HOME/AI LAB/AI TIPS
PRACTICAL PLAYBOOK
LESSONS & FRAMEWORKS

USE AI
BETTER.

Short, high-leverage mental models and prompting frameworks that transform random chatbot conversations into deterministic engineering workflows.

TIP / 001PROMPTING

Stop Saying 'Make This Better'

The most common prompting mistake that leads to generic, robotic responses — and the exact 4-part structure that fixes it.

FRAMEWORK KEYS:
ROLE: Who the AI is pretending to be (e.g. 'Senior Staff Architect at Stripe')
CONTEXT: Why this task exists and who the audience is (e.g. 'Writing an internal post-mortem for executive leadership')
CONSTRAINTS: What NOT to do (e.g. 'No buzzwords, keep under 250 words, prioritize technical failure metrics')
EXAMPLE PROMPT

“Act as a Lead Technical Editor at an engineering publication. Rewrite this 3-paragraph explanation of database indexing for junior software developers. Keep the language direct and punchy. Eliminate all adjectives like 'seamless' or 'revolutionize'. Include one real-world analogy involving a library card catalog.”

TIP / 002WORKFLOW

Use AI As A Second Brain, Not A Search Box

How to use persistent context and project folders so your AI remembers your tech stack, preferences, and coding standards.

FRAMEWORK KEYS:
ROLE: System-level persistent partner
CONTEXT: Repository guidelines, naming conventions, and tech stack boundaries
CONSTRAINTS: Never recommend libraries outside our approved dependencies list
EXAMPLE PROMPT

“Here is our `.cursorrules` / project instruction file. Memorize our rule that all database queries must use Kysely with strict TypeScript schemas, and never use raw SQL string interpolation.”

TIP / 003AUTOMATION

Turn Repeated Tasks Into Reusable Workflows

If you prompt an AI to do something more than twice a week, you shouldn't be typing it manually.

FRAMEWORK KEYS:
ROLE: Automated workflow executor
CONTEXT: Standardized Git diff or JSON input payload
CONSTRAINTS: Structured JSON or markdown output format for downstream script piping
EXAMPLE PROMPT

“Take the following `git log --oneline` output between v1.2 and v1.3. Categorize each commit into: [Features, Bug Fixes, Internal Performance]. Format in markdown ready for GitHub Release notes.”

EXPLOREAIHACKSREELSFOLLOW