Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: I'm building an AI-proof writing tool. How would you defeat it?

auth-auth.vercel.app

Post reply on HN
Print view View original

11–20 of 32 posts

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#11

Post by JoshuaDavid »

The following by "Claude Slopson" (Claude Opus asked to write an answer that was obviously AI) scored 87% authentic:

> Ah, what a fantastic question

> For me, it's Breaking Bad–and honestly? It's not just a show, it's a masterclass in storytelling that fundamentally reshaped the television landscape.

> What keeps drawing me back? The way it seamlessly blends moral complexity with edge-of-your-seat tension is nothing short of breathtaking. Walter White's transformation isn't just compelling–it's a profound meditation on identity, ambition, and the human condition itself.

> But here's the thing–it's also deeply rewatchable. Every frame is meticulously crafted. Every detail matters. The foreshadowing alone is chef's kiss!

> Whether you're a first-time viewer or a seasoned fan, Breaking Bad offers something for everyone. It's a testament to what happens when visionary creators push the boundaries of their medium.

> In an era of endless content, some shows simply transcend. This is one of them.

> 10/10, would recommend! What's YOUR comfort rewatch? Drop it below!

(HN strips the emojis, but don't worry–they were there)

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#13

Post by hackingonempty »

A paper on this was just submitted.

"We demonstrate that this class of defenses is insecure against two practical attack classes: the copy-type attack, in which a human transcribes LLM-generated text producing authentic motor signals, and timing-forgery attacks, in which automated agents sample inter-keystroke intervals from empirical human distributions."

https://arxiv.org/abs/2601.17280v1

There are tons of USB dongles available for about $10 containing a programmable microcontroller. You connect to the dongle from another device over wifi and it can act like a keyboard typing out whatever text you want using a model of human typing, perhaps even your own.

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#14

Post by callmeed »

ephou7 wrote: post #7

I used this: #!/usr/bin/python3 import subprocess import time import random with open("/tmp/x") as f: t = f.read() for c in t: subprocess.call([ "xdotool", "type", c ]) time.sleep(abs(random.gauss(0,0.07))) And pasted a random Hacker News comment: Authenticity Score 81 Highly Authentic Words per minute: 162 Keystroke variance: 52ms Paste attempts: 0 Window/tab switches: 4 Pauses (≥10s): 0 DOM manipulations: 0 You fai…

That's cool, thanks for sharing.

Is there a way to detect this approach?

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#15

Post by callmeed »

dpoloncsak wrote: post #4

>Pasting and DOM manipulation are disabled to ensure all writing is original. >We track telemetry such as typing speed, pauses, tab changes, and window focus events. People figure out ways around this for like...Runescape bots and other low-stake situations. I don't think it would hold up to anything other than casual users. Seems like an agent could whip something up in Auto-HotKey or something. I get this is the ex…

School/students were my target user when I created this. But also mostly just a fun toy.

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#16

Post by callmeed »

jryan49 wrote: post #12

I pulled up a 2nd browser next to your site. I typed the prompt into Gemini, and manually typed what Gemini said and removed em dashes as I typed. It said it was 100% genuine.

Maybe I need to punish tab/windows switches more severely?

Or perhaps require webcam and do eye tracking?

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#17

Post by jryan49 »

callmeed wrote: post #16
jryan49 wrote: post #12

I pulled up a 2nd browser next to your site. I typed the prompt into Gemini, and manually typed what Gemini said and removed em dashes as I typed. It said it was 100% genuine.

Maybe I need to punish tab/windows switches more severely? Or perhaps require webcam and do eye tracking?

I only switched tab/windows once (to input the question). So I'm not sure how severely you'd want to rate that. The webcam doing eye tracking might help if the person isn't good a typing and can copy one side without checking if the other is okay...

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#18

Post by gibybo »

I clicked begin and immediately alt tabbed to ChatGPT. I took a screenshot of the prompt and pasted to ChatGPT, with the instruction "Write a 1 paragraph response as if you were a human". I alt tabbed back to the writing website and started typing ChatGPTs response. I intentionally added a few pauses in places where I thought I might pause, and deleted words a couple times. I also slightly changed the last sentence from "down, plus the assumption that..." to "down. And.." to give it the human touch of poor grammar, although I'm not sure if that mattered.

I scored 100 "Highly Authentic"

Instructions to ChatGPT:

[What's something you pretend to understand but actually don't? Why do you fake it?] Write a 1 paragraph response as if you were a human.

Response:

Honestly, I probably nod along to conversations about highly technical stuff like advanced financial instruments or super niche software internals when I only grasp the surface, because stopping to ask for a full breakdown can derail the moment or make me look out of my depth. It's not about trying to deceive anyone, it's more about social momentum and not wanting to slow things down. And I can fill in the gaps later if I need to.

Re: Show HN: I'm building an AI-proof writing tool. How would you defeat it?

#19

Post by cyode »

callmeed wrote: post #14
ephou7 wrote: post #7

I used this: #!/usr/bin/python3 import subprocess import time import random with open("/tmp/x") as f: t = f.read() for c in t: subprocess.call([ "xdotool", "type", c ]) time.sleep(abs(random.gauss(0,0.07))) And pasted a random Hacker News comment: Authenticity Score 81 Highly Authentic Words per minute: 162 Keystroke variance: 52ms Paste attempts: 0 Window/tab switches: 4 Pauses (≥10s): 0 DOM manipulations: 0 You fai…

That's cool, thanks for sharing. Is there a way to detect this approach?

Most text composition involves backspaces and cursor movement. This script simulated neither afaik, though I’m sure it could have without much more difficultly.

Probably you will need to track many signals like those and use a model that takes them all into account.

Post reply on HN