🤖 AI Engineering · Fine-tuning & Adaptation

Decide between prompting, RAG, and fine-tuning

Pick the cheapest adaptation layer that solves the problem using a decision ladder.

foundation~20 minAI EngineersML EngineersLLM App Developers

Steps

1Try prompt engineering first: most 'model gaps' are specification gaps
2If failures are missing knowledge, add retrieval, not weights
3If failures are format/tone/style consistency, consider light fine-tuning
4Estimate 12-month cost per option including maintenance, not just training
5Prototype the winner in one week; if gains are marginal, revert to cheaper tier
6Document the decision and revisit when base models jump a generation

Common Pitfalls

▲Fine-tuning to inject facts that go stale within months
▲Skipping the cheap baseline entirely

Commands

Install with skills CLI

$ npx skills add aniruddhaadak80/skills --skill fine-tuning-adaptation-when-not-to-finetune

Install globally

$ npx skills add aniruddhaadak80/skills --skill fine-tuning-adaptation-when-not-to-finetune -g

Tags

#strategy#architecture#cost#ai-engineering#fine-tuning-adaptation

Install this skill

$ npx skills add aniruddhaadak80/skills --skill fine-tuning-adaptation-when-not-to-finetune -a opencode

SKILL.md

# Decide between prompting, RAG, and fine-tuning

> Pick the cheapest adaptation layer that solves the problem using a decision ladder.

## When to Use This Skill

Pick the cheapest adaptation layer that solves the problem using a decision ladder.

## Steps

1. Try prompt engineering first: most 'model gaps' are specification gaps
2. If failures are missing knowledge, add retrieval, not weights
3. If failures are format/tone/style consistency, consider light fine-tuning
4. Estimate 12-month cost per option including maintenance, not just training
5. Prototype the winner in one week; if gains are marginal, revert to cheaper tier
6. Document the decision and revisit when base models jump a generation

## Common Pitfalls

- Fine-tuning to inject facts that go stale within months
- Skipping the cheap baseline entirely

View source on GitHub ↗

Clean, deduplicate, and balance instruction-response pairs so fine-tuning learns behavior rather than noise.

🤖 AI Engineering·~60m

Open →

Decide between prompting, RAG, and fine-tuning

Steps

Common Pitfalls

Commands

Tags

Related skills