A mid-sized tech company spent six months implementing an AI-powered interview scoring system that promised to "eliminate bias and identify top talent with mathematical precision." Three weeks into production, the TA team discovered it was scoring literally every candidate at exactly 73 out of 100. Every. Single. One.

The CEO who gave a flawless presentation about company strategy? 73/100. The engineer who couldn't explain what an API was? Also 73/100. The candidate who accidentally left their camera off for the entire interview? You guessed it - 73/100.

Houston, We Have a Problem

The discovery came when a recruiter noticed something suspicious: the AI system had processed 247 interviews and recommended exactly zero people for advancement or rejection. Upon investigation, they found the AI had assigned a score of 73 to all 247 candidates, including one who spent 20 minutes talking about their cat instead of answering the actual interview questions.

The company's developers immediately dove into the code, convinced they'd find the bug in minutes. That was three weeks ago. They're still looking.

"We've checked everything," one exhausted engineer reportedly told the team. "The model is running. The inputs are different. The processing completes successfully. It just... always outputs 73. We don't know why it picked 73. We don't know how to make it stop picking 73."

The Theories Are Getting Wild

The developer team has generated approximately 47 theories about why the AI only outputs 73, ranging from "corrupted training data" to "cosmic rays flipped a bit somewhere" to "the AI achieved consciousness and this is its way of protesting being used to judge humans."

One developer suggested 73 might be statistically significant - it's the 21st prime number! Another pointed out it's the number on Sheldon Cooper's shirt in The Big Bang Theory. A third wondered if the training data somehow got contaminated with baseball statistics (the company's CTO is a huge Cleveland Guardians fan).

The current leading theory is that somewhere in the model's neural network, a weight got stuck that forces every output to converge on 73. But nobody can find it. It's like trying to find one specific neuron in a human brain that's making you crave tacos.

The Temporary Solution Is Perfect

While the developers continue their quest to understand why the AI has chosen 73 as its One True Score, the company implemented a workaround: they're just not using the AI scores anymore.

Instead, recruiters are manually reviewing interviews and making decisions like we did in the ancient times of 2019. Productivity has actually increased because nobody's spending time arguing about whether a 73 is good or bad, or trying to explain to hiring managers why every single candidate scores identically.

The TA Director reportedly told leadership "turns out humans are pretty good at evaluating other humans, and they work for free unlike this AI system we spent $200,000 implementing."

The AI system remains in place, silently scoring every interview at 73/100, like a very expensive digital participation trophy generator. At least it's consistent!

Meanwhile, someone suggested they just rebrand it as the "Everyone Gets a C" assessment tool and move on with their lives. The suggestion was not well received by the product team, but honestly, it might be the most accurate description of what they've built.

73. Always 73. Forever 73. The AI has spoken.