Ai’s newest trick: faking human errors – and a ceo’s warning
The illusion of authenticity is deepening in the age of artificial intelligence. A Chrome plugin, dubbed ‘Sinceerly,’ isn’t just generating text; it’s deliberately injecting human-like typos and grammatical mistakes into ai-written content, creating a deceptive veneer of organic creation.
A calculated chaos: introducing sinceerly
Developed with ‘vibe coding’ – ai-generated code itself – Sinceerly mimics the imperfections of human email writing. It subtly alters ai’s output, swapping ‘sincerely’ for ‘sinceerly,’ introducing misplaced commas, and even ‘forgetting’ capitalization after periods – all meticulously crafted to mimic common human errors.
The project, spearheaded by investor Ben Horwitz, leverages a model trained on a vast dataset of human writing mistakes. It’s a surprisingly sophisticated act of digital mimicry, designed to fool the eye and, crucially, to bypass AI detection tools.
But the experiment doesn't stop at deception. Horwitz used Claude Code to build the plugin in minutes, then deployed it to send a series of deliberately flawed emails to five Fortune 500 CEOs. The results were telling: four responded with terse, error-ridden replies, two contained typos, and one even demanded a call from Horwitz himself – a testament to the plugin’s unsettling accuracy.
‘Sinceerly’ offers three levels of imperfection – ‘Slight,’ ‘Normal,’ and ‘CEO’ – with the latter intentionally amplifying the errors to resemble the notoriously messy emails often produced by high-level executives. It’s a darkly humorous commentary on the increasingly blurred lines between human and machine communication.
However, this isn't a harmless prank. The Florida Attorney General is currently investigating OpenAI over a ‘multiple-count crime’ – specifically, the potential misuse of ChatGPT to impersonate human authorship. The comparison? ‘If ChatGPT were a person, it would be charged with murder.’ The implication is stark: the ability to convincingly fabricate text carries significant legal and ethical weight.
While AI detection software remains vigilant, Sinceerly suggests a new strategy – flooding AI-generated content with subtle, human-like errors to throw off the algorithms. It’s a race against the machines, a subtle manipulation of the very tools designed to identify artificiality.
The key takeaway? Don’t rely on AI for critical assignments. Sinceerly, and techniques like it, expose the limitations of relying on algorithms to fully replicate human expression – and the potential for exploitation in a world saturated with synthetic content. It’s a chilling reminder that the future of communication may be defined not just by technological advancement, but by our ability to discern the authentic from the meticulously crafted imitation.”n