Mining digital exhaust
summary
Workshop shows how to turn personal traces into concrete findings by actually scraping, coding and prompting: uses Chrome DevTools Protocol (CDP) + Playwright and agentic Copilot/Codex/Claude to scrape logged‑in pages (LinkedIn invites → ~600 rows), builds a 120‑line Python scraper, and runs bookmarklet scrapers for WhatsApp. Analysis finds a 23.5 kg drop with a near‑linear phase (≈1 kg/week, R² ≈ 0.97) then plateau; YouTube watch activity concentrates noon–5 pm (≈48% of views). Emphasizes postmortem prompts, style‑transfer narration (Gladwell/NYT), and reusable topic‑modeling tooling.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:27Personal data analysis
- 2:53Data extraction techniques
- 9:59Building scrapers
- 27:09Scraping LinkedIn invites
- 47:36Bookmarklets for scraping
- 1:00:13Data analysis insights
- 1:21:42Style transfer for narratives
- 1:48:42Personal data insights
- 2:01:05Topic modeling
- 2:19:24Visual insights generation
- 2:39:10Summary
related talks
19mHow to get clean web data for chatbots and LLMs
10mFrom Bots to Brains: The Evolution of Web Scraping
40mPython For Self-Trackers: How To Become A Data-Driven You
01h 05mWhat the hell am I doing?
49mHow to get smarter while doing your job
23mBeyond Scraping: The Future of Web Data Access with WebMCP
29mGiant Pile of Data
29m5 Simple Steps to Create Meaningful Features from Clickstream Data