Twitxy Saved

Yoni Fraimorice

@yonifra

Senior software engineer & AI champion 👑 @microsoft, ex-@wix, husband, and father of three cute bandits
Israel yonifra.com Joined August 2009
361 Following 95 Followers 958 Posts

Latest posts 5

Yoni Fraimorice @yonifra
OpenAI מדווחת שסוכניה הפיקו 948 ניצולי הסלמת הרשאות ייחודיים בלינוקס, לפי ניתוח הווריאנטים שלה, בעלות כ-1.2 מיליון דולר בקרדיטים. הניסוי נערך בסביבה מבוקרת ונדרש לתוקף כבר להיות בעל גישה מקומית. זה דיווח החברה, לא אישור עצמאי לכל הממצאים.

openai.com/index/a-thousa…
openai.com A thousand fewer paths to root OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel.
Yoni Fraimorice @yonifra
מפתח PostgreSQL רוברט האס קושר את עיכוב גרסה 19 לעומס דיווחי באגים שנמצאו בעזרת AI. בפוסט מ-30.10 הוא מציין עלייה של כ-60% בקומיטים שתיקוניהם הוחלו גם על גרסאות קודמות. הזווית: גילוי מהיר יותר לא מבטל את הצורך לאמת ולתקן. זו הערכתו, לא סיבה יחידה שהוכחה.

rhaas.blogspot.com/2026/10/postgr…
Yoni Fraimorice @yonifra
לפי Ars Technica, גוגל מתכננת לכבות כברירת מחדל כלי AI וגם Search בחשבונות תלמידים המנוהלים בידי בתי ספר, עד שהמנהל יפעיל אותם. לא תהיה אפשרות להשאיר חיפוש ולכבות רק AI. השינוי צפוי בחודשים הקרובים, לא חסימה שכבר חלה על כל תלמיד.

arstechnica.com/ai/2026/11/goo…
arstechnica.com Google will let schools skip AI tools, but they have to give up search, too Google will offer schools AI tools that purportedly don't give away the answers.
Yoni Fraimorice @yonifra
בתיקון ל-S-1, Anthropic אומרת שהערכות לגבי מודלים עתידיים מצביעות על אפשרות שתעמוד בקרוב בסף שלה להאצת התקדמות בעזרת AI. לפי המסמך, אף מודל נוכחי לא הוגדר כחוצה אותו. זו הערכה לגבי העתיד, לא הכרזה ש-Claude כבר בונה את הדור הבא לבדו.

techcrunch.com/2026/11/02/ant…
techcrunch.com Anthropic warns that it's very close to RSI capability, aka self-building AI | TechCrunch RSI is the point at which AI grows so good at training itself that it spawns a dramatic acceleration of AI capabilities. To the AI doomers, RSI represents the cataclysm from which there is no return.
Yoni Fraimorice @yonifra
OpenAI מציגה את Grid Protectors: קמפוסי AI עתידיים שיפחיתו צריכת חשמל כשהרשת בעומס. בפיילוט המתוכנן בג'ורג'יה החברה מתחייבת לממן את מלוא עלויות התשתית והשירות החשמלי, עם עד 1,000 מגה-וואט של גמישות ביקוש. התחייבות לעתיד, לא חיסכון שנמדד כבר.

openai.com/index/grid-pro…
openai.com Introducing the Grid Protectors Program OpenAI’s Grid Protectors Program will work with utilities to help future compute campuses support grid reliability and protect consumer electricity costs.