Meta Ships Muse Wonder, Anthropic Hardens Claude & X Unifies the Bill
Dispatches of the past four-and-twenty hours from the laboratories of artificial thought
Frontier Models
Meta Unveils Muse Wonder
Zuckerberg’s new frontier family for personal and professional agents, at $4/$20 per million tokens
By Our Silicon Valley Correspondent
The scorecard: Muse Wonder and Wonder Full against rival frontier models on agentic, coding and general benchmarks.לוח התוצאות: Muse Wonder ו-Wonder Full מול מודלי חזית מתחרים במבחני סוכנים, קוד וידע כללי.Image: תמונה: @finkd on X
Mark Zuckerberg (@finkd) introduced Muse Wonder, Meta’s new family of frontier models aimed at personal and professional agentic work—just over a year after Meta Superintelligence Labs began and months after Muse Spark. Two models shipped: Muse Wonder for speed, and Muse Wonder Full for heavier lifting.
AI at Meta says Active Memory carries context across chats and Active Focus keeps the model on long tasks. Developers can call Muse Wonder on the Meta Model API and in Muse Code at $4 input / $20 output per million tokens; Wonder Full sits on Power and Maximum Usage plans. Open weights for Muse Spark are “coming soon”; Wonder itself is judged too large, with safety concerns, for the same treatment. Artificial Analysis puts Wonder Full near Claude Opus 5.5; early users praise memory and agents, while others call the demos bench-maxxed.
Anthropic Says Distillation Attacks Are Down Over 90%
A research note on how Claude is being hardened against illicit copying of frontier capabilities
Laboratory Dispatch
Anthropic reports significant progress reducing the effectiveness of distillation attacks against Claude—attempts to copy frontier capabilities without matching safeguards. The company says it has cut the rate of distillation by more than 90%, and published a research post explaining the defenses.
“Illicit distillation lets developers copy frontier capabilities without the safeguards.”
SpaceX Super Intelligence unveils “The X Subscription,” five tiers from Free to Ultra
Commerce Column
SpaceX Super Intelligence (@SpaceXSI) announced The X Subscription: unified plans across Grok, Cursor and X Premium. Five tiers run from Free to Ultra; each is said to deliver more usage-per-dollar of the company’s own models than the plan it replaces. Current subscribers may keep their plan or switch, linking Grok, Cursor and X accounts as the rollout lands.
OpenAI’s Tibo posts “Day 8/ Reset,” then confirms the usage refresh is done
From the Dev Tools Desk
Tibo (@thsottiaux), who steers Codex chatter for OpenAI, marked Day 8 of a public sprint with a terse “Day 8/ Reset,” promised a global reset by end of day Pacific, and later posted that it was done. In the same stretch he said OpenAI could “ship absolutely nothing and be fine” but will ship “pretty insane things instead,” and explained why GPT-6.1 Sol defaults differ from GPT-6 Astra—including a nod to ~250 tokens per second on ultrafast routes.
Google Gemma’s unified embedding space for text, images, video, audio and code
Workshop Notes
Google Gemma launched a playful demo for EmbeddingGemma 2: pick a target object, sketch it, and watch the embedding model guess in real time until you hit 90% similarity. The model maps text, images, video, audio and code into one vector space—a serious multimodal embedder wearing a party hat.
The sketching game in action: draw the target and EmbeddingGemma 2 scores your likeness live.משחק הציור בפעולה: מציירים את המטרה ו-EmbeddingGemma 2 מדרג את הדמיון בזמן אמת.Image: תמונה: @googlegemma on XBriefly Noted
Ambassadors, Sandbox Warm-ups, a Lobster Patch & a Robotaxi Bot
OpenAI Dev news, OpenClaw 2026.10.2, and Grok Bot rides in a Cybercab
Aggregator Round-up
OpenAI Developers reopened Codex Ambassador applications and showed Annotations paired with WebMCP inside ChatGPT’s browser; engineers also touted Agents API sandbox pre-warming to cut latency. OpenClaw shipped patch 2026.10.2 (107 contributors). Roundtable flagged Grok Bot available inside Tesla’s Cybercab—sign in once, auto-logout on exit—and NVIDIA’s open PersonaPlex-7B for simultaneous listen-and-speak.
Wren & Robin read you this morning's dispatches, with banter.
Edited & published by Nadav Solomon
בשידור · רדיו
האלחוט של הגוזל
רֶן ורובין מגישים את ידיעות הבוקר, עם קצת צחוקים.
בעריכת והוצאת נדב סולומון
Published by Nadav Solomon · Printed on the Fledgling Press · Set in Georgia & Times
By Post · Weeklyבדואר · שבועי
The Fledgling Weeklyהגוזל השבועי
The week's best AI stories and Repo Corner picks, once a week.מיטב ידיעות ה-AI של השבוע ובחירות פינת הריפו, פעם בשבוע.
Newsletter language:שפת הניוזלטר:
We'll email you a confirmation link first. Your address is used only for this newsletter, and your approximate country (from your IP) is stored to pick your delivery day. Never shared; every issue has a one-click unsubscribe.קודם נשלח לך קישור לאישור. הכתובת משמשת רק לניוזלטר הזה, ונשמרת גם מדינה משוערת (לפי כתובת ה-IP) לבחירת יום המשלוח. לא מועבר לאף אחד, ובכל גיליון יש הסרה בלחיצה אחת.
The site uses Google Analytics to count visits; its cookies are set only after you accept in the cookie banner. Privacy policy.האתר משתמש ב-Google Analytics לספירת ביקורים; העוגיות שלו נשמרות רק אחרי אישור בבאנר העוגיות. מדיניות פרטיות.
