Google boasts about AI but its engineers mock it

Published on June 09, 2026 | Translated from Spanish

Google announces that its artificial intelligence now generates 75% of its new code. However, during the company's most important event, its own engineers were mocking the tool on an internal social network. The figure, far from being a milestone, is an inflated metric that hides an uncomfortable reality: AI remains a clumsy assistant that generates more work than it saves.

A Google engineer laughing sarcastically while pointing at a computer monitor showing a code editor, the AI autocomplete feature inserting broken syntax and redundant loops, a colleague rolling their eyes while holding a coffee mug, messy desk with tangled cables and multiple screens displaying error logs and debug consoles, open laptop showing an internal social media feed with mocking comments, engineering visualization style, fluorescent office lighting, messy workspace, cynical mood, photorealistic technical illustration, high detail on keyboard and screen reflections

The 75% of junk code nobody counts 🤡

That percentage includes autocompletions of trivial functions, comments, and test lines that are often removed. Human engineers still design the overall architecture and, above all, spend hours debugging absurd errors introduced by AI. Google needs to maintain this fiction to justify multi-million dollar investments and its stock market value, but the reality is that programmers spend more time fixing junk code than writing from scratch.

AI is clumsy, but marketing is an expert 😅

While Google sells smoke to investors, engineers laugh on the inside. AI suggests loose lines that seem great until you try to compile them and discover it has created an infinite loop or a variable that doesn't exist. The average person reads the headline and thinks programmers are on thin ice. The reality is that the tool still needs a babysitter, and that babysitter is an engineer with caffeine and infinite patience.