🛠️ Tools & Frameworks
The AI engineering toolchain — LangChain, LlamaIndex, DSPy, LangGraph, LangSmith, Braintrust, Inspect, AGENTS.md.
The workflow
flowchart LR
A[Pick framework] --> B{Use case}
B -->|Agents| C[LangGraph /<br/>CrewAI / AutoGen]
B -->|RAG| D[LlamaIndex /<br/>Haystack]
B -->|Orchestration| E[LangChain /<br/>DSPy]
B -->|Eval| F[Braintrust /<br/>LangSmith]
C --> G[Build, ship,<br/>observe]
D --> G
E --> G
F --> G
Pick a framework that matches your primitive — agents, retrieval, orchestration, or evals — not all four at once.
Key takeaways
Videos (31)
Building Great Agent Skills: The Missing Manual
Design skills deliberately around invocation tradeoffs, clear procedures, useful references, and minimal context.
DSPy: The End of Prompt Engineering - Kevin Madura, AlixPartners
DSPy lets you write LLM software as typed, modular Python programs that survive model swaps and can be auto-optimized rather than hand-prompted.
Full Workshop: Setting Yourself Up for Success —Jason Liu, OpenAI Codex
Invest in persistent context and shared workflows so agents support ongoing projects across coding and knowledge work.
Vercel AI SDK Masterclass: From Fundamentals to Deep Research
Vercel AI SDK provides a small, provider-agnostic set of primitives (generate/stream + tools) that scale from a hello-world prompt up to a multi-step deep-research agent.
Function Calling is All You Need — Full Workshop, with Ilan Bigio of OpenAI
Most agent and RAG architectures collapse to a function-calling loop — master that one primitive and the rest of the agent stack falls out.
The Dark Arts of Web Automation: Teaching Agents to Use Websites Like Humans — Corey Gallon, Rexmore
Turn browser exploration into reusable programs with explicit verification at each step.
On Engineering AI Systems that Endure The Bitter Lesson - Omar Khattab, DSPy & Databricks
Build LLM systems whose intent (signatures, metrics) is separate from prompts, so optimizers can re-compile them as scale-based models keep leapfrogging.
Agent Frameworks Considered Harmful — Rémi Louf, .txt
Simple event-driven agent definitions still require a reliable runtime that preserves history and handles distributed-system failures.
Your Agent Can Now Train Models — Merve Noyan, Hugging Face
Hugging Face Hub plus open coding agents and inference routing let you spin up, fine-tune, and serve open models from chat with near-zero glue code.
Full Walkthrough: Writing & Using Skills — Nick Nisi and Zack Proser
Skills give Claude portable, composable, scriptable units of expertise that route in by description and stop you from re-explaining context every conversation.
What if the harness mattered more than the model? - Aditya Bhargava, Etsy
Invest in harness design and controlled tool access to improve agent capability independently of model upgrades.
A Practitioner's Guide to Graphs - Tim Ainge, Good Collective
Graph value comes from consistent modeling and algorithms that exploit relationships, not from graph visualization alone.
Skill issue: Lessons from skilling up coding agents to use Langfuse - Marc Klingen, Clickhouse
Product skills should guide agents toward current documentation and application-specific workflows, then improve through trace feedback.
[Workshop] AI Pipelines and Agents in Pure TypeScript with Mastra.ai — Nick Nisi, Zack Proser
Mastra lets TypeScript devs build production AI pipelines with strongly-typed workflows and a single-stack alternative to Python frameworks.
Cooking with Agents in VS Code — Liam Hampton, Microsoft
Choose agent execution modes according to how much supervision each task needs and coordinate them from a shared development environment.
Tool Calling Is Not Just Plumbing for AI Agents — Roy Derks
Agents are only as good as their tools — invest in well-described, schema-typed, framework-independent tool layers, not just the agent loop.
Human-in-the-Loop Automation with n8n — Liam McGarrigle
n8n's visual workflows plus human-in-the-loop nodes let non-developers build inspectable, fixable agentic automations rather than opaque end-to-end agents.
Scaling Agents on Kubernetes with acpx and ACP — Onur Solmaz, OpenClaw
Standardized agent interfaces make remote coding sessions and repeatable maintenance workflows easier to orchestrate.
Building Reactive AI Apps: Matt Welsh
AI.JSX brings React's composability to LLM programs so frontend devs can stream and compose model calls as nested components.
The Unreasonable Effectiveness of Separating the Task from the Model — Maxime Rivest & Isaac Miller
Separate task contracts and quality criteria from implementation so AI programs can be composed and optimized safely.
Backlog.md: Terminal Kanban Board for Managing Tasks with AI Agents — Alex Gavrilescu, Funstage
Decompose big features into atomic markdown tasks with explicit acceptance criteria and an MCP-exposed workflow so AI agents stay in-scope and reviewable.
Building an ACP-Compatible Agent Live — Bennet Fenner, Zed
ACP separates coding-agent execution from editor UI through a small session-oriented protocol.
Ship Agents that Ship: A Hands-On Workshop - Kyle Penfound, Jeremy Adams, Dagger
Dagger turns agent loops into portable containerized workflows that ship to CI, letting you treat coding agents as just another reusable Dagger module.
Realtime multiplayer, automation, and you! — Idan Gazit, GitHub
Pair intelligent automation with deterministic boundaries and collaboration mechanisms that keep teams aligned.
Pragmatic AI with TypeChat: Daniel Rosenwasser
Use TypeScript types as both schema guidance and validator-driven repair loop to make LLM JSON outputs reliably consumable by typed application code.
remobi.app: Don't change your terminal workflow for mobile
Mobile agent supervision can reuse a developer's existing terminal session and custom workflow.
Skills are the New SDKs - Elvin Aghammadzada, DataRobot
Treat skills as an agent-facing interface that supplies the right operational knowledge at the right time.
Git push get an AI API: Ryan Fox-Tyler
Productionizing AI features means composing functions, models, and traditional code/dictionary validation — git push is the deploy primitive.
A Song of Types and Agents - Roberto Stagi, Ratel
TypeScript is a practical agent-building language when shared application types and ecosystem integration matter most.
[Full Workshop] How to add secure code interpreting in your AI app: Vasek Mlejnsky
Adding secure AI code execution to an app is now a few-hundred-line integration with e2b sandboxes plus Vercel AI SDK function calling.
Hypermode Launch: Kevin Van Gundy
Hypermode bets that the winning AI dev platform is one that lets you swap models, prompts and data with zero friction so iteration speed compounds.