Turn repeated actions into a workflow
Choose a repeated activity. What starts it, what inputs does it need, what happens in order, and what is produced? A schedule is only one trigger; a button, file, or event can trigger work too.
Whether automation is worthwhile depends on frequency, failure cost, and maintenance. Walk through a small sample before scheduling it indefinitely.
Collect articles, extract titles and links, remove duplicates, and create a daily digest draft. Make the draft reliable before adding sending.
Take this with you · Describe the workflow and output criteria before choosing tools.
Understand the application you're working with
Webpages generally combine HTML structure, CSS presentation, and JavaScript behavior. The DOM represents page structure in the browser. A page may request backend data. Mobile apps may be native, cross-platform, or contain web views.
Not every application needs a remote backend, and reading the screen isn't always the best way to obtain data. Find where information and actions actually live before selecting a connection method.
A displayed price may already come from a structured endpoint. Reading that endpoint and moving a mouse to find a price have different maintenance costs.
Take this with you · Find the source of the data before automating the interface.
APIs and pages have different boundaries
A documented public API often provides explicit formats, permissions, and rate limits. Internal page endpoints can change and require understanding the authorized and permitted use.
When no suitable interface exists, parsing or browser automation may help, but login, layout changes, and dynamic loading affect reliability. robots.txt is one crawler instruction mechanism, not a substitute for access permission.
For a weather digest, establish the source, permitted frequency, and error responses before implementing requests.
Take this with you · The access method affects reliability and maintenance.
Extract fields you can validate
Read structured JSON by field name and HTML through the DOM and selectors. Models can help extract information from free text, but may omit, confuse, or invent fields.
Validate required fields, formats, and unusual values regardless of how extraction happens. Preserve traceable sources. A missing price shouldn't silently become zero, nor an uncertain date become a confirmed fact.
A digest item needs a title and link. Flagging or skipping a missing title is better than presenting an uncheckable item as complete.
Take this with you · Treat extracted content as data that still needs validation.
Act on observable page states
Browser automation can navigate, fill fields, click buttons, read content, and capture screenshots. Reliable steps wait for observable conditions instead of always sleeping a fixed number of seconds and guessing readiness.
Consider login, permissions, layout changes, and failure messages. A submitted click doesn't prove a server accepted the operation. Check feedback and the final data.
After submitting a form, check the success state or result record. Keep steps and screenshots as evidence when something fails.
Take this with you · Check the result of an automated action, not just that the action happened.
Scripts, agents, and safe repetition
A fixed script follows predefined steps. An agent can choose the next tool based on a goal and feedback. Flexible decisions require clear permissions, checks, and stopping conditions. Both approaches can work together.
Deduplication and idempotent design help avoid duplicate creation or sending on repeat runs. Failures need bounded retries, records, and a route to human review. One successful run doesn't prove daily reliability.
A digest can deduplicate links, skip incomplete entries, and produce a draft first. Sending later needs clear recipients, confirmation rules, and sent-item records.
Take this with you · Describe successful, failed, and repeated execution in a professional automation brief.
Run an article-digest pipeline
Four sample inputs contain a duplicate link and a missing title. Add validation and deduplication to see the output change.
Uses local fixtures. It doesn't visit websites, send email, or create scheduled tasks.
A daily digest retries after a network problem. How can you reduce duplicate sending?
Choose an answer before reading the explanation.
Write an implementable digest brief that covers failure and repetition.
Think first, then uncover help as needed. Later chapters expect more complete descriptions and clearer checks.
Bring this chapter's context to your AI. Going deeper is optional; you can keep reading without it.
Explain triggers, steps, outputs, and failures using a daily inbox routine.
One approach you can make your own
I'm learning computers and AI-assisted programming from scratch. In this chapter I learned: Automation：Break repeated work into steps and interfaces；Distinguish scripts, browser automation, and agents；Design validation, deduplication, failure handling, and output checks Explain triggers, steps, outputs, and failures using a daily inbox routine. Check my understanding first. Explain one small question at a time with a concrete example and the limits of any analogy. Ask a scenario question, then adapt to my answer. Distinguish facts, guesses, and uncertainty; suggest hands-on ways to check. Confirm the version before giving tool-specific advice, and don't invent features or references.
Complete this chapter
Mark the sections as read, finish the experiment, and pass the scenario check to record completion. You can visit the next chapter at any time.
Read the original references
This is an introduction. These official resources help you check details and explore further.
Playwright · Auto-waitingAnthropic · Workflows and agentsYou've connected machines, programs, networks, data, and AI. Keep learning while building something that matters to you.