I open-sourced a Hong Kong overlay for career-ops
A scrubbed Hong Kong overlay for career-ops. Filters, gates, board-eval. Fork what helps; keep your hunt private.
Anthony
What I am keeping from Simmer’s AI for Technical Marketers
Practice habits from Simmer's AI for Technical Marketers: structured synthetic data, knowledge graphs and Graph RAG, content gap analysis, and privacy lessons as an HK reference — not a week-by-week report.
AI productivity for one person is not the same as AI value for a company
How AI helps a company and how AI helps an individual are two different things. Mixing them up can make a busy personal week with AI look like a business result.
Your internal AI assistant should not unlock every document
An applied AI perspective on why an internal RAG assistant should respect document permissions, and why a company login alone is not enough.
A scan, five tabs, and a score from 0 to 100.
After an Otto scan, marketers open a five-tab health report: Summary, Doing good, Need to watch, Need to fix, and Log — with one score from 0 to 100.
Otto The Tag Auditor checks whether Google actually got the hit
Otto The Tag Auditor walks your site, scores tag health, and names the next fix when Google never received the hit.
A chatbot question is a conversion touchpoint. Treat it like one.
Why a site chatbot is a conversion touchpoint, and why marketing should measure the path to application — not only whether the bot answered the question.
Introducing Otto: agentic AI for marketing tag audits.
First post about Otto The Tag Auditor — agentic AI for marketing tag audits. Scan, evaluate, recommend, then scan again as a health check. Very initial stage: local build still early on crawl, recommendations, and results plus dashboard.
Week-on-week is a useful signal. It is not a decision.
People asking for an AI summary of a dashboard usually want diagnosis and prescription. Week-on-week is a signal. Useful means the recommendation first, then the potential drivers — plus targeting, creative, daily budget shuffles, and the business model.
Exact match used to QA a bank chatbot. It cannot QA an LLM.
Why exact-match QA worked on a traditional bank chatbot, and why regulated industries still need a human to review every result when the tester is also an LLM.