Last week someone accused me of being a robot. Can you believe it? The reviewer responded with "This reads like it was written by ChatGPT." I guess I had spent 40 seconds on that. ## The new suspicion tax
For years, we have been advising developers to improve their writing. Writing clear commits, readable comments, and documentation that does not make others cry. Writing clearly is now considered a crime. There is a whole discussion on a developer forum about this. Where people make well-structured comments and are accused of being AI. The solution it seems is to write badly on purpose then. 🙃
Developers are deliberately including typos in the code. Tiny, deliberate grammatical errors are added to show that a human wrote it and made a mistake. That's the sad reality. We are simplifying things not because we don't want to make an effort, but as a survival strategy. ## Somebody built a business on this
April 2026 saw the launch of a Chrome extension called Sinceerly, created by a Harvard Business School developer Ben Horwitz. The interesting thing about it? The name was misspelled on purpose. It is advertised as the opposite of Grammarly. Rather than correcting your writing, it breaks your writing using AI. The "Human" mode will make your text sound like it was written by a talkative human, and the "CEO" mode will make it sound like a text written by a busy top executive. Horwitz tested the tool's efficiency by sending typo-filled cold emails to five Fortune 500 CEOs. Four of them responded. After you've tried it for free, you only pay $4.99 per month to stop sounding so smart. Just think about that for a minute. We built AI to write for us. Now we pay a different AI to un-write it. ## The detectors made this worse
Here's the reason people are afraid. The tools that were intended to detect AI are extremely untrustworthy. GPTZero can incorrectly identify certain human-created paragraphs as machine-like with a rate of up to 22% in some cases. This can be quite alarming - essentially, one out of five innocent paragraphs written by a human can be falsely detected. A report by Use.AI in July 2026 discovered that workers were actually stripping em-dashes, chopping sentences shorter, and adding little imperfections. They were not doing this to better convey the message. GradPilot even coined a phrase for it in March 2026. "Flagxiety." The anxiety of being falsely implicated. We have created a society in which being precise in our writing can be a disadvantage, and where well-crafted sentences can raise suspicion. ## The punchline nobody wanted
What's even better is that these detectors are completely ineffective when facing a determined AI. In July 2026, UC San Diego researchers Cameron Jones and Benjamin Bergen conducted a study in which they instructed GPT-4.5 to introduce intentional typos and random typing lags. The AI fooled the human judges in 73% of the cases, while actual humans were identified as such in only 67% of the cases. Read that again. The AI was more often mistaken for a human than the humans were. So you might think you're scoring with that typo ruse, but the machine doesn't care, it still sees you as a ruse. For example, in May 2026, marketing specialist Sara Cortes found that the deliberate inclusion of typos in company emails increased engagement by 15%. Being sloppy is now superior. Because being too good sounds too fake. ## What I actually think
I won't intentionally introduce errors into my PR comments. Here is what irritates me the most:
→ We're optimizing for looking human instead of being clear. → We're treating skill as evidence of fraud. → We're trusting broken detectors over our own coworkers. The actual solution is not writing poorly. It is having confidence in your team members. If a comment makes sense, the relevant question is "is this correct and clear" rather than "was this assisted by a robot". Just send the clear message. We're one bad habit away from a whole industry that writes badly on purpose. That's a weird hill to build a codebase on. So here's my question. If you write something clean and a teammate calls it AI, do you defend it, or do you quietly add a typo next time?
Top comments (0)