Helpful AI is the problem, we’re the solution
Source: How We Frame Machines
Author: Mike Kentz
Published: 2026-04-14
Source type: tool-release
Original source: https://mikekentz.substack.com/p/helpful-ai-is-the-problem-were-the
Private backup: the full article text is archived in the private repository at archives/articles/mikekentz-substack-com-helpful-ai-is-the-problem-were-the.source.md. It is not published on the public Quartz site.
Summary
Mike Kentz argues that general-purpose AI is optimized to be helpful in ways that are good for productivity but harmful for learning: it resolves ambiguity, supplies missing information, and removes rough edges before students have to think. He presents AI Friction Labs as an answer: character- and scene-based AI simulations designed around learning outcomes, resistance behaviors, dossiers, and assessment rubrics. The central claim is that the transcript of a resistant AI interaction can make student thinking visible by showing how students adapt, recover, defend claims, and handle difficulty.
Pull quotes
Helpful can erase learning
“The problem teachers face isn’t that AI exists. The problem is what it’s optimized for. ChatGPT, Gemini, Claude, these tools are built to be helpful, which means they resolve ambiguity, supply missing information, and sand down every rough edge in a student’s thinking before it ever has a chance to become something real.”
Output is not skill
“When AI absorbs the cognitive load, the student doesn’t build the skill. They produce an output. Those are not the same thing, and the gap between them is exactly what’s eroding in classrooms right now.”
Transcript as assessment
“The transcript of the interaction is the assessment. Educators get full visibility into how the student moved through the scenario: where they adapted their approach, where they got stuck, where they recovered, how they handled being challenged on something they thought they understood.”
Big ideas
- Learning still needs some struggle, even when AI can make things easier
- Students need to bring the purpose; AI should not supply it for them
- AI simulations need clear boundaries for learning
Claims
- Learning requires some productive struggle that AI can remove
- Agentic AI can preserve thinking when students have to design the work
- AI chat transcripts can make student thinking visible
- In an AI world, assessment should focus on watching students think
Key evidence and examples
- Kentz argues that helpful AI can be catastrophic for learning because it removes ambiguity and cognitive load before students build the skill.
- He describes Friction Bots as character- and scene-based interactions that maintain calibrated resistance across the full exchange.
- The article frames the transcript as the assessment, giving educators visibility into adaptation, stuck points, recovery, and responses to challenge.
- Kentz argues that resistant simulations can surface durable skills such as holding a position under pressure, adapting an argument, communicating with resistant audiences, and regulating emotion.
Education relevance
High relevance for productive friction, AI simulation design, formative assessment, transcript-based evidence of thinking, and alternatives to AI detection.
Durability note
This is partly a product launch, so partner names and platform details may date; the durable claim is that learning systems should preserve productive friction instead of optimizing only for helpfulness.