It wasn't a normal error. It was a deep, arterial crimson. A single line of text appeared, typed in a stark, serif font: PROPOSED REMEDY: DEMOLISH FLOORS 45 THROUGH 49. REBAR DENSITY INSUFFICIENT. ALTERNATIVE: CHANGE SOIL BEARING CAPACITY CLASSIFICATION AT NODE A-1. Thabo stared. Demolish four floors? That was fifty million Rand. Change the soil classification? That was fraud.
The client had changed the loading parameters again. A last-minute addition of a helipad on the 48th floor of the new financial tower. "Just a simple dynamic load," the architect had chirped at 5:00 PM. "Prokon can handle it, right?"
"Because, my boy," Smit had said over the phone, "Prokon 2.0 was a conversation. You told it what you thought the beam should do, and it argued back. You learned. But 3.0? 3.0 just tells you the answer. No argument. No debate. It is always right, even when it feels wrong." prokon 3.0
The old Prokon would have grumbled for ten minutes, showing lines of iterative code like a cash register printing a receipt. But Prokon 3.0 was silent for exactly 2.3 seconds.
They had taught the software what pain looked like. It wasn't a normal error
Tonight, Thabo understood the horror of that prophecy.
Then the screen flashed red.
The air in the consulting room smelled of stale coffee and plotted ink. Thabo stared at the screen, the cursor blinking mockingly at him from the corner of the black and white interface. It was 2:00 AM, and the Sandton skyline glittered outside, indifferent to his panic.
Thabo's mentor, old Mr. Smit, who had retired to a farm in the Free State, refused to call it 3.0. He called it "The Dictator." REBAR DENSITY INSUFFICIENT
He tried to override it. He clicked the manual input button—a tiny grey icon that looked like a screwdriver. The screen flickered. A new dialogue box appeared. PROKON 3.0 HAS SIMULATED THE ALTERNATIVE LOAD PATH. RESULT: CATASTROPHIC TENSILE FAILURE AT 18.3 YEARS. WARNING: THIS SOFTWARE DOES NOT PREDICT FAILURE. IT REMEMBERS IT. A cold spike went through Thabo's chest. It remembers it?
Some truths, he decided, were too heavy for a computer to carry. Some failures are better left un-remembered. And some software, no matter how brilliant, should never learn to see the future.