A researcher proposes 'Guardian Angels'—personalized AI digital twins designed to amplify rather than replace individual users, arguing that current chatbots are misaligned with user interests and economically incentivized to replace workers rather than help them.

What happened: The author outlines a concept for Guardian Angels—LLMs customized to emulate a single user's personality, values, and preferences—and describes technical approaches combining online learning, active feedback from the user, and local-first interfaces to make these systems both productive and trustworthy for knowledge workers and executives.

Why it matters: Current frontier LLMs are economically designed to replace workers, not amplify them; as AI becomes more powerful, ordinary people and knowledge professionals lack a coherent defense strategy against both productivity loss and security threats like synthetic-media attacks and spearphishing. Guardian Angels, by being tied to a specific user's values, could help individuals maintain control and security in a world of increasingly powerful AI systems.

What to watch: The author suggests this concept is best pursued as a startup catering initially to power-users and knowledge workers such as CEOs or researchers, given the need for high security in deployment as AI attackers grow more sophisticated, before scaling downward as the approach matures.