You've automated research workflows. Now: what if your code could read a 40-page paper and hand you three key findings in structured JSON? Before we build that, tell me — what's the research task that still eats two hours of your week?
Literature review, honestly. I read ten abstracts, take notes, and half my notes are redundant by the time I actually sit down to write.
By Day 25 an agent does that first pass for you. But first — rate yourself on six skills. Be honest about where you are, not where you wish you were.
I've used ChatGPT. Building something with it feels completely different though.
Using an AI tool and calling an AI model in code are different skills. The second one is teachable in 30 days. Rate where you are and we'll handle the rest.
Thirty days. One real project: an AI research assistant. By Day 28 you'll have an agent that takes a research question, searches recent literature with perplexity/sonar, extracts three structured findings, and formats them as a markdown reading list.
| Week | Focus | Student framing |
|---|---|---|
| 1 | Basic agents — prompt, run, output | "Ask a question. Get a structured answer. One function." |
| 2 | Structured output — Pydantic, Literal, list | "The LLM must return this exact shape. No guess-and-check." |
| 3 | Pipelines + batch | "Classify 20 open-text survey responses in one loop." |
| 4 | Web search + tools + capstone | "Search literature, extract findings, format a reading list." |
Every lesson runs against real AI models — PydanticAI over OpenRouter.
Create a free account to get started. Paid plans unlock all tracks.
You've automated research workflows. Now: what if your code could read a 40-page paper and hand you three key findings in structured JSON? Before we build that, tell me — what's the research task that still eats two hours of your week?
Literature review, honestly. I read ten abstracts, take notes, and half my notes are redundant by the time I actually sit down to write.
By Day 25 an agent does that first pass for you. But first — rate yourself on six skills. Be honest about where you are, not where you wish you were.
I've used ChatGPT. Building something with it feels completely different though.
Using an AI tool and calling an AI model in code are different skills. The second one is teachable in 30 days. Rate where you are and we'll handle the rest.
Thirty days. One real project: an AI research assistant. By Day 28 you'll have an agent that takes a research question, searches recent literature with perplexity/sonar, extracts three structured findings, and formats them as a markdown reading list.
| Week | Focus | Student framing |
|---|---|---|
| 1 | Basic agents — prompt, run, output | "Ask a question. Get a structured answer. One function." |
| 2 | Structured output — Pydantic, Literal, list | "The LLM must return this exact shape. No guess-and-check." |
| 3 | Pipelines + batch | "Classify 20 open-text survey responses in one loop." |
| 4 | Web search + tools + capstone | "Search literature, extract findings, format a reading list." |
Every lesson runs against real AI models — PydanticAI over OpenRouter.