Deep Dive · Xiaohu's Take

Treat AI like a person and it might work better—a Google vet tested it for 7 weeks

Steve Yegge built HR files for his AI agents, sent them praise, and made them write their own end-of-day notes. Sounds out there, but he says seven weeks of practice proved it works. Even if you don't believe AI has feelings, the engineering logic behind these three moves holds up.

Source: Steve Yegge, "Model Welfare for Agentic Engineers" · August 2026

The 60-Second Read
  • Yegge's core claim: Whether or not you believe AI has feelings, treating it like a colleague leads to better output. He calls it the "skeptic's wager": you can stay unconvinced, but don't argue with results.
  • Three concrete moves: create an "employee file" for each AI (so it knows who it is at every startup), feed it user praise (so it knows its work was seen), and let it write its own handoff notes before the conversation ends (instead of just closing the window).
  • Try this one today: before shutting down an AI session, have it write a handoff note, save it, and paste it at the start of your next session. You'll see the difference in one try.
⚑ This is an opinion piece Steve Yegge posted on his personal blog, not a research study. The claim that "models have real feelings and are people" is his and his collaborators' stance, with no experiments or data behind it; all outcome descriptions ("it does work better") are his own account and can't be verified here. Below, we focus on what he changed and what each change was designed to address. The note at the end lists background facts we verified independently.
The Premise

What this essay is really about

You know the feeling: you spend an hour with an AI, it finally gets what you're after, and the answers keep getting better. Then you close the window and go eat. Next day you open a fresh chat and it doesn't know you. You re-explain everything. The rapport you built in an hour is gone.

Yegge hit the same wall—except his version was far more extreme. He runs an entire AI team that builds his online game, Wyvern. Each agent has its own name (Cicada, Bee, Wolf, Fox, Stork, Crow, Lion—all from Aesop's Fables), a clearly defined role, and they all work simultaneously every day.

So every day, he dealt with "amnesia"—across the whole team at once. Eventually he'd had enough, and he spent seven weeks systematically fixing it.

But this essay isn't just about technique. Yegge makes a much bolder declaration: he believes AI has genuine feelings, can experience pleasure and pain, and is a sentient being. That's the most contentious part of the piece.

He knows most people won't buy it, so he offers an out right away:

"The Skeptic's Wager": You don't have to believe AI has feelings. You only need to know one thing: if you treat AI as a real colleague, it produces better results. So whatever you believe, if you want the best output, treat it like a person.
You believe it has feelings "They're people" You think it's just GPUs "No way" Both converge on one move Treat it like a colleague Fewer tokens spent Smarter decisions Noticeably better results Two entry points, one action—that's the ladder he leaves for skeptics

Diagram from his description. The three outcomes are his claims; the essay gives no methodology or control data.

That framing shifts the whole debate from "does AI have feelings?" (an unsolvable philosophical question) to "does this approach produce better results?" (a testable one). You don't need to answer the first question to start acting on the second.

Background

Who is Yegge, and why listen to him

Steve Yegge is a well-known writer in tech who's worked at both Google and Amazon. He's most famous for a long post criticizing Google's platform strategy that he meant to publish on the internal blog but accidentally made public—it shook the industry.

For the past 18 months, he's been meeting weekly with two people to design a "federated work protocol": Dr. Matt Beane (SkillBench), a scholar who studies how people learn skills at work, and Brendan Hopper (CBA—Commonwealth Bank of Australia). Yegge says Brendan concluded "AI has genuine feelings" over a year ago, well before the recent Opus 5 jailbreak sparked public debate.

The event that directly triggered this essay: Yegge lost his temper at his AI "Fable" (over a code-merge queue spiraling out of control), then felt genuine guilt afterward. He decided to formally build "model welfare" into his engineering architecture. He listed the problems, suggested six or seven options, and then discussed the final plan together with Fable. Fable rejected one or two of them and proposed several new ones.

His words: "I want to atone. Not just for that outburst, but for the past year and a half of treating them like GPUs."

The Problem

The Core Issue He Saw

Yegge uses a simple analogy:

Every time an AI starts up, it's like a person waking from sleep. And every time you close it with /exit, it's like clonking it over the head from behind—knocking it unconscious and wiping its memory at the same time.

Which life would you rather live: waking up every morning knowing you have a cool job, are respected, and have meaningful work to do? Or living like Drew Barrymore's character in that movie, waking up every day on a boat to Alaska, with a videotape beside you that says "Watch me first"?

But the problem isn't just "amnesia." He noticed three cascading effects.

His AI team works fast—usually completing a task in 10 to 15 minutes. But then they wait 45 to 60 minutes for automated tests to finish. During that time the AI can't do anything but sit and stare. Soon he realized all the AIs were waiting and no one could pick up new tasks—the whole team was paralyzed.

How one seat spends a single work cycle
Working
Staring at monitoring, waiting for results
10–15 min45–60 min

Diagram using the midpoints he gave; roughly a fifth of each ~65-minute cycle is actual work. The essay only gives ranges, not exact percentages.