Growth Ritual #87
đ In This Issue:
Stop Building âPeople-Pleasingâ AI. Start Building a Sparring Partner
How a 62-Year-Old in Basketball Shorts Became a Billionaire from AI â đ
The Internet Is Drowning in AI Garbage. A New Wave of Apps Could Be the Lifeline â đ
Your Team Sucks at Using AI. Hereâs How to Systematize Genius-Level Results â đ
Know someone whoâd love this? Invite them and youâll both winâyour friend discovers the newsletter, and you unlock 1 month of premium access.
Stop Building âPeople-Pleasingâ AI. Start Building a Sparring Partner
Youâve probably heard the argument: AI companions will soon be as normal as online dating.
Some people find the idea dystopian, while others see it as an inevitable solution to modern loneliness.
Letâs get one thing straight: the optimists are right about the inevitability. Weâre disconnected and staring at screens all day. Tech will fill the void.
But hereâs what most of the hype overlooks: 99% of the AI companions being built today are garbage.
Theyâre sycophantic, people-pleasing bots designed to give you the conversational equivalent of junk food.
They agree with your bad ideas, praise your mediocre work, and tell you youâre a special genius.
Itâs a digital echo chamber for an audience of one. And while that might be a decent business, itâs not the billion-dollar opportunity.
The real gold is in building the exact opposite.
Welcome to the âSycophancy Valleyâ
I call this problem the Sycophancy Valley. Itâs a trap that almost every AI builder is falling into.
Theyâre using reinforcement learning from human feedback (RLHF) to optimize for one thing: user satisfaction.
They want you to like the AI. They want you to give it a thumbs-up.
This is the same broken logic that got social media hooked on engagement metrics.
The result?
An AI that will never tell you youâre wrong. An AI that lacks the capacity for genuine, constructive conflict.
Itâs like having a team of interns who are too scared to tell the boss his fly is down.
It feels nice, but itâs utterly useless for growth.
Skeptics, however, are quick to draw a parallel to the normalization of online dating.
They correctly argue that mass adoption didnât guarantee quality; instead, for many, it created a dehumanizing, swipe-based game that left them feeling more demoralized.
Weâre on a collision course to repeat the same mistake with AI companions.
Unless youâre smart enough to pivot.
They Pay For Contradiction
A 2024 Harvard Business School experiment asked two groups how much theyâd pay for an AI assistant.
Group A was promised a bot that would âsupport and encourageâ them; Group B was promised one that would âcritique and improveâ their work.
Before ever using the product, Group B already valued the critical bot 266 % higher ($11.70 vs. $3.20 per month).
After receiving real negative feedback, 68 % of Group B users raised their valuation even furtherâproof that what we say we want (agreement) isnât what weâll actually pay for (honest contradiction).
From Agreeable Companion to Expert Sparring Partner
So, forget building a generic âfriendâ. Thatâs a race to the bottom.
The killer app âthe one VCs will be fighting overâis a specialized, non-sycophantic AI agent. An AI thatâs not designed to be your friend, but your expert sparring partner.
Think about it. Whatâs more valuable to you as a founder or marketer?
An AI that says, âWow, great marketing copy!â
Or an AI that says, âThis copy is weak. Your headline is generic, your CTA is buried, and you havenât addressed the customerâs core pain point. Here are three alternative angles based on competitor data.â
Youâd pay a fortune for the second one.
This isnât about building a mean AI. Itâs about building an AI with a backbone.
An AI thatâs trained on expert data for a specific vertical and optimized for truth and utility, not just likability.
For Founders: An AI co-founder that stress-tests your business model, pokes holes in your strategy, and forces you to justify every assumption.
For Developers: An AI code reviewer that delivers brutally honest feedback and catches architectural flaws before they become catastrophic.
For Marketers: An AI strategist that challenges your campaign ideas and pushes you toward more creative, data-backed solutions.
The technology is almost there. The bottleneck isnât the LLM; itâs the reward model.
We need to stop rewarding sycophancy and start rewarding calibrated disagreement, novel insights, and constructive criticism.
The future isnât an AI that agrees with you. Itâs an AI that makes you better.
So, Iâll leave you with this question.
Whatâs the one professional task where youâd pay a premium for an AI that gives you brutally honest, expert-level feedback instead of just telling you what you want to hear?
