Insert Friction

It’s the moment a student reaches for AI, before it answers, that matters most for staying in charge. This is what I built to help our kids think for themselves before AI thinks for them.

A few months ago, I wrote about a conversation I had with a 15-year-old friend of my daughter while waiting in line for a Freya Skye concert. As a reminder, she’s a sharp student who uses AI a lot. If she types “c” in her browser’s search bar, ChatGPT autocompletes, and she’s pulled to go there. She thinks she’s using AI reasonably well, but she feels guilty about it and worries that AI is slowly eroding what makes her thinking her own. I’ve thought about the implications of that conversation often.

Then, just this week, I was talking with a 30-something man who uses AI all day at work and then also reached for AI to help with a greeting card. He bemoaned that he should have been able to do that himself, but it was just too easy and automatic to use AI, and he worried about the slow slip of his writing abilities.

It’s a pattern I’m seeing over and over with kids, peers, and co-workers. AI use is genuinely helpful, but it’s becoming too frictionless.

By the time a student feels guilty about AI’s help on an essay, the decision that mattered has already passed. It happened earlier, at the blank prompt box, in the half-second before they typed “write my essay” and hit enter. That half-second is when the tool’s speed and confidence start shaping the work. It’s where a student either drives or becomes a passenger.

And right now, it’s far too easy to move from thinking about using AI to having AI do all the thinking for you.

So I built something that lives in that moment between reaching for AI and handing your thinking over to it. It’s a browser extension called f/riction.

Not all friction is bad

I’ve made the case before that not all struggle is the same. Some friction is just waste: clunky tools, busywork, rules that never helped anyone learn. AI should clear that out, and it does. But some friction is the actual work. Forming an argument, weighing evidence, getting stuck and pushing through, defending a point of view. That struggle doesn’t get in the way of learning. It is essential to learning.

We’ve moved from a world where friction was the default to one where it’s a choice. f/riction is my attempt to put a little of the right kind back, exactly where it gets lost.

What f/riction does

When a student opens ChatGPT, Claude, Gemini, Grok, or Google AI Mode, f/riction steps in first with short planning steps to help them think for themselves before AI thinks for them. It takes a minute or two and walks through four moves that line up with the AI Waypoints I’ve written about before.

Own your start. It asks what you’re working on and what you already have. A thesis, a rough take, a first attempt, the place you got stuck. AI is very good at filling empty space, so if you bring nothing, AI sets the frame and the direction. f/riction asks you to put something of your own in first.

Outsource with intent. It asks what kind of help you actually want, whether that’s brainstorming, explaining a concept, or pressure-testing a draft. You name the job you want AI to do instead of handing over the whole thing.

Keep doing hard things. This is the piece I'm most excited about. It fundamentally changes a user's experience with the AI. You don’t just ask for an answer or type of help, you also tell the AI how to work with you. You can ask AI to check your direction, flag what’s uncertain, push back, ask for your reasoning, or point out your blind spots. Those selections carry through and add friction to the AI conversation itself to sharpen a user’s thinking.

Preserve untouched space. The last step asks what you want to keep for yourself. Your argument, your voice, your final answer. You decide what’s yours before any work is handed to AI.

All of that becomes a single opening prompt that sets the guardrails for the conversation. Each step contains strong defaults that nudge a student towards framing a productive conversation with AI, but each default is also editable to maintain student agency.

The final prompt that is paste-ready into the chat box ensures AI pauses to ask for what the student has before starting to answer, shapes the AI’s interaction to provide the type of help and challenges the student asked for, and steers the AI away from doing the work the student said they want to keep for themselves.

As a small side benefit, f/riction also gives students the option to start with sources instead of the AI overview when they run a Google search. Everything is moving in the direction of AI’s synthesized answer being the default response, limiting student exploration and scrutiny of sources. f/riction pushes back on this trend by taking the student to the source list first, giving them a chance to take a different path instead of defaulting to reviewing AI’s answer.

A few things f/riction is not. It’s not an AI detector or content moderator. It’s not a filter. It doesn’t monitor anyone, send chats to a server, or report behavior to a teacher or parent. It doesn’t force the student to use it.

Policing starts from suspicion. f/riction starts from agency. It assumes what I keep coming back to: our kids want to learn. Where they need our help is in preparing to use the most powerful cognitive technology ever invented without losing the essential development moments that will shape their future in the process.

How to get f/riction to your kids

I’ll be open about where this is. It’s early. It’s a free Chrome browser extension. It doesn’t collect personal data or send student work anywhere. I built it because I think it can bring real value to students. My kids (and wife) actually like it.

The extension is available through a private Chrome Web Store link, which means I’m handing out the download link individually. I want to know who’s using it so I can have real conversations to make it better (immense thanks to those who tested v1).

If any of this resonates, message me.

If you’re a parent and you want to try it alongside your kid, send me a note and I’ll send you the link. Then tell me what you think. What helped, what felt clunky, what your kid said about it.

If you’re a teacher, a school leader, or a librarian, I especially want to hear from you. Librarians in particular are stepping into an important role as schools work out what AI literacy/fluency looks like. I’d love to find a few schools willing to try f/riction with students this fall so we can watch what actually changes and write about the results (in a way that fully protects student privacy). Grades 7-10 are probably ideal, but I’m open to discussing others.

At the upcoming education innovation conference in Colorado (InnEdCo), I’m hosting a breakout session focused on how parents and teachers can work together to help students develop future-ready skills while protecting their development. Tools like f/riction are just a small part of that much bigger, critical conversation. We won’t protect learning by pretending kids will avoid AI. We protect it by helping them use AI without losing themselves in it. That takes fluency, it takes friction, and it takes all of us being part of the solution whether we feel ready or not.

— Whit