Dashboard
Learning Paths
demoSeven tracks, A through G — every track taught against Claude and OpenAI side by side.
Foundations
Call Claude and OpenAI from a script. Pick the right one per task.
Building agents
From "model that answers" to "agent that acts." Plans, tools, recovery.
Retrieval-augmented generation
Ground answers in private data — embeddings, chunking, citations.
Voice and text agents
Realtime voice + text chat. Latency budgets, turn-taking, barge-in.
Connecting agents to data and tools
MCP + safe data access. Let agents read and act on your real systems.
Evaluating agents
Golden sets, LLM-as-judge, A/B testing. Know when an agent gets better.
Fine-tuning and model customization
When prompts and RAG aren't enough. OpenAI fine-tune + Claude customization.
What's new this week
Track A CLI now compares both providers head-to-head
2 new lessons in Track B build the same research agent on each
Added to Track C — measurable +12% recall@5 over single-stage
Track D voice agent: <250ms first audio chunk
Track E now ships a starter MCP server template