Season 4 · Weeks 13–16 · 4 weeks
Season 4: Will.
FRIDAY decides its next step by itself: planning your week, using many tools, working while you sleep. Power needs limits.
FRIDAY starts deciding the next step by itself: an agent loop with a budget, a clean context, memory that forgets on purpose, approval before risky actions, and one plug (MCP) for your calendar and files. Then make its work survive restarts and ship a morning briefing that runs every day by itself (Project 4).
How you learn
An agent loop with a budget, memory that forgets, an MCP server and a job queue, written by hand.
What you build
A week planner, long-term memory with approvals, your calendar and files over MCP, and a daily briefing.
The weeks
- Week 13It decides the next step
You can build an agent loop, choose when to use a workflow, set step, time and money limits, explain the SDK, and show each step in a live feed.
- Week 14A clean desk
You can keep an agent's context small and sharp, give it a long-term memory that forgets on purpose, gate risky actions behind a person's yes, and grade the whole road it takes.
- Week 15One plug, many tools
You can build an MCP server that any AI app can plug in: tools agents use well, stateless requests, sign-in with OAuth, and a pause to ask the person when it matters.
- Week 16Work that survives
You can run FRIDAY's work as durable background jobs, test it with fakes, meter every token against a budget, package skills, and ship a release.
Projects
- Project 4: The morning briefing
Every morning FRIDAY gathers your calendar, tasks and notes and briefs you, by itself, within budget. Weeks 13–16.
Skills you prove
Tiles on the periodic table of skills that this season lights up.
- Raw agent
- Build a tool-using agent with an MCP server in an afternoon, without a framework (week 13)
- Context
- Design the context for a long-running agent: compaction, memory, sub-agents (week 14)
- MCP
- Build an MCP server that gives any AI app safe, documented tools (week 15)
Checkpoint test: Will
After week 16. A morning briefing that runs by itself for a week.
- Blank file, 45 minutes: an agent loop with a step limit and a cost cap that passes your tests
- Your morning briefing has run by itself every day for a week, and survived a restart
- FRIDAY asks before every risky action, and a trajectory eval fails if it doesn't
- Explain to a non-technical friend what FRIDAY can and can't do on its own