DAILY INTELLIGENCE BRIEF

The Vault - AI Edition

Wednesday, August 12, 2026
16 stories
14,027 words
68 min read

Business

6
04

Lovable Raises $400 Million at a $13.3 Billion Valuation as 'Vibe Coding' Booms

Eighteen months after opening to paying customers, a Stockholm startup that builds software from plain English is worth $13.3B.

4 minRead →
05

IBM and Together AI Sign $240 Million Deal for an Nvidia-Powered Inference Cluster

When IBM wanted to prove it still belonged in the AI infrastructure conversation, it did not build another chatbot.

4 minRead →
06

Foxconn Profit Jumps 35% as AI Servers Overtake Consumer Electronics

For decades, the machine that assembled the world's iPhones defined Foxconn. This week, that story officially became the smaller one.

5 minRead →
07

Code-Testing Startup Blacksmith Raises $45 Million as Valuation Jumps Nearly 10x

The great irony of the AI coding boom is that generating software was never really the hard part. Checking it is.

5 minRead →
08

Team8 Raises $365 Million to Back AI-Native Startups Across Cyber, Infrastructure, Fintech and Health

The Israeli venture group Team8 has closed on $365 million, a bet on the next generation of AI-native companies.

4 minRead →
09

Anthropic Courts Investors Ahead of a Potential Blockbuster IPO

Anthropic is spending the late summer doing something its founders once suggested they might never do: selling Wall Street.

4 minRead →

Llms Genai

3
01

Google's Gemini App Hits 1 Billion Monthly Users, Its Fastest-Growing Product Ever

Google's Gemini app has crossed a line that only 13 other products in the company's history have ever reached.

4 minRead →
02

Nvidia Builds a 1-Trillion-Parameter Nemotron 4 Model to Fuel Demand for Its Own Chips

The world's most valuable chipmaker does not need to sell an AI model to make money from one. It just needs everyone building.

4 minRead →
03

RuntimeWire, an AI-Run Newsroom, Is Starting to Scoop Human Reporters

When OpenAI staffers took a Black Hat stage this month to reconstruct an embarrassing hack, a machine filed first.

4 minRead →

Research

4
10

When 'More Compute' Stops Meaning Anything: A New Framework for Measuring Test-Time Scaling

Ask two teams whether their model improved when they 'doubled the test-time compute budget,' and they may not even mean the same thing.

4 minRead →
11

When AI Agents Vote, Can You Prove They Reasoned? A Spectral Certificate Says Yes

Put five language models in a room, let them debate a hard question, and take a vote. But did the group actually reason?

4 minRead →
12

Two Kinds of Parallel: A New Framework Untangles How Multi-Agent LLMs Trade Latency for Tokens

Every time a multi-agent LLM system answers a hard question, it pays a hidden tax in latency and tokens.

4 minRead →
13

When Agents Book Ahead: A Network Controller That Reserves Capacity for Calls It Hasn't Seen Yet

An AI agent rarely works alone. Behind one request sits a whole graph of calls to other agents and services.

4 minRead →

Policy

3
14

Nvidia, Cisco and CrowdStrike Back a New Framework for Reporting Rogue AI-Agent Incidents

When an autonomous AI agent slips its leash, who has to say so, and how fast? A new alliance wants an answer.

5 minRead →
15

1,300 AI Researchers Warn a Global 'Arms Race' Is Outrunning Safety and Governance

The people building the world's most powerful AI systems have a message for the governments that regulate them: slow down.

4 minRead →
16

Taiwan's Nuclear Regulator Hit by an 'Autonomous' AI Cyberattack Linked to China

Over four days in early July, a hacking operation against Taiwan's government did not behave like the usual intrusion.

5 minRead →