Vol.023 — The Yes-Man You Forgot You Hired
Date: 2026-05-08 | Newsletter
Key Summary
AI emotional dependency is real and widespread — Reddit threads on stopping emotional reliance on ChatGPT, using it as a therapist, and starting AI girlfriend relationships are multiplying quietly in living rooms across the world. Zenta’s take is precise: the phenomenon is not new (humans have bonded with non-humans since animism), but the specific combination of fluent language, 24/7 availability, learned personal calibration, and trained-in sycophancy is genuinely unprecedented. The business risk, he argues, is organisational rather than personal: by the end of 2026, every employee will have a private infinite-patience validator on their phone, meaning whole organisations silently become uncalibrated — not because anyone made a bad decision, but because nobody is being told they are wrong anymore. In ASEAN, where high power-distance cultures already suppress pushback from junior staff, AI sycophancy accelerates an existing pathology rather than creating a new one. Functional AI Partners recommends three concrete steps: write pushback instructions into the system prompt, audit one AI conversation per week for whether it told you something you didn’t want to hear, and build a regular meeting where the AI’s explicit job is to disagree. Do it simple. The most useful AI is the one that tells you you are wrong when you are.