I finished Automation Foundations. I can call tools, send things, idempotency mindset is there. What's next?
The reliability layer. Foundations taught you how to write a script that works. Patterns teaches you how to write a script that keeps working — at 9am every morning, when the network blips, when one item out of ten fails, when you re-run because the script crashed midway.
What does that look like in code?
Six new primitives. Retry with backoff — the call flaked, try again with a delay. Dedup keys — track which items you've already processed, in a tool not in memory. State across runs — read "last_processed_id" from a Sheet at start, write the new value at end. Partial failure — 7 of 10 items succeeded, log the 3 failures, don't crash. Structured logs — tagged lines like {"step": "send", "status": "ok"} that future-you can grep. Pre-flight validation — check inputs before causing side effects.
And week 4?
Patterns at scale. You'll page through hundreds of items, cache reads, chain transforms, and orchestrate 4 tools with reusable helpers. The final synthesis composes 5 prior primitives on a 3-element generic input — same minimum-demonstration shape Foundations used.
Same six prompts again?
Same six. Rate yourself today. Strongly disagree is fine — the prompts come back on day 30 to mark your delta.
Automation Patterns adds the reliability primitives to your kit. Foundations taught the what — calling tools, sending, creating, appending. Patterns teaches the how do I make this survive Tuesday — the patterns that turn a script you ran once into a script you can schedule.
Week 1 — Multi-step chains and branching. 3-step chains (read → transform → write), reshaping one tool's response for another, conditional routing, aggregation, partition, recognizing specific error classes from tools.
Week 2 — Reliability primitives. Retry with exponential backoff, predicting idempotency, dedup keys persisted to a Sheet, state across runs, partial-failure handling (process 5, fail 1, report counts), structured logging.
Week 3 — Schedule-safe production patterns. Auditing scripts for "safe at 9am every day", incremental processing using last-run state, failure self-alerts, configuration extraction, pre-flight validation, the full reliable-chain synthesis.
Week 4 — Patterns at scale. Pagination over many pages, streaming-style read+process, caching read responses, chained-result transformations, multi-tool orchestration with reusable helpers, a final synthesis combining 5 prior primitives.
Write Python automations that you can leave running on a cron — they retry transient failures, dedup against state stored in a tool (not in memory), log their progress in a parseable shape, fail loud on misconfiguration, and recover cleanly when re-run. The kit you need before any meaningful real-world automation.
Automation Foundations completed (or equivalent fluency with Composio tool calls). Python comprehensions and exceptions are used freely — Python Patterns recommended but not strictly required.
Rate each statement honestly on the 1-5 scale. The same prompts come back on day 30 to mark your delta. No grade — just calibration.
I finished Automation Foundations. I can call tools, send things, idempotency mindset is there. What's next?
The reliability layer. Foundations taught you how to write a script that works. Patterns teaches you how to write a script that keeps working — at 9am every morning, when the network blips, when one item out of ten fails, when you re-run because the script crashed midway.
What does that look like in code?
Six new primitives. Retry with backoff — the call flaked, try again with a delay. Dedup keys — track which items you've already processed, in a tool not in memory. State across runs — read "last_processed_id" from a Sheet at start, write the new value at end. Partial failure — 7 of 10 items succeeded, log the 3 failures, don't crash. Structured logs — tagged lines like {"step": "send", "status": "ok"} that future-you can grep. Pre-flight validation — check inputs before causing side effects.
And week 4?
Patterns at scale. You'll page through hundreds of items, cache reads, chain transforms, and orchestrate 4 tools with reusable helpers. The final synthesis composes 5 prior primitives on a 3-element generic input — same minimum-demonstration shape Foundations used.
Same six prompts again?
Same six. Rate yourself today. Strongly disagree is fine — the prompts come back on day 30 to mark your delta.
Automation Patterns adds the reliability primitives to your kit. Foundations taught the what — calling tools, sending, creating, appending. Patterns teaches the how do I make this survive Tuesday — the patterns that turn a script you ran once into a script you can schedule.
Week 1 — Multi-step chains and branching. 3-step chains (read → transform → write), reshaping one tool's response for another, conditional routing, aggregation, partition, recognizing specific error classes from tools.
Week 2 — Reliability primitives. Retry with exponential backoff, predicting idempotency, dedup keys persisted to a Sheet, state across runs, partial-failure handling (process 5, fail 1, report counts), structured logging.
Week 3 — Schedule-safe production patterns. Auditing scripts for "safe at 9am every day", incremental processing using last-run state, failure self-alerts, configuration extraction, pre-flight validation, the full reliable-chain synthesis.
Week 4 — Patterns at scale. Pagination over many pages, streaming-style read+process, caching read responses, chained-result transformations, multi-tool orchestration with reusable helpers, a final synthesis combining 5 prior primitives.
Write Python automations that you can leave running on a cron — they retry transient failures, dedup against state stored in a tool (not in memory), log their progress in a parseable shape, fail loud on misconfiguration, and recover cleanly when re-run. The kit you need before any meaningful real-world automation.
Automation Foundations completed (or equivalent fluency with Composio tool calls). Python comprehensions and exceptions are used freely — Python Patterns recommended but not strictly required.
Rate each statement honestly on the 1-5 scale. The same prompts come back on day 30 to mark your delta. No grade — just calibration.
Create a free account to get started. Paid plans unlock all tracks.