🔍 Proof of Incompetence

Multiple users and publications documented GPT-5's failure at elementary algebra that any middle school student could solve.

This represents a catastrophic regression from previous models that could handle basic arithmetic correctly. The error was documented by OpenAI's own developer community and reported across multiple tech publications.

Users posted screenshots showing GPT-5 making elementary mistakes in number comparison that demonstrate fundamental logical failures.

This type of error is particularly damaging because it shows GPT-5 failing at the most basic numerical reasoning that any calculator or computer system should handle flawlessly.

When asked to create a simple US map, GPT-5 invented completely fictional state names that don't exist anywhere in reality.

This represents a significant regression from GPT-4o's geographic knowledge and demonstrates how GPT-5's training may have actually corrupted basic factual information.

Multiple users documented GPT-5 creating maps with completely nonsensical state labels, including renaming Oklahoma as the fictional "Gelahbrin."

This error spread rapidly across social media as evidence of GPT-5's regression in basic geographic knowledge, something that should be fundamental for any AI system.

OpenAI's own launch presentation contained catastrophically inaccurate charts that made GPT-5 look worse than it claimed to be.

Sam Altman was forced to acknowledge the error on Twitter, writing "wow a mega chart screwup from us earlier." The fact that OpenAI couldn't even get their own presentation charts correct became a symbol of the entire botched launch.

Users documented GPT-5 failing at basic counting tasks that any human could perform correctly.

These failures represent basic cognitive tasks that should be trivial for any AI system claiming to have "PhD-level intelligence."

Data scientist Colin Fraser documented GPT-5 failing a basic mathematical proof about repeating decimals.

While GPT-5 got the final answer correct, its reasoning and proof methodology were flawed, showing it doesn't understand the mathematical concepts it's working with.

When asked to create historical timelines, GPT-5 invented completely fictional US presidents.

This demonstrates how GPT-5's training process may have corrupted basic historical knowledge, creating confident but completely false information.