The 'AI Confidence Theater' problem: everyone claims life-changing workflows, few can show them

Elena Verna, who works at an AI company and uses the tools all day, argues that the loudest voices in AI have created a performance she calls ‘confidence theater’ — a steady stream of ‘life-changing’ claims that collapse the moment you ask people to actually demonstrate them. When she does ask, what she usually sees are modest, genuinely useful workflows: summarizing Slack, drafting emails, running scheduled research. Helpful, but nowhere near the employee-replacing super-agents being advertised, most of which fire inconsistently and produce good output only when hand-fed precise context.

Her core point is that the exaggeration isn’t harmless. It poisons real adoption — people who try a hyped ‘system,’ watch it fail, and conclude all AI is nonsense. It has also revived hustle-culture posturing in a new form: instead of bragging about 5am workouts and revenue milestones, people now flex how many tokens they burn and how many agents ‘run their business,’ with no business outcomes to back it up. Hiring is a particular casualty, because AI hands everyone the vocabulary of expertise (RAG, MCP, vector databases, memory) without the competence behind it, pushing employers toward work trials and case studies since verbal interviews no longer filter anything.

Verna traces the phenomenon to three forces: social media rewards shocking, exaggerated claims with attention; AI genuinely can do impressive things, so fabricated demos are hard to distinguish from real ones (unlike, say, faking an ‘insane Dropbox setup’); and vendor marketing sells certainty about outcomes that are inherently probabilistic. The net effect is a fake baseline that makes legitimately time-saving uses feel embarrassingly basic and leaves people demoralized rather than excited — which is why she wants the theater to stop.