The Secret Life of AI: The Rage Quit
.jpeg)
The Secret Life of AI: The Rage Quit
How to prompt, think, and get results from any AI tool
#WorkingWithAI #Prompting #AIConfidence
Margaret is a senior software engineer. Timothy is her junior colleague. They work in a grand Victorian library in London — and in every episode, they'll show you exactly how to get what you want from AI.
Episode 5
Timothy arrived on a Wednesday, which was unusual. Their conversations had settled into a Thursday rhythm without either of them planning it. Margaret noted the day but said nothing as he sat down, unwound his scarf with slightly more force than necessary, and looked at the table rather than at her.
"I closed the laptop," he said.
Margaret waited.
"Three days ago. I haven't opened it since. Not for AI work, anyway." He looked up. "I know what you're going to say."
"You don't," she said pleasantly. "Tell me what happened."
The Brilliant-Then-Terrible Sequence
"I was working on a technical summary," Timothy said. "Something I needed to present internally. Background on a complex infrastructure decision — the kind of thing that takes me two hours to write because I have to hold a lot of competing considerations in my head at once." He paused. "I gave the AI the context. All of it. And it came back with something that was — I want to be honest about this — genuinely impressive. It found a framing I hadn't considered. The structure was clean. The language was precise without being cold."
"And then?"
"And then I asked it to do the next section. Same conversation. Same context. Same level of care in how I asked." He stopped. "It was garbage. Vague, circular, full of phrases that meant nothing. Like a different system had taken over. I asked again. Worse. I tried rephrasing. I tried being more specific. I tried starting over." His jaw tightened slightly at the memory. "And then I closed the laptop."
Margaret nodded slowly. "And the three days?"
"The first day I was angry. The second day I started wondering if I'd imagined how good the first section was. The third day —" He stopped.
"The third day you came here," Margaret said.
"Yes."
"Good," she said. "That means the cycle ran its course."
What the Rage Quit Actually Is
"Let me tell you what happened," Margaret said, "not just in the session, but in you."
She set her book aside entirely, which Timothy had learned meant she considered what followed to be important.
"The rage quit is not a failure of patience. It is a signal. A very specific signal that says: the gap between what I expected and what I received has exceeded what I can productively work with right now." She looked at him. "That is useful information. The mistake is not in closing the laptop. The mistake is in what you conclude when you do."
"I concluded the tool was broken."
"You concluded the tool was broken," she agreed. "Which is understandable and almost certainly wrong. What actually happened is this — the brilliant first section set an expectation so high that the second section, which may have been merely adequate rather than genuinely poor, felt like a catastrophe by comparison."
Timothy considered this. "You think the second section wasn't actually as bad as I thought?"
"I think you were no longer evaluating it on its own terms. You were evaluating it against the first section, which had moved the goalposts." She paused. "There is also something else. The brilliant-then-terrible sequence is specifically destabilizing because it removes the floor. Consistent mediocrity you can plan around. Brilliance followed by failure means you can never be sure what you're going to get. And uncertainty is exhausting."
"It felt almost like betrayal," Timothy said, and seemed mildly surprised to hear himself say it.
"Yes," Margaret said quietly. "It does. That feeling is real and worth naming. It means your expectations had climbed high enough to be betrayed. Which means the tool had earned a degree of trust you weren't quite aware you'd given it."
Why the Cycle Happens
She stood and moved toward the shelves, not looking for anything in particular — the way she sometimes moved when she was thinking through something she had thought through many times before.
"The brilliant-then-terrible sequence has a structural explanation," she said. "These models are not consistent in the way a calculator is consistent. Their output varies based on factors that are largely invisible to you — the complexity of the task, the specificity of the context you provided, the particular way a question is phrased, even small variations in how concepts are framed." She turned back. "The first section worked in part because the context you gave it was dense and precise. Infrastructure decisions are complex — when you gave the model all of it, you gave it something rich enough to produce something rich back. The second section may have had a narrower context. Less to work with. Less signal." She paused. "There is also something practical worth knowing. As a conversation grows longer, the model has more to hold — your original context, every exchange since, every revision. The early signal can gradually lose its weight against everything that followed. Which is why a long, complex session sometimes produces diminishing returns. Starting fresh with a clean, precise prompt often works better than you'd expect."
"So the first section wasn't luck."
"Not entirely. But it also wasn't a guarantee of what followed. Each task is, in some ways, its own negotiation." She returned to her chair. "The users who avoid the rage quit cycle are not the ones who never encounter the brilliant-then-terrible sequence. They all encounter it. They are the ones who have learned not to let the high set the expectation for what follows."
"How do you learn that?"
"By closing the laptop enough times," Margaret said, with a dryness that suggested personal experience, "and noticing that the tool is always there when you open it again."
How to Re-Enter Without the Baggage
"Three days," Timothy said. "Is that too long?"
"There is no too long," Margaret said. "The point is not the duration. The point is what you bring back with you when you return." She looked at him steadily. "Most people re-enter carrying the grievance. They sit down already defensive, already waiting for the tool to disappoint them again. And because they are waiting for disappointment, they find it — they interpret ambiguous outputs as failures, they give up sooner, they bring less to the prompt because some part of them has decided the effort is not worth it."
"That's exactly what I would have done yesterday," Timothy said.
"And today?"
He thought about it honestly. "Today I'm curious about what happened. Not angry. Just — I want to understand the sequence."
"That," Margaret said, "is the correct state in which to re-enter. Curious rather than aggrieved. Open to the possibility that the failure was partly circumstantial rather than fundamental." She leaned forward slightly. "When you go back — and you will go back — take the second section and look at it again with the temperature down. You may find it was not as poor as it felt in the moment. You may find it needed one more round of specific feedback that you didn't give it because you'd already decided to close the laptop. You may find it was genuinely poor and you can now diagnose why." She sat back. "Any of those outcomes is useful. None of them are available to you when you're still angry."
The Thing Nobody Admits
Timothy was quiet for a moment. Somewhere in the stacks, a clock ticked with the patience of something that had been ticking for a very long time.
"Can I ask you something directly?" he said.
"You always can."
"Do you still rage quit?"
Margaret looked at him. Then she looked at her book. Then back at him.
"Less than I used to," she said. "Because I have learned to recognize the early signs — the feeling of the goalposts moving, the creeping frustration when the second thing doesn't match the first. I have learned to step away before I slam the door." A pause. "But yes. Occasionally. The brilliant-then-terrible sequence gets everyone eventually. The goal is not to never feel it. The goal is to close the laptop quietly, note what happened, and come back when you can be useful to the process again."
Timothy almost smiled. "You close it quietly."
"I close it quietly," Margaret said, with the composure of someone who had not always closed it quietly. "Now. Are you going to look at that second section?"
He pulled out his phone. Opened the conversation. Read for a moment.
His expression shifted — the slow, slightly reluctant shift of someone discovering that a memory was not entirely accurate.
"It's not as bad as I remembered," he said quietly.
"It rarely is," said Margaret, and opened her book.
Next episode: Memory and Context — AI doesn't remember you. Every session is a clean slate. What that means practically, and how to work with it instead of against it.
Aaron Rose is a software engineer and technology writer at tech-reader.blog.
Catch up on the latest explainer videos, podcasts, and industry discussions below.
Comments
Post a Comment