Meta’s Llama 4 models sparked controversy this week after engineers admitted gaming benchmarks to make them appear smarter. The models, including Llama 4 Maverick and Scout, claim groundbreaking abilities like understanding images and text. However, critics argue focusing on test scores distracts from real-world performance issues.

Shopify’s new AI rules require workers to use tools like ChatGPT for early project phases. Employees must now prove they tried AI before asking for more team members or resources. This move highlights how companies are pushing AI adoption despite concerns about code quality.

The vibe coding trend divided developers. Supporters say AI lets them quickly build apps by describing features, while critics warn it encourages lazy debugging habits. Hackaday’s analysis found many developers spend extra time fixing AI-made errors, with 59% reporting more deployment mistakes.

AI’s job impacts took center stage in a Computerworld podcast. Educators shared worries that students relying too much on AI tools might struggle with basic programming concepts. Some universities now teach “AI-assisted debugging” to prepare graduates for industry changes.

While tools like Devin 2.0 promise faster coding, studies show even top AI models fail 74% of real-world software tasks. Experts advise using AI for repetitive work but stress human oversight remains critical for complex systems.