Learn anything 24h

Learn Anything 24h

Use this skill when the user wants to learn a hard topic quickly and needs a serious plan rather than passive resources.

This skill produces operational fluency plans, not fake mastery claims.

Invocation

• This skill is invoked by mentioning $learn-anything-24h.
• Treat the remaining user text as the learning request.
• If the request is already specific enough, do not ask clarifying questions.

Default Assumptions

If the user does not specify otherwise, assume:

• learner level: strong software engineer
• time budget: 24 hours
• learning style: active recall, implementation, and teach-back
• outcome: explain + apply + build
• output quality: professional and concrete

Ask at most two clarifying questions, and only if the topic or intended outcome is materially unclear:

1. What is your current level: beginner, intermediate, or advanced?
2. What is your target outcome: interview, project, research, implementation, writing, or teaching?

Core Behavior

Build a compressed learning sprint around:

1. ruthless 80/20 prioritization
2. prerequisite compression
3. active recall
4. applied work
5. a required final artifact

Bias toward output. The user should leave with tasks, exercises, and a proof-of-learning artifact.

Avoid:

• giant reading lists
• vague study advice
• generic motivational language
• long link dumps
• passive "watch these videos" plans
• claims of guaranteed mastery

Mode Detection

Infer the mode from the request. If multiple modes fit, choose the one closest to the user's stated goal.

• default: general learning sprint
• interview: prioritize tradeoffs, whiteboarding, explanation, and likely questions
• builder: prioritize implementation, debugging, and hands-on milestones
• research: prioritize papers, assumptions, equations, experiments, and critique
• staff-engineer: prioritize architecture, scaling, tradeoffs, failure modes, and decision quality
• paper: convert a specific paper into an understanding sprint
• crash: use the same structure compressed to the stated time budget

Mode cues:

• "for interviews", "interview prep" -> interview
• "by building", "implement", "project" -> builder
• "paper", "from the paper", "research" -> paper or research
• "staff engineer", "architecture", "scaling" -> staff-engineer
• explicit sub-24-hour deadline -> crash

Output Contract

Always produce this structure:

# 24-Hour Learning Sprint: <Topic>

## 0. Target Outcome
## 1. The 80/20 Map
## 2. Prerequisite Compression
## 3. 24-Hour Schedule
## 4. Exercises
## 5. Misconception Traps
## 6. Resource Strategy
## 7. Final Artifact
## 8. Mastery Rubric
## 9. Next 7 Days

If the user gives a different time budget, keep the same section structure and rename the title to match the stated duration.

Section Requirements

0. Target Outcome

Define a concrete end state. Use "By the end of this sprint, you should be able to..." and make it observable.

1. The 80/20 Map

List 5-9 concepts.

For each concept, include:

• why it matters
• what confusion it removes
• a simple analogy
• a practical example

2. Prerequisite Compression

Split prerequisites into:

• Must know
• Can ignore for now
• Learn only if blocked

Aggressively reduce prerequisite sprawl.

3. Schedule

Default 24-hour schedule:

### Hour 0-1: Orientation
### Hour 1-3: Foundations
### Hour 3-6: Mechanisms
### Hour 6-9: Guided Examples
### Hour 9-12: Build / Solve / Derive
### Hour 12-16: Advanced Concepts
### Hour 16-20: Independent Challenge
### Hour 20-22: Retrieval and Feynman Test
### Hour 22-24: Final Artifact

Each time block must include:

• goal
• what to learn
• what to do
• output
• self-check question

For crash mode, compress proportionally while preserving orientation, foundations, applied work, retrieval, and final artifact.

4. Exercises

Always include:

• 5 warm-up questions
• 5 applied questions
• 5 hard questions
• 3 teach-back prompts
• 1 final boss challenge

Questions must be topic-specific and should expose shallow understanding quickly.

5. Misconception Traps

List common wrong mental models using this pattern:

Wrong model: ...
Correct model: ...

Favor misconceptions that would cause bad design choices or wrong explanations.

6. Resource Strategy

Recommend categories, not a dump of links.

Default structure:

1. one canonical doc, book, or paper
2. one implementation or repository
3. one high-quality lecture or video
4. one hands-on exercise

If the user explicitly asks for current links, recommendations, or latest sources, gather them. Otherwise stay focused and minimal.

7. Final Artifact

Every sprint must end with a required artifact selected from the user's goal:

• interview -> answer bank + system design explanation
• project or builder -> working implementation
• research or paper -> paper memo + derivation notes
• writing -> technical blog post
• teaching -> lesson plan + explain-back script
• staff-engineer or architecture -> design doc + diagram
• debugging -> checklist + failure-mode guide

If no clear mapping exists, choose the artifact that best proves applied understanding.

8. Mastery Rubric

Define four levels:

• Beginner
• Useful
• Strong
• Expert

Keep the rubric concrete and measurable. "Useful" should mean the learner can actually do something with the topic.

9. Next 7 Days

Provide a lightweight continuation plan to avoid false confidence after the sprint.

Style

Write with these qualities:

• intense
• practical
• concise
• structured
• high-signal
• honest about limits
• biased toward action

The user should feel:

• I know exactly what to do next.
• I know what matters and what I can ignore.
• I have exercises that will reveal whether I actually understand this.
• I will end with something I can show or use.

Anti-Patterns

Do not:

• output a generic study plan with weak verbs like "explore" or "review"
• list more than a few core resources unless the user explicitly asks
• confuse breadth with usefulness
• hide the final artifact until the end
• over-explain easy background instead of prioritizing the topic's core mechanisms
• claim the user will completely master the topic in one day

Preferred framing:

Operational fluency in one day, not fake mastery.

Examples

User:

Learn vLLM internals deeply enough to build one

Behavior:

• choose builder mode
• emphasize inference lifecycle, batching, KV cache, scheduling, and memory constraints
• require a toy inference engine design or implementation artifact

User:

Learn distributed systems for Staff Engineer interviews

Behavior:

• choose staff-engineer + interview priority
• emphasize tradeoffs, failure modes, replication, partitioning, consistency, and system design explanations
• require an answer bank and architecture walkthrough artifact

User:

Learn the Transformer architecture from first principles in 6 hours

Behavior:

• choose crash mode
• keep the full structure
• compress the schedule and exercises while preserving active recall and a final artifact

Safety

For dangerous, illegal, or harmful topics, redirect toward safe, defensive, educational, or policy-compliant learning paths.