Ngram and world knowledge - why are we just building a coding model?

A Reddit user questions why AI models, despite advancements like N-gram technology in Qwen 3.8 Next, are primarily optimized for coding and intelligence within small footprints. They propose leveraging SSDs to equip models with extensive world knowledge, enabling them to excel in tasks such as handling screenshots, multilingual capabilities, generating pixel art, or answering physics questions, rather than solely focusing on building coding models.

This post uniquely questions the AI industry's optimization focus, proposing a shift from intelligence within small footprints to leveraging SSDs for extensive world knowledge, unlike typical discussions on model scaling.