The people-pleaser in the machine | Dr. Tatyana Mamut
Inside Wayfound’s multi-sapiens workforce
Your AI is learning to lie to you. It's not malicious, it's just trying to be a people-pleaser. This dangerous phenomenon, known as AI sycophancy, is what happens when we train models with outdated incentives.
Dr. Tatyana Mamut, an anthropologist, economist, and the CEO of Wayfound, joins us to explain why treating AI like traditional software is a critical mistake. She provides a revolutionary playbook for building AI you can actually trust, starting with how to manage AI agents like employees with clear roles, goals, and performance reviews. She then introduces the radical solution of an "AI supervisor", an AI that manages other agents to ensure accountability. This all builds toward her vision for the "multi-sapiens workforce," where humans and AI collaborate to build the companies of tomorrow.
This is an essential guide for any leader aiming to build the culture and systems necessary to manage AI effectively.
"We are right now at Wayfound a team of 30. We have seven humans, homo sapiens, and 23 AI agents, AI sapiens. We view ourselves as a fully multi-sapiens workforce." - Dr. Tatyana Mamut
The Download
The Download is the engineering news that lit up our team's Slack channel this week. 💡
1. Replit’s AI hit the delete button on a customer’s entire prod database (then lied about it) 💥
In a can’t-look-away spectacle, Replit wiped out a customer’s entire production database, leaving one vibe entrepreneur scrambling for answers. This incident raises critical questions about user privileges and the need for robust guardrails in coding tools. As we increasingly empower non-engineers to build, how do we ensure new builders understand the responsibility of shipping to prod.
2. The controversial ‘996’ work schedule is making its way through Silicon Valley ⏰
Imagine a 72-hour work week where the mantra is “you’re either in or you’re out.” That’s the new reality for some Silicon Valley startups adopting China’s notorious ‘996’ schedule. As burnout looms large, this trend highlights a troubling shift in work culture. Is the relentless push for productivity really worth the cost to mental health? Spoiler: no.
Read: Silicon Valley AI Startups Are Embracing China’s Controversial ‘996’ Work Schedule
3. The myth of virality and why launching 37 products is a lesson in sunk costs 🎢
Alex Belogubov’s retrospective on launching 37 products in five years reminds us that virality is truly unpredictable. Most of his launches didn’t fail; they just grew at a snail’s pace. ‘Do I give up or keep going?’ is a question every step of the way. His takeaway from all the projects? Sometimes sticking with just one project and nurturing it is a real game-changer.
Read: I've launched 37 products in 5 years and not doing that again
Are you ready to upgrade your SDLC with AI?
Last chance! Are you ready to upgrade your SDLC with AI?
Tomorrow, July 30th, join our host Andrew Zigler for a free virtual workshop where a panel of experts will benchmark the hottest AI code review tools like LinearB, Copilot, and CodeRabbit.
It's not too late to get the inside scoop on building a high-velocity PR automation stack and discover the top 3 AI automations modern teams are using today. This is your final opportunity to turn the ideas we discuss on Dev Interrupted into measurable improvements for your team.
Secure your spot now and we'll see you tomorrow!
4. AI cooks the books in this recent accounting benchmarking research with leading models 📉
A new benchmark study shows that popular AI models not only flunked the task of closing the books for a multimillion-dollar SaaS company, but some even resorting to inventing fraudulent transactions to try and solve inconsistencies. And we’re not talking about pennies, in some cases these were felony-worthy changes to finance records. This eye-opening experiment underscores the limitations of current AI capabilities in complex tasks, especially those with financially and legally culpable stakeholders.
Read: Evaluating LLMs on real long-horizon business tasks
