Prefer Search Based Tools (2023) (abhijeetbodas2001.github.io)

The author argues that modern software falls into two camps: directive tools that do what you ask, and recommendation-driven apps that push content at you. Because of the rise of recommendation algorithms, most apps—from social feeds and YouTube to chat clients and phone launchers—default to algorithmic suggestions that distract users and reduce productivity. The practical remedy recommended is to favor or configure search-based workflows: disable recommendation panes, set new tabs to a search engine, use clients or extensions that hide suggestions (NewPipe, YouTube search add-ons), “permanently archive” chats and rely on the search box to open conversations, or adopt minimal launchers like KISS. The goal is intentionality—only surface content when the user explicitly searches for it. For the AI/ML community this is a useful critique and design brief: recommender systems increase engagement but also attention costs and cognitive friction; shifting interfaces toward search changes the technical priorities. Better in-app search requires investments in fast indexing, semantic/query understanding, relevance ranking, privacy-preserving local search, and configurable defaults that avoid opaque nudges. Designers and ML engineers should balance recommendation-driven discovery with robust, prominent search affordances and user controls so systems enable deliberate retrieval rather than passive consumption.