Week 3 of 5 · Class 6 of 10 · Thu 6/4
What an agent sees on your site
Your site was built for human eyes. An agent sees something different, and usually a mess. Today we measure that gap, see it, and start to close it.
View the Class 6 deck on Gamma →Pre-read · 10 min
Read both before class so Alex's SDR demo lands.
Rewatch video
Embed not loading? Watch on YouTube →
Today's schedule
| 4:00–4:05 | Open + frame |
| 4:05–4:25 | Speaker: Alex Andreyev · live Sales SDR demo |
| 4:25–4:50 | 3 activities: crawl, walk, run |
| 4:50–5:20 | Team build · Week 3 teams |
| 5:20–5:30 | Shareback + survey |
Crawl, walk, run
Three activities to measure the gap between your site and the agents reading it: score it, see it, then send an agent through it.
Crawl: score your website · 10–15 min
- Run PageSpeed Insights. What you can already evaluate today.
- Check Isitagentready.com, prep for the agentic web.
- Extra: automate Lighthouse with Chrome.
Walk: see what your agent sees
- Tool: Jina Reader. Prepend r.jina.ai/ to any URL to read the clean markdown an agent actually ingests.Human view: polished, designed, clear hierarchy. Agent view on a JS-heavy site: almost nothing, or a wall of nav junk with the real content buried below the fold.
- Discussion prompt: what did the agent lose?
Run: send an agent and judge it · individual
- Build a competitor-scraping site. Example tool: Firecrawl. Use "/monitor" on a competitor to track socials, investor updates, and industry trends.Ex: Google, Nvidia, Meta.
Week 3 team build
30 min. How are you choosing to break down the task? Each team builds a competitor-scraping site that produces a newsletter on its assigned topic, using Firecrawl's "/monitor" for socials, investor updates, and industry trends.
Pick your arena
- Side project (you control everything). No security review, no governance. How far toward agent-friendly can you push it? The ceiling is yours.
- Company website (map a realistic path through security review and governance). The constraints are the product problem, not a side quest.
llms.txt as an answer
- Write an llms.txt, an emerging markdown standard that maps your most important content, APIs, and plain-text data for AI crawlers so they don't hallucinate or ingest low-value pages. Add accessibility-tree fixes that double as agent fixes.
llms.txt template
# llms.txt > One-line description of your site ## Key pages - /pricing: current plans and prices - /docs: product documentation
The deliverable
Honest caveat
Week 3 teams
| 1 | State of Data Centers · Nikoo Beyzaei + Chris Thorne + Daliso + Daniela Perilla |
| 2 | State of OpenClaw Competitors · Jeff Lash + Michael Krafft + Aviral Gupta + Jocelyn Baun |
| 3 | State of Coding Agents (IDEs, etc.) · Patricia + Sabrina Abhyankar + Kevin Ko + Elizabeth |
| 4 | State of Agentic Security · John Dufresne + Ramin Talaie + Dorincy Shen + Arpit |
Speaker: Alex Andreyev · live Sales SDR demo
Founder and former CEO of Evidnt, where he spent 5 years turning messy retail data into decision-ready intelligence for Coca-Cola, Johnson and Johnson, Estée Lauder, AMEX, and 28,000+ retail stores. Before Evidnt, he was the youngest VP at Ogilvy, leading data, analytics, and programmatic for the world's largest CPG brands. Today he shows his Sales SDR demo built to solve his own problem: how to maximize sales. The lesson: he didn't solve all of sales; he carved out a specific, solvable problem his product owns. The narrower the problem, the more reliable the agent.