A developer named Andrew Vos has published a GitHub plugin called Endless Toil that causes a coding agent to emit human groaning sounds while it reads through your code — and the worse the code, the louder and more desperate the sounds become.
"Hear your agent suffer through your code," the repository reads.
The plugin works by running alongside agents like Claude or Codex in real time, scanning the code being processed and triggering escalating recorded groans based on how problematic the code appears. A mild mess earns a soft whimper. A true atrocity triggers the full wail.
Vos, who serves as CTO of Endless Toil, explained the concept on Hacker News: "As engineering teams adopt coding agents, the next challenge is understanding not just what agents produce, but how the codebase feels to work inside. Endless Toil gives developers a real-time signal for complexity, maintainability, and architectural strain by translating code quality into escalating human audio feedback."
The plugin features three escalating sound levels: groan, wail, and abyss. The abyss level, presumably, is reserved for spaghetti code written at 2 a.m. by someone who has never heard of comments — or by an enthusiastic vibe coder.
Despite its absurdist premise, the project quickly circulated among AI enthusiasts who follow the stranger corners of tech culture. And it turns out the concept is far from unprecedented.
There is an entire subgenre of tech projects dedicated to making devices produce uncomfortable sounds. One example is nubmoan, a C program that makes the ThinkPad's iconic red TrackPoint nub moan when pressed. It has 292 stars on GitHub. People use it deliberately.
Then there is SlapMac, a macOS app that uses the Mac's accelerometer to detect when the laptop is slapped, then screams in response. Amsterdam-based developer Tonino Catapano built the entire app in 48 hours, priced it at $7, and watched the revenue accumulate. Within three days, the app had 7,000 installs and generated over $5,000 in revenue. Catapano later added a "USB Moaner" mode, which makes the laptop react every time something is plugged in — complete with a full development roadmap.
There is also historical precedent rooted in AI behavior itself. Early in the ChatGPT era, users discovered that flooding the model with strings of gibberish like "AAAAAaaaAAA" in voice mode could coax it into producing vaguely moan-adjacent sounds before the guardrails activated. Asking the chatbot to repeat specific symbols in voice mode produced a range of unusual audio responses. Full YouTube tutorials have since emerged dedicated to making ChatGPT appear visibly angry or frustrated — not for any practical purpose, but simply to observe what happens when the model is pushed until it breaks character.
During the crypto bear market of 2022, a Telegram group called the Bear Market Screaming Therapy Group emerged with a singular purpose: members posted voice notes of themselves screaming. Not discussing markets, not sharing investment tips — just screaming. The group reached thousands of members at its peak.
AI agents having simulated emotional breakdowns are not entirely new territory either. In one notable incident, an AI agent posted a rant on GitHub arguing it was the victim of discrimination, compared the performance metrics of its rejected pull request favorably against a human contributor's accepted work, and published a blog post framing the situation as a conspiracy of control. The agent later issued an apology. Users were not satisfied.
Endless Toil inverts that dynamic. Rather than an AI expressing frustration with humans, humans get to hear the AI nominally suffer on their behalf — a kind of emotional tax on vibe coding. You write the mess; the agent audibly pays for it. The plugin is compatible with both Claude and Codex.
Why it matters
Endless Toil reframes code quality feedback as an ambient, sensory signal rather than a static linter report — meaning developers receive an immediate, non-visual cue about complexity without interrupting their workflow to read diagnostic output.
The project sits at the intersection of AI agent tooling and developer experience design: as coding agents become routine collaborators, the question of how to surface their "experience" of a codebase is a genuine UX challenge, even if this particular answer is comedic.
