facto framework (millions of… | ||
| Spec-Driven Development: Agentic Coding at FAANG Scale and Quality | Al Harris | 01h 04m |
Kiro claims that Spec‑Driven Development turns prompts into structured EARS requirements, designs, task lists and end‑to‑end code using MCP servers, CDK and… | ||
| DSPy: The End of Prompt Engineering | Kevin Madura | 01h 13m |
Claims that DSPy replaces brittle prompt tinkering by letting developers write typed Python programs with LLM calls as first‑class functions, using Signatures… | ||
| Automating Large Scale Refactors with Parallel Agents | Robert Brennan | 01h 16m |
Claims that parallel agents orchestrated from a cloud sandbox can automate massive refactors and dependency fixes at scale, using the Open Hands agent SDK and… | ||
| Build a Prompt Learning Loop | SallyAnn DeLucia, Fuad Ali | 52m |
Builds a practical prompt optimization loop that reads failure examples, asks an LLM to rewrite the system prompt using human explanations and rule-checker… | ||
| Building durable Agents with Workflow DevKit & AI SDK | Peter Wielander | |
Converts a local coding agent into a durable, production-ready workflow using Vercel’s open‑source Workflow DevKit and the AI SDK, with Next.js and a sandbox… | ||
| Claude Agent SDK [Full Workshop] | Thariq Shihipar | 01h 52m |
Workshop shows an implemented Agent Loop (Context → Thought → Action → Observation) inside a built Harness that wires tools, prompts and a file system for… | ||
| Welcome to AIE CODE | Jed Borovik | 04m |
Emcee opens the single-track AI Engineering CODE Summit with an engineer-focused framing: AI coding as the day’s central problem, and a roadmap of sponsors… | ||
| Building Intelligent Research Agents with Manus | Ivan Leo | 01h 22m |
Demos show the Manus API creating asynchronous agent tasks that query private connectors, upload large files, and return structured reports with task IDs and… | ||
| AGI: The Path Forward | Jason Warner, Eiso Kant | 16m |
Poolside builds models from scratch and demonstrates a VS Code agent called Malibu agent that converts an ADA codebase to Rust, writes about 1152 lines of… | ||
| How Claude Code Works | Jared Zoneraich | 01h 06m |
Argues Claude Code succeeds by replacing brittle DAGs with a single-threaded Master Loop that iterates tool calls until done, relying on stronger models rather… | ||
| Why Agent Hype can fall short of reality | Joel Becker | 21m |
METR measures time horizons by mapping human task completion minutes to model 50% success points (Claude 3 Opus ≈ 4 min, o1‑preview ≈ 15 min) and finds a… | ||
| Amp Code: Next Generation AI Coding | Beyang Liu | 19m |
Amp Code builds an opinionated terminal-and-editor coding agent with a custom TUI framework, streaming diffs, editor integrations (Emacs, Neovim, JetBrains, VS… | ||
| Making Codebases Agent Ready | Eno Reyes | 16m |
Argues that the bottleneck for agent adoption is environment readiness not model quality and gives a measurable framework: eight categories (linters, build… | ||
| The 3 Pillars of Autonomy | Michele Catasta | 25m |
Claims a production coding agent that makes nontechnical users fully hands-off by combining autonomous testing (generating Playwright scripts, programmatic… | ||
| From Arc to Dia: Lessons learned building AI Browsers | Samir Mody | 18m |
Arc becomes Dia, a ground-up AI-native browser that moves prompts, tools, models and context inside the product so anyone at the company can prototype and… | ||
| AI Kernel Generation: What's working, what's not, what's next | Natalie Serrino | 19m |
Automated agents generate low‑level kernels for PyTorch and sometimes improve performance: a CLI experiment finds a candidate 22% faster than torch.compile on… | ||
| Code World Model: Building World Models for Computation | Jacob Kahn | 17m |
A 32 billion parameter dense transformer models program execution by generating execution traces with local variables and predicting state transitions, trained… | ||
| Coding Evals: From Code Snippets to Codebases | Naman Jain | 18m |
Traces the shift from snippet benchmarks to whole-repo evaluation with concrete practices: dynamic LiveCodeBench releases (six versions) that use 30–50 inputs… | ||
| Hard Won Lessons from Building Effective AI Coding Agents | Nik Pash | 14m |
Argues that frontier models outperform elaborate stacks: Gemini 3.0 immediately tops terminal bench leaderboards running on a stripped-down Terminus harness… | ||
| RL Environments at Scale | Will Brown | 19m |
Describes an open stack for scaling RL environments: an environments hub plus the verifiers toolkit and a Prime RL trainer that runs thousands of rollouts on… | ||
| Efficient Reinforcement Learning | Rhythm Garg, Linden Li | 20m |
An RL stack uses asynchronous pipeline RL with dedicated sampling and training GPUs, in‑flight weight updates, and importance‑ratio correction; a test with… | ||
| Don't Build Agents, Build Skills Instead | Barry Zhang, Mahesh Murag | 16m |
Claim: package procedural knowledge as organized folders of files and scripts that a code-enabled agent loads at runtime instead of reengineering new agents… | ||
| No Vibes Allowed: Solving Hard Problems in Complex Codebases | Dex Horthy | 21m |
Claims that current models can handle brownfield repos at scale: Claude code works on a 300k‑LOC Rust codebase and a workflow ships what would be a week’s work… | ||
| Defying Gravity | Kevin Hou | 25m |
Google Antigravity ships an agent-first IDE with three linked surfaces: an editor with agent sidebar, an agent manager hub for parallel agents and inboxed… | ||
| Building Cursor Composer | Lee Robinson | 16m |
Cursor Composer is a coding agent model trained with RL to be both fast and smart, claiming about four times more efficient token generation versus peers and… | ||
| nothing matches | ||