Thirty days. From your first 3-step chain to a synthesis combining six primitives across one tool with multi-tool shape. Same six prompts you saw on day 1 — rate yourself again.
Where do I go next?
Three directions, depending on what you build next.
AI Patterns if you want LLMs in the same reliable-chain shape. The kit you have here transfers directly — retry around complete(), dedup on prompts, structured logs of generations, partial failure when one item's response doesn't parse. The orchestration is identical; the side effect is complete() instead of gmail.send.
Automation Advanced (when v1 expands) for webhooks, durable workflows, observability dashboards, cross-service consistency. The next layer up — what teams build when one script becomes a system.
Or just build something real. Pick a workflow you do every day manually. Apply the Patterns kit: 3-step chain → retry → dedup against a Sheet → structured logs → schedule it. The kit is the same; only the inputs change.
Rate the prompts below as honestly as you did on day 1.
Four weeks. Twenty-four practice scripts. Your kit:
| Capability | Lessons |
|---|---|
| Compose chains | L1-L7 (week 1: 3-step chain, reshape, route, aggregate, partition, error classes) |
| Reliability primitives | L8-L14 (week 2: retry, idempotency, dedup, state, partial failure, structured logging) |
| Schedule-safe patterns | L15-L21 (week 3: audit mindset, incremental, alerts, config, validation, synthesis) |
| Patterns at scale | L22-L28 (week 4: pagination, streaming, caching, transforms, helpers, final synthesis) |
Every one demonstrated on the smallest possible generic input. The skill: pattern recognition + clean assembly.
Deferred to Automation Advanced:
None are required for your first dozen real-world automations. The kit you have is enough.
→ AI Patterns — LLMs with the same reliable-chain shape. Retry around complete(). Dedup on prompts. Structured logs of generations. Same kit, new side effect.
→ Or pick a workflow you do every day. Apply the Patterns kit:
# 1. Pre-flight — what are the required inputs?
# 2. Helpers — one per tool you talk to
# 3. State — read last-run from a Sheet
# 4. Loop with retry + dedup + structured log per item
# 5. Summary log + alert on failure
# 6. Schedule itThat skeleton is the shape of every production automation you'll write. Rate the prompts below as honestly as you did on day 1.
Thirty days. From your first 3-step chain to a synthesis combining six primitives across one tool with multi-tool shape. Same six prompts you saw on day 1 — rate yourself again.
Where do I go next?
Three directions, depending on what you build next.
AI Patterns if you want LLMs in the same reliable-chain shape. The kit you have here transfers directly — retry around complete(), dedup on prompts, structured logs of generations, partial failure when one item's response doesn't parse. The orchestration is identical; the side effect is complete() instead of gmail.send.
Automation Advanced (when v1 expands) for webhooks, durable workflows, observability dashboards, cross-service consistency. The next layer up — what teams build when one script becomes a system.
Or just build something real. Pick a workflow you do every day manually. Apply the Patterns kit: 3-step chain → retry → dedup against a Sheet → structured logs → schedule it. The kit is the same; only the inputs change.
Rate the prompts below as honestly as you did on day 1.
Four weeks. Twenty-four practice scripts. Your kit:
| Capability | Lessons |
|---|---|
| Compose chains | L1-L7 (week 1: 3-step chain, reshape, route, aggregate, partition, error classes) |
| Reliability primitives | L8-L14 (week 2: retry, idempotency, dedup, state, partial failure, structured logging) |
| Schedule-safe patterns | L15-L21 (week 3: audit mindset, incremental, alerts, config, validation, synthesis) |
| Patterns at scale | L22-L28 (week 4: pagination, streaming, caching, transforms, helpers, final synthesis) |
Every one demonstrated on the smallest possible generic input. The skill: pattern recognition + clean assembly.
Deferred to Automation Advanced:
None are required for your first dozen real-world automations. The kit you have is enough.
→ AI Patterns — LLMs with the same reliable-chain shape. Retry around complete(). Dedup on prompts. Structured logs of generations. Same kit, new side effect.
→ Or pick a workflow you do every day. Apply the Patterns kit:
# 1. Pre-flight — what are the required inputs?
# 2. Helpers — one per tool you talk to
# 3. State — read last-run from a Sheet
# 4. Loop with retry + dedup + structured log per item
# 5. Summary log + alert on failure
# 6. Schedule itThat skeleton is the shape of every production automation you'll write. Rate the prompts below as honestly as you did on day 1.
Create a free account to get started. Paid plans unlock all tracks.