Daily Practice

Daily Practice is the review queue on the dashboard.

It is the part of the product that makes the platform different from a normal article site. Reading creates exposure. Daily Practice creates repetition.

What Appears in Daily Practice

The dashboard should pick a small number of review items each day. The current design targets one item per key category, so the learner does not only review the most recent topic.

Examples:

• one Transformer Systems card
• one AI Agent Frameworks card
• one card from a weak concept
• one card from a recently completed lesson

The exact scheduling can evolve, but the intent is stable: make review light enough to do daily and broad enough to prevent forgetting.

Review Items

A review item can be a flashcard, a small question, or a challenge prompt.

Good review items are short and specific:

• “What is stored in the KV cache?”
• “Why does a causal mask prevent information leakage?”
• “What is the difference between an action and an observation in an agent framework?”
• “Where does the runtime boundary sit in an OpenHands-style architecture?”

Bad review items are too broad:

• “Explain transformers.”
• “Explain AI agents.”
• “Summarize the whole article.”

Recall Score

Each roadmap can have a recall score. The score should represent how reliably the learner remembers the cards attached to that roadmap.

A simple first version can track:

• attempted cards
• correct answers
• missed cards
• cards due for review
• solved challenge decks

Later versions can use a spaced repetition scheduler, but the product does not need a perfect algorithm to be useful. It needs a consistent review loop.

Why This Matters

The topics in this platform are dense. A learner can read about attention masks, KV cache, FlashAttention, or agent event streams and still lose the structure a few days later.

Daily Practice makes the platform ask: can you still reconstruct the idea now?