Tuesday 22 September 2026
Top stories today:
Xiaomi’s MiMo-V2.6 Pro debuts as new top open-weight model
Meta’s Muse Mac app flaw exposed user authentication tokens
SpaceXAI’s Grok 4.7 debuts with improved self-verification
Google unveils $899+ Googlebook hardware with Android integration
Oura, smart ring maker, seeks to raise up to $2.2B in U.S. IPO
0. Data & Calendar
All values as of 6 AM ET / 3 AM PT, other than S&P500 and NASDAQ close (4 PM ET / 1 PM PT).
All times are ET.
Listen to our AI-generated podcast summarizing today’s newsletter (beware of hallucinations):
1. Xiaomi’s MiMo-V2.6 Pro debuts as new top open-weight model
MiMo-V2.6 Pro surpassed Alibaba’s Qwen3.8 Max and Z.ai’s GLM-5.3 on the Artificial Analysis Intelligence Index:
An MoE model with 1.02T parameters (42B active).
Costs $0.435/$0.87 per 1M input/output tokens, vs. GPT-5.6 Sol at $4/$20 (≤272K input tokens), Qwen3.8 Max at $2/$6, and GLM-5.3 at $1.4/$4.4.
MiMo-V2.6-Flash also debuted at $0.14/$0.28.
MiMo-V2.6-Pro UltraSpeed delivers up to 20x faster output than MiMo-V2.6 Pro, at $4.35/$8.70.
MiMo-V2.6 Pro and Flash are accessible via Xiaomi’s AI Studio, MiMo Code, and MiMo Desktop, with API access through the MiMo Open Platform.
Model weights are available through Hugging Face.
Get the chance to peek inside founders and leaders’ brains and see how they think about going from zero to 1 and beyond.
Join thousands of weekly readers at Google, OpenAI, Stripe, TikTok, Sequoia, and more.
Check all the tools and more here, and outperform the competition.
*Sponsored.
3. Meta’s Muse Mac app flaw exposed user authentication tokens, researcher finds
Patrick Wardle discovered the 0-day affecting the Muse app.
Meta said it issued a hotfix to patch the flaw.
Muse topped U.S. charts with 902K+ downloads in the 6 days after launch, vs. Meta AI’s 773K over the same post-launch period.
Separately, Shopify plans to let Muse complete purchases at its stores via one-tap checkout.
Your agentic finance team:
The Scout studies your business before you need to tell it anything, then builds your profile and custom chart of accounts and picks the metrics that matter most.
The Accountant automatically categorizes and reconciles your transactions, so books are right every day, not 10 days after month-end.
The CFO answers revenue, burn, and runway questions from your live books, and drills from any number down to the transaction behind it.
*Sponsored.
5. SpaceXAI’s Grok 4.7 debuts with improved self-verification, long-context handling
Grok 4.7 scored 46 on the Artificial Analysis Intelligence Index, bringing SpaceXAI into the top 4 AI labs:
Costs $2/$6 per 1M input/output tokens, the same as Grok 4.6, vs. GPT-5.6 Sol’s $4/$20, Claude Fable 5’s $10/$50, and Qwen3.8 Max’s $2/$6.
500K-token context window, unchanged from Grok 4.6.
Multimodal models are only as good as the data behind them. Our trainers work across text, audio, and image pipelines—handling transcription, labeling, and annotation tasks with speed and consistency.
Audio transcription and speech data labeling
Image and video annotation for vision model pipelines
Multilingual coverage across Portuguese, Spanish, and English
Part of our vetted LATAM talent network, working in U.S.-aligned time zones.
*Sponsored.
7. Google unveils $899+ Googlebook hardware with deep Android integration
Oct. 4: the new devices will go on sale in the U.S.
Up to 2.8K OLED displays with touchscreen support and Intel or Qualcomm processors.
