Sign In

BUILD-AND-FIND: An Effort-Aware Protocol for Evaluating Agent-Managed Codebases

์ž‘์„ฑ์ž
  • Haebom
์นดํ…Œ๊ณ ๋ฆฌ
Empty

์ €์ž

Jhen-Ke Lin

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ๋…ผ๋ฌธ์€ AI ์—์ด์ „ํŠธ๊ฐ€ ์ƒ์„ฑํ•œ ์ฝ”๋“œ ๋ฒ ์ด์Šค๋ฅผ ํ‰๊ฐ€ํ•˜๋Š” ์ƒˆ๋กœ์šด ํ”„๋กœํ† ์ฝœ์ธ BUILD-AND-FIND๋ฅผ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค. ์ด ํ”„๋กœํ† ์ฝœ์€ ์ฝ”๋“œ์˜ ๊ธฐ๋Šฅ์  ์ •ํ™•์„ฑ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ์ดํ›„ ์—์ด์ „ํŠธ๊ฐ€ ์ฝ”๋“œ์˜ ์˜๋„์™€ ์„ค๊ณ„ ์„ ํƒ์„ ์–ผ๋งˆ๋‚˜ ์‰ฝ๊ฒŒ ํŒŒ์•…ํ•˜๊ณ  ์žฌํ˜„ํ•  ์ˆ˜ ์žˆ๋Š”์ง€์— ์ดˆ์ ์„ ๋งž์ถฅ๋‹ˆ๋‹ค. BUILD-AND-FIND๋Š” ์ฝ”๋“œ ์ƒ์„ฑ์˜ ๋ช…ํ™•์„ฑ๊ณผ ๊ฒ€ํ†  ์šฉ์ด์„ฑ์„ ์ธก์ •ํ•˜์—ฌ, ์—์ด์ „ํŠธ๊ฐ€ ์ƒ์„ฑํ•œ ์ฝ”๋“œ ๋ฒ ์ด์Šค๋ฅผ ๋‹จ์ˆœํžˆ ๊ธฐ๋Šฅ์ ์ธ ๊ฒฐ๊ณผ๋ฌผ์ด ์•„๋‹Œ, ํ˜‘์—…์„ ์œ„ํ•œ ํšจ๊ณผ์ ์ธ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ๋„๊ตฌ๋กœ ํ‰๊ฐ€ํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
ํ–ฅ์ƒ๋œ ์ฝ”๋“œ ๋ฒ ์ด์Šค ํ‰๊ฐ€: ๊ธฐ์กด์˜ ๊ธฐ๋Šฅ์  ์ •ํ™•์„ฑ ํ‰๊ฐ€๋ฅผ ๋„˜์–ด, ์ฝ”๋“œ์˜ ๋ช…ํ™•์„ฑ, ์ดํ•ด ์šฉ์ด์„ฑ, ๊ทธ๋ฆฌ๊ณ  ์žฌํ˜„ ๊ฐ€๋Šฅ์„ฑ์„ ์ข…ํ•ฉ์ ์œผ๋กœ ํ‰๊ฐ€ํ•  ์ˆ˜ ์žˆ๋Š” ์ƒˆ๋กœ์šด ๊ธฐ์ค€์„ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค.
โ€ข
AI ์—์ด์ „ํŠธ ํ˜‘์—… ์ด‰์ง„: ์—์ด์ „ํŠธ๊ฐ€ ์ƒ์„ฑํ•œ ์ฝ”๋“œ ๋ฒ ์ด์Šค๊ฐ€ ํ›„์† ์—์ด์ „ํŠธ์˜ ์ž‘์—… ํšจ์œจ์„ฑ์— ๋ฏธ์น˜๋Š” ์˜ํ–ฅ์„ ์ •๋Ÿ‰์ ์œผ๋กœ ์ธก์ •ํ•จ์œผ๋กœ์จ, AI ํ˜‘์—… ์‹œ์Šคํ…œ์˜ ์„ฑ๋Šฅ์„ ๊ฐœ์„ ํ•˜๋Š” ๋ฐ ๊ธฐ์—ฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
โ€ข
์ธ๊ฐ„์˜ ๊ฐœ์ž… ๋ฐ ๊ฐ๋… ํ•„์š”์„ฑ: ์‹คํ—˜ ๊ฒฐ๊ณผ, ๋ณต์žกํ•œ ์ž‘์—…์˜ ๊ฒฝ์šฐ ์—ฌ์ „ํžˆ ์ธ๊ฐ„์˜ ๊ฐ๋… ๋ฐ ๋””๋ฒ„๊น…์ด ํ•„์š”ํ•จ์„ ์‹œ์‚ฌํ•˜๋ฉฐ, BUILD-AND-FIND ํ”„๋กœํ† ์ฝœ์ด ์ž๋™ํ™”๋œ ํ‰๊ฐ€๋ฅผ ์ œ๊ณตํ•˜์ง€๋งŒ ์‹ค์ œ ๊ฐœ๋ฐœ ๊ณผ์ •์—์„œ์˜ ๋ณต์žก์„ฑ์„ ์™„์ „ํžˆ ์ œ๊ฑฐํ•˜์ง€๋Š” ๋ชปํ•จ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
๐Ÿ‘
Made with Slashpage