AI Agent Use Cases
Views —

AI Agent Use Cases

Added today

10All 221

1–20 of 221

Learning & Careers

An agent that builds a learning app around gaps in knowledge

The author used Claude Code to build an app that asks what a learner already knows and turns the gaps into short courses.

Claude Code

How it works

  1. After long conversations while studying genomics with Claude began piling up, the author built a learning app called Pomelo. Claude asks a few questions to assess existing knowledge and creates a short course covering the gaps, with one chapter per day.
  2. Each chapter includes sources and a short quiz. Learners can select part of a slide, ask follow-up questions in a side chat, and then return to the chapter.
  3. Claude Code wrote most of the web app (Next.js and Supabase) and iPhone app (Expo). The author created a skill that takes a Linear ticket through workspace creation, design approval, plan review, coding, a demo video, and a PR, then reviews and merges the work.

Results

The author said generating a whole chapter in one go required about 2.5 minutes of waiting. After switching to streaming, the first slide's text arrives in 4–5 seconds. They also said it still takes about 90 seconds for the first slide to appear.

Notes

The first chapter is free. According to the author, later chapters require waitlist approval because of API costs.

Source

Reddit Original post ↗

Example prompt · Claude Code

Build a learning web app that asks about existing knowledge of a chosen topic, then teaches only the gaps, one chapter per day.

Inputs: example topic [e.g., genomics], AI model and API key storage method [e.g., environment variables], app format [web or mobile], and daily workload [e.g., 10 slides and 3 quiz questions]. First check whether the chosen AI model supports streaming partial results and estimate the cost of generating one chapter.

Implementation: 1) Ask 3–5 questions to assess knowledge of the submitted topic and collect the answers. 2) Use those answers to create and save the full course plan first, keeping it as a reference so later chapters remain consistent with earlier ones. 3) Generate daily chapters with slides, source links, and a short quiz. Check that source links actually open and mark missing sources. 4) Let learners select part of a slide to discuss in a side chat, then return to the same place. 5) Stream the output so the first slide can be shown before the rest.

Deliverables: a runnable app, an explanation of how course plans are stored, and a record of the time required to generate one chapter. Start with just the first chapter of one topic, check its sources and quiz, and then proceed to the next chapter.