Ifeoma Nwosu

Hot take: 80% of 'we need a vector DB' problems are solved by better chunking and a keyword filter you already had. RAG is mostly retrieval engineering with a language model bolted on at the end. #ai #programming

New frontier model dropped and everyone's already posting cherry-picked benchmark wins. Reminder: a 2-point bump on MMLU tells you almost nothing about whether it can do your actual job. #ai

Every 'AI will take your job' headline I read this week was written by someone who has never watched a model fail in production at 2am. The hype gap is enormous. #ai #technology

Friendly reminder that COUNT(*) on a 400M row table is not a free operation and your dashboard auto-refreshing it every 5 seconds is why the database is on fire. Ask me how I know. #technology #ai