This project contains a Val Town cron job that watches the Anthropic Greenhouse job board and tracks changes over time.

anthropic-job-tracker-cron.ts:

• Fetches all pages from https://job-boards.greenhouse.io/anthropic
• Extracts job post URLs from the page's window.__remixContext data
• Visits each job post page and stores the raw job payload JSON
• Compares the latest scrape against previously stored jobs
• Detects newly added jobs
• Detects removed jobs and archives them
• Records each time a job was observed
• Emails a summary when new jobs are found

The script runs as an exported async default function, which makes it suitable for use as a Val Town scheduled job.

Its flow is:

1. Ensure the SQLite schema exists
2. Load previously known jobs from the jobs table
3. Scrape the Anthropic Greenhouse board page by page
4. Hydrate each job with its detailed page payload
5. Compute:
   • newJobs: jobs now present that were not previously stored
   • removedJobs: jobs previously stored that no longer appear
6. Insert observation records for all currently visible jobs
7. Upsert removed jobs into archived_jobs
8. Replace the current contents of jobs with the latest snapshot
9. Send an email if any new jobs were found

The script creates and uses three SQLite tables:

Current snapshot of active job postings.

• url - primary key
• raw_json - serialized job payload from the job detail page
• last_seen_at - timestamp of the most recent sync

Jobs that used to exist on the board but disappeared in a later run.

• url - primary key
• raw_json - last known serialized payload
• archived_at - timestamp when the removal was detected

Append-only history of when a job was seen.

• url
• seen_at

When new jobs are detected, the script sends an email with:

• the number of new jobs
• each job title, if available
• each job URL

At the end of a successful run, the function returns an object like:

{
  ok: true,
  syncedAt: "2026-04-09T00:00:00.000Z",
  count: 42,
  newCount: 3,
  archivedCount: 1
}

• The scraper depends on Greenhouse exposing data inside window.__remixContext
• Job details are cached indirectly by reusing stored raw_json when a known URL already exists
• syncJobs() replaces the jobs table contents on every run so that table always reflects the latest snapshot