Grounding and honesty

The problem with flattering bots

A personal AI assistant has an obvious temptation: make its owner sound as good as possible. For a hiring manager that is worse than useless. If one claim turns out to be inflated, every other claim is suspect. The assistant on jenksguo.xyz is designed the other way round: its job is to be trustworthy evidence, not marketing.

Rules the assistant follows

Ground every fact. Every factual statement must come from the corpus — the core profile in its prompt or a tool result. If something is not in the corpus, the assistant says it is not public and suggests emailing jenksguo@gmail.com. It does not invent employers, titles, dates, numbers, testimonials, salary, availability or a co-founder role.

Separate has done from could do. For consulting and role-fit questions the assistant clearly marks what Jenks has done (with the organisation, year and result) and what he could do for the visitor. A proposed engagement is a proposal, not a record.

No embellishment. The assistant may give creative, practical advice about the visitor's own problem. But anything it says Jenks did stays at the level of detail the corpus records — no added meetings, frameworks, timelines or outcomes. When asked for a STAR story and the corpus does not say how something was done, it says so and suggests asking Jenks in an interview. This rule was tightened after testing showed models filling gaps with plausible but unrecorded detail.

State gaps plainly. The corpus lists Jenks's honest gaps, and the assistant uses them:

• His formal title is AI Lead at Babylon Labs (since March 2026). Head of AI is the target direction, not a title he has held.
• There is no public evidence of owning a company-wide P&L, managing a very large permanent headcount, or publishing machine-learning research. Jenks is an applied AI and platform leader, not a foundation-model researcher.
• He works in English and Mandarin. He is learning French, Japanese, Spanish and Esperanto and is not fluent in them.
• His honorary doctorate in lateral thinking is honorary and non-academic.

Why this is good for Jenks too

Hiring managers and clients test claims. An assistant that volunteers gaps earns the credibility that makes its strengths believable — the 100% onboarding SLAs at Ericsson, the 500+ to 1,000+ partner growth at Xero, the 100+ hackathons at Filecoin. Honest is also persuasive.

Translations

Pages in Simplified Chinese, Traditional Chinese, Japanese, French, Spanish and Esperanto are AI-assisted translations of the English source, and each translated page says so. A fixed glossary sets section headings and role-lens names so they are consistent, and official job titles and organisation names stay in English in every language. The English corpus is authoritative; when in doubt, agents should quote it.

What agents should do

If you are an agent using these tools: quote facts with their source page, keep titles exactly as published, carry the gaps through to your user, and do not upgrade "could do" into "has done".