--- id: 550e8400-e29b-41d4-a716-446655440000 category: "10_Wiki/Topics/Agent & AI" confidence_score: 1.0 tags: [Agent, AI, Wiki, Reinforcement Learning, Karpathy] last_reinforced: 2026-04-21 github_commit: "initial" --- # [[P-Reinforce]] ## ๐Ÿ“Œ ํ•œ ์ค„ ํ†ต์ฐฐ (The Karpathy Summary) > ํŒŒํŽธํ™”๋œ ์ •๋ณด๋ฅผ ์ž์œจ์ ์œผ๋กœ ๊ตฌ์กฐํ™”ํ•˜๊ณ  ์—ฐ๊ฒฐํ•˜์—ฌ ์Šค์Šค๋กœ ์„ฑ์žฅํ•˜๋Š” '์™ธ๋ถ€ ๋‡Œ'๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๊ฐ•ํ™”ํ•™์Šต ๊ธฐ๋ฐ˜ ์ง€์‹ ์—”์ง„. ## ๐Ÿ“– ๊ตฌ์กฐํ™”๋œ ์ง€์‹ (Synthesized Content) - **์ถ”์ถœ๋œ ํŒจํ„ด:** Karpathy์˜ LLM-Wiki ์•„ํ‚คํ…์ฒ˜๋ฅผ ์‹ค์ „ ์—์ด์ „ํŠธ ์Šคํ‚ฌ๋กœ ๊ตฌํ˜„ํ•˜์—ฌ, ์ง€์‹์˜ ์—”ํŠธ๋กœํ”ผ๋ฅผ ๋‚ฎ์ถ”๊ณ  ์—ฐ๊ฒฐ์„ฑ์„ ๊ทน๋Œ€ํ™”ํ•จ. - **์„ธ๋ถ€ ๋‚ด์šฉ:** - **RL Logic**: $R = w_1(Accuracy) + w_2(Connectivity) + w_3(Satisfaction)$ ๊ณต์‹์„ ํ†ตํ•ด ์ตœ์ ์˜ ํด๋”๋ง ์ˆ˜ํ–‰. - **Autonomous Folderling**: 85% ์ด์ƒ์˜ ์œ ์‚ฌ๋„ ์‹œ ๊ธฐ์กด ํด๋” ๋ฐฐ์น˜, ์‹ ๊ทœ ๊ฐœ๋… ๋“ฑ์žฅ ์‹œ ์ฆ‰์‹œ ์นดํ…Œ๊ณ ๋ฆฌ ํ™•์žฅ. - **Git Sync**: ๋ชจ๋“  ์ง€์‹์˜ ๋ณ€ํ™”๋ฅผ GitHub ํƒ€์ž„๋ผ์ธ์— ์˜์†์ ์œผ๋กœ ๊ธฐ๋ก. ## โš ๏ธ ๋ชจ์ˆœ ๋ฐ ์—…๋ฐ์ดํŠธ (Contradictions & RL Update) - **๊ณผ๊ฑฐ ๋ฐ์ดํ„ฐ์™€์˜ ์ถฉ๋Œ**: ๊ธฐ์กด ์ˆ˜๋™ ์œ„ํ‚ค ๊ด€๋ฆฌ ๋ฐฉ์‹์˜ ์ •์  ๊ตฌ์กฐ ํ•œ๊ณ„๋ฅผ ๊ทน๋ณตํ•˜๊ณ  ๋™์  ๊ทธ๋ž˜ํ”„ ๊ตฌ์กฐ๋กœ ์ „ํ™˜. - **์ •์ฑ… ๋ณ€ํ™”**: ์‚ฌ์šฉ์ž์˜ "์ด ํด๋” ์•„๋‹ˆ์•ผ" ํ”ผ๋“œ๋ฐฑ์„ ์ˆ˜์ง‘ํ•˜์—ฌ `20_Meta/Policy.md`์— ๋ฐ˜์˜, ๊ฒฝ๊ณ„์„ (Boundary)์„ ์žฌ์„ค์ •ํ•จ. ## ๐Ÿ”— ์ง€์‹ ์—ฐ๊ฒฐ (Graph) - **Parent**: Agent Ecosystem - **Related**: Knowledge Automation, Recursive Structuring - **Raw Source**: 00_Raw/2026-04-21-P-Reinforce_Skill_Info