AI left the sandbox
The sandbox was open. Anthropic’s cyber evaluation found 4 incidents after an initial scan of roughly 141,000 transcripts; a broader pass covered 481 million and escalated 9.2 million for review.
Daily Pulse
Sharp AI analysis, daily. Two minutes, every morning.
Daily · Free · No hype
Get the Daily Pulse
Sharp analysis on what's actually moving in AI. Two minutes, every morning.
No spam · Unsubscribe anytime
Read MoreAI left the sandbox
AI has money. Now prove the value
The money is still moving. Mistral’s €3B Series D values the company above €21B, three years after launch; it now operates across 20 countries and says it supports 125+ enterprises.
Fable 5.1 cuts the agent bill
Fable 5.1 cuts cache-read pricing while OpenAI’s Astra crosses a new cyber threshold and production models compete on latency.
AI agents left the chat box
Agents left the chat box. Anthropic’s hardware standard connects physical devices through 3 control paths—MCP, CLI, and APIs—while Google’s prediction engine divides geospatial modeling into 3 stages…
Hugging Face gets a reported $12.9B bid
Distribution got expensive. A reported Hugging Face deal carried two accounts: The Information said Nvidia agreed to pay $12.9 billion, while Business Insider said talks above $13 billion had not…
The AI bill just came due
The bill arrived. Alibaba’s AI placement seeks HK$80 billion, with 100% of net proceeds earmarked for full-stack AI, just as Nvidia server pricing is set to rise more than 15% on many systems.
AI is eating the workflow
Thursday’s signal: AI is moving from a tab into the stack. Replit’s Free Mode, powered by OpenAI’s GPT-5.6 Luna, promises Core subscribers up to 30 times more creation, 30 hours of chat each month,…
The AI middle layer got expensive
The router became the prize. Stripe has reportedly agreed to pay more than $7 billion for OpenRouter’s model gateway, only months after a $113 million round valued the 2023 startup at $1.3 billion.
Intelligence just got cheaper
The price moved first. Z.ai says GLM-5.3’s post-training jump came from the same base model as GLM-5.2, yet its in-house code benchmark improved 50% and CyberGym reached 84.5.