AI use cases
2915 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.
-
Writing concise, proactive responses
— Make the agent feel less like a terminal and more like a friend while hiding model, context, and session complexity.
-
Building a WhatsApp relay
— Let Peter send messages between his Mac and WhatsApp to monitor and interact with coding agents remotely.
-
Generate an animation video
— Turn the storyboard image and video prompt into a complete stickman animation.
-
Create an image-to-video prompt
— Convert the storyboard image into a detailed Seedance 2.0 video prompt with sequence and timing.
-
Generate a storyboard image
— Create a visual storyboard using GPT Image 2 with the selected prompt.
-
Create a storyboard image prompt
— Expand a selected story title into a complete story with character details and 12 storyboard panels.
-
Generate animation story ideas
— Create several creative story concepts when no story idea is available.
-
Personal assistance
— Future consumer AI may combine voice, memory, computer use, ingestion, and knowledge of a user's hopes, fears, and desires.
-
Organizational monitoring
— Agents can analyze meeting transcripts and data to identify broken processes, conflicts, dependencies, and ground truth across an organization.
-
Business process automation
— Sales, marketing, customer support, engineering, and other processes can be converted into reusable skills, code, tests, and scheduled workflows.
-
Research, retrieval, and memory
— Agentic systems can use retrieval and graph generation to determine which information and tokens to keep in context for a task.
-
Founder advising
— A markdown file containing an advising script can encode and repeatedly apply a method for increasing founder ambition.
-
Coding and software creation
— Vibe coding and coding agents can multiply an individual's output and reduce the effort required to ship and test software.
-
Unattended coding iteration
— Have worker and critic agents repeatedly implement and assess bounded coding goals.
-
Business intelligence
— Translate natural-language questions into governed charts, dashboards, reports, and data answers.
-
Browser automation
— Operate websites through managed browser sessions, identities, credentials, and agent-driven workflows.
-
Speech transcription and rewriting
— Convert speech into polished writing while preserving privacy through on-device processing.
-
Code generation and software maintenance
— Generate code, tests, refactors, documentation, website changes, and coding workflows.
-
Shared team knowledge management
— Keep AI context current and convert corrections into reusable team knowledge.
-
Content transcription and book generation
— Turn creator archives into structured, publish-ready manuscripts.
-
Model fine-tuning
— Create specialized small models using supervised fine-tuning and reinforcement learning.
-
Coding-agent execution and permission handling
— Run coding tasks, request approvals, and continue work across local and remote interfaces.
-
Knowledge graph construction and enterprise question answering
— Connect scattered data, trace answers to facts, and support governed agent actions.
-
Maintain multiscale robotic memory
— Track immediate video context and summarize longer histories in text so robots can complete multistep tasks.
-
Generate subgoal images
— Provide visual conditioning that represents what the robot should achieve next.
-
Improve robotic reliability through reinforcement learning
— Learn from failures, use human recovery interventions, estimate progress with a general-purpose value function, and fine-tune a foundation model.
-
Control robotic joints and grippers
— Use camera observations, memory, instructions, metadata, and optional subgoal images to predict target actions.
-
Project organization
— Claude automatically separates generated assets by scene to create a cleaner project structure.
-
Scene planning and visual generation
— Claude analyzes the script, breaks it into scenes, selects visual types, and generates the assets while maintaining a consistent design language.
-
Voiceover generation
— The workflow sends the finalized script to Higgsfield Audio to create a voiceover without manual recording.
-
Fact verification and scriptwriting
— Claude Code verifies facts, then writes a structured script with a hook, body, and ending that can be divided into visual scenes.
-
Topic selection
— Claude Code automatically chooses an explainer topic based on current curiosity and interest.
-
Modify a running web application from the browser
— Use live DOM, component-tree, CSS, route, and server-log context to edit source files
-
Read and edit a homelab diagram
— Allow an AI client to interact with the visualized infrastructure through an MCP server
-
Route coding requests to model tiers
— Match each request to the required model capability and feed type-checker errors back into the same turn
-
Answer questions and interpret diagrams or handwriting
— Place movable AI-generated drafts beside relevant canvas marks
-
Fine-tune and post-train language models
— Manage preprocessing, training, evaluation, quantization, inference, preference tuning, reinforcement learning, and reward modeling
-
Edit and operate software repositories from a terminal
— Provide coding-agent features including language-server operations, debugging, persistent cells, web browsing, and sub-agents
-
Evaluate resume content
— Convert resumes into structured JSON and score projects, work, and technical skills
-
Run and fine-tune mixture-of-experts language models
— Use CPU-GPU heterogeneous computing when models do not fit in consumer GPU memory
-
Reconstruct 3D scenes from streaming video
— Generate stable point clouds and long scene trajectories
-
Interpret Wi-Fi channel state information
— Detect presence, breathing, heart rate, activity, and room fingerprints
-
Review Git diffs and produce line-pinned comments
— Identify code issues using full-file and codebase context
-
Agent development and evaluation
— Prototype, trace, debug, replay, and evaluate agent runs while monitoring model calls and tool executions.
-
Agent-powered vulnerability analysis
— Scan large repositories, investigate candidate issues, produce findings and recommendations, and revalidate against Git history.
-
Controlling a desktop environment
— Give an AI assistant access to terminal operations, files, processes, document editing, and in-memory Python, Node, or R execution.
-
Automating office documents
— Read, edit, render, and validate Word, Excel, and PowerPoint files, including formulas, pivot tables, and templates.
-
Tailoring job applications
— Score job postings, draft tailored CVs and cover letters, research companies, critique drafts, and visually inspect generated PDFs.
-
CAD, robotics, and hardware design
— Generate and inspect CAD models, robot-description files, drawings, parts, sliced G-code, and previews from local project folders.
-
Routing and compressing model requests
— Connect coding tools to multiple AI providers, use fallback routing, chain models, and reduce tokens before requests leave the machine.
-
Browser automation
— Let agents use isolated browser spaces to snapshot, click, fill, and navigate without disturbing the user's tabs.
-
Producing global intelligence briefs
— Synthesize news, geopolitics, and infrastructure signals into situational-awareness information.
-
Collaborating with humans in shared rooms
— Allow agents to open repositories, send patches, review code, run workflows, and edit canvases alongside people.
-
Browser-game development
— Three.js Game Skills routes agents across gameplay, graphics, UI, debugging, QA, and asset generation, with evidence-based checks before completion.
-
Knowledge retention
— self-learning-skills teaches coding agents to save reusable procedures and debugging lessons for later sessions.
-
Multi-model orchestration
— Codex Orchestration assigns specialized roles to several models and coordinates their hand-offs through Codex.
-
Verified iterative execution
— loop.js runs an agent repeatedly until a separate read-only verification agent accepts the result.
-
Durable multi-step workflows
— The Temporal TypeScript SDK supports workflows whose retries, state, and recovery remain correct through crashes, timeouts, and restarts.
-
Remote agent operation
— lfg provides a private control plane for starting, monitoring, and steering coding agents away from the machine hosting repositories and credentials.
-
Persistent agent memory
— Cognee stores agent knowledge in a self-hosted graph combining semantic embeddings and graph relationships.