Sungjae Lee
Founder & CEO of blex LLC. Editor overseeing Creeta's research and editorial process.
An 80-year conjecture fell to AI. What was actually proved?
Unnamed OpenAI model disproved Erdős's unit-distance conjecture. Sawin's n^1.014 is the first exponent gain in 80 years.
Copilot Cowork, 받는 사람 아니면 허락을 묻는다
PromptArmor의 2026년 5월 POC에 따르면 Copilot Cowork의 Skills 인젝션 + 자기 전송 공백 = 조용한 M365 유출.
Copilot Cowork asks permission — unless you're the recipient
Copilot Cowork's Skills injection + self-send gap = silent M365 exfiltration, per PromptArmor's May 2026 POC.
Grok Build의 /goal: 'Complete'가 뜨면 무엇을 확인했나?
xAI의 /goal: Grok Build가 목표를 받아 완료까지 실행하며, 내장 검증과 4가지 조작 명령을 제공한다.
Grok Build's /goal: when 'Complete' appears, what was checked?
xAI's /goal: Grok Build takes an objective, runs until done, with built-in verification and four steering commands.
요양원 AI가 캔자스시티 버스를 스캔하게 된 경위
KCATA의 RideKC가 실시간 안면인식을 추가했다. 세 가지 감시목록, SafeSpace Global, 공개 정책·감사 부재.
How a nursing-home AI ended up scanning Kansas City buses
KCATA's RideKC adds live face recognition: three watch lists, SafeSpace Global, no published policy or audit.
TVM 0.25의 TIRx가 Triton이 감춘 것을 드러낸다
TVM v0.25.0은 TIRx를 추가한다. 명시적 Blackwell 오케스트레이션, 18패스 lowering 체인, 29개 타일 프리미티브, autoscheduler 없음.
TVM 0.25's TIRx reveals what Triton deliberately conceals
TVM v0.25.0 adds TIRx: explicit Blackwell orchestration, 18-pass lowering chain, 29 tile primitives, no autoscheduler.
Fugu beats each component it calls — and the asterisks matter.
Fugu coordinates frontier LLMs via one API. What the benchmark sheet covers, what it omits, and caveats for builders.