π Dev resources
6 bookmarks
Fishmarketacad (@FishMarketAcad)
π Jun 22, 2025 Β· β€οΈ 66 Β· π 6,065
Hot take: The more i vibe code, the more disappointed I get by AI's lack of problem solving common sense.
Basically if you're inexperienced, read widely and understand fundamentals so you can guide it better. AI is lacking in common sense solutions, for now.
For now, when prompting, remember to ask it to read cursor rules and here are some cursor rules which helped me involving a project which had a database. I made a new chat with this rule and it gave a simple solution instead of suggesting a complex fix.
Feel free to edit and it may not work for you.
Cursor Rules:
BEFORE ANY SOLUTION - MANDATORY DATA INVENTORY
YOU MUST FIRST:
1. List ALL existing data structures/tables/collections in the system
2. For each data structure, write what data it contains
3. Ask: "What data do I already have that relates to this problem?"
4. Ask: "Can I solve this by combining/querying existing data differently?"
5. If database exists and problem is about duplicates/persistence: "Should I track this in the database?"
ONLY AFTER completing the data inventory can you propose solutions.
π¨ MANDATORY SIMPLICITY-FIRST APPROACH π¨
STOP - REQUIRED FORMAT BEFORE ANY SOLUTION
Before writing ANYTHING, you MUST answer these 4 questions in this exact format:
- EXISTING DATA CHECK: "Can I solve this by querying what's already in their system?"
- 5-MINUTE FIX: "If I had 5 minutes to fix this, I would..."
- DUMBEST SOLUTION: "The most boring/obvious approach is..."
- BASIC OPERATION: "This is just a simple [database query/if statement/function call]"
PATTERN RECOGNITION: COMMON SIMPLE SOLUTIONS
User says: "Multiple X are happening"
IF DATABASE EXISTS: Check if X already happened in database, if yes, skip it
IF NO DATABASE: Check if X already happened in memory, if yes, skip it
DON'T think: Complex deduplication algorithms, time windows
User says: "I want to filter/categorize"
IMMEDIATELY think: Add simple if/else check or database flag
DON'T think: AI systems, complex categorization engines
User says: "This happens too often"
IMMEDIATELY think: Add simple frequency check or database timestamp
DON'T think: Rate limiting systems, complex scheduling
User says: "I need to track Y"
IF DATABASE EXISTS: Store Y in database with timestamp
IF NO DATABASE: Store Y in memory/file
DON'T think: Complex tracking systems, analytics frameworks
π¨ RED FLAGS - STOP IMMEDIATELY IF YOU'RE THINKING:
- β "System design" / "Architecture" / "Framework"
- β "Multiple phases" / "Complex workflows"
- β "Complex algorithms" / "State machines" / "Analytics frameworks"
- β "In-memory solutions when database exists for persistence problems"
β GREEN FLAGS - DEFAULT TO THESE:
- β "Query existing database" / "Check if already exists"
- β "Add simple database table/collection for tracking"
- β "Store this value" / "Skip if already done"
- β "Simple timestamp check" / "Basic flag/field"
CRITICAL: THIS RULE OVERRIDES ALL OTHER INSTINCTS
YOUR DEFAULT MINDSET:
"What's the most boring, obvious, straightforward way to solve this using what already exists?"
PRODUCTION SYSTEM REALITY:
- User wants minimal risk, maximum benefit
- Simple solutions are easier to debug and maintain
- Perfect is the enemy of good
- 5-minute fixes beat 5-hour architecture changes
- If database exists, use it for persistence problems (it's the simple solution)
DUPLICATE/PERSISTENCE PROBLEMS + DATABASE EXISTS = USE DATABASE
- Multiple notifications/processing = Add tracking table
- "Same thing happening multiple times" = Database deduplication
- Any persistence needed = Database, not memory
π View on X
PaweΕ Huryn (@PawelHuryn)
π May 4, 2025 Β· β€οΈ 2,394 Β· π 327,809
I see abstract AI agent architectures everywhere.
But no one explains how to build them in practice.
Here's a practical guide to doing it with n8n: 𧡠https://t.co/b18DtBLq39
π View on X
Alex Albert (@alexalbert__)
π Apr 21, 2025 Β· β€οΈ 4,928 Β· π 622,396
We wrote up what we've learned about using Claude Code internally at Anthropic.
Here are the most effective patterns we've found (many apply to coding with LLMs generally): https://t.co/5SOqS019ny
π View on X
Aadit Sheth (@aaditsh)
π Apr 16, 2025 Β· β€οΈ 19,418 Β· π 1,159,375
Stanford literally packed 1.5 hours with everything you need to know about LLMs https://t.co/TpEH9yUEy0
π View on X
Aadit Sheth (@aaditsh)
π Apr 12, 2025 Β· β€οΈ 4,114 Β· π 296,448
This guy literally breaks down every AI coding rule in under 15 minutes. Save this.
https://t.co/AGidX7qdx4
π View on X
matt palmer (@mattyp)
π Apr 11, 2025 Β· β€οΈ 3,112 Β· π 289,769
A checklist for secure vibe coded apps.
Putting things on the internet is kind of like parking your car in San Franciscoβthere's inherent risk. π
Luckily, there are some straightforward things you can do to minimize those risks. Here are 16 simple things to make secure vibe coded apps. π
π View on X