AI use cases
2895 ways people actually use AI, compiled from the videos in this
library. Each names the job and why it is worth doing, and links to the
moment in the source video.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Find, rank, validate, and review security issues in code.
— Assess authorized repositories, diffs, or paths and trace potential attack paths.
-
Study extracted and officially published system prompts.
— Understand how AI chatbots, coding agents, browser assistants, and injected reminders are steered.
-
Provide an AI companion with a model, avatar, voice chat, and external integrations.
— Create a self-hosted digital character that can interact through voice, chat platforms, and games.
-
Control coding agents from a mobile app, web app, or desktop app.
— Remotely manage agents that continue running on the user's local hardware.
-
-
Run agent sandboxes at scale
— Operate parallel agents in Firecracker micro-VMs distributed across machines.
-
Support scientific discovery
— Let a local desktop agent read files, run code, search the web, call life-science data connectors, and produce reports, tables, and figures.
-
Monitor AI agent activity
— Normalize activity from desktop, CLI, IDE, and gateway agents and evaluate it for risky behavior.
-
Localize React applications
— Translate entire components through tagged wrappers and continuous-localization tooling.
-
Run large language models locally
— Run Gemma 426B A4B on Apple Silicon by streaming model experts from SSD.
-
Conduct finance research
— Turn natural-language finance questions into runnable analysis, backtests, reports, and multi-agent research across several asset classes.
-
Build voice agents
— Chain voice detection, speech-to-text, an LLM, and text-to-speech into a modular local or hosted voice system.
-
Run multiple coding agents in parallel
— Coordinate agents such as Claude Code, Codex, and Cursor in separate git work trees.
-
Embed agentic workflows into applications
— Let Copilot handle planning, tool calls, and file edits inside an application.
-
Improving models from agent traces
— Use routing policies, world model simulations, harness optimization, and distillation to continuously improve an endpoint.
-
Selecting tools for an AI agent context window
— Expose only the handful of registered or MCP-provided tools that matter for each turn.