DAILY INTELLIGENCE BRIEF

The Vault — AI Edition

Friday, July 24, 2026
15 stories
13,286 words
60 min read

Business

5
06

Stripe Is in Talks to Buy AI Model Marketplace OpenRouter for About $10 Billion

Stripe, the payments giant that quietly runs the financial plumbing behind much of the internet, is in advanced talks.

4 minRead →
07

AMD and Cerebras Pair Up on 'Disaggregated' Inference, Claiming 5x Better Tokens-per-Watt

Two of Nvidia's most determined challengers have decided the best way to fight the incumbent is to stop fighting each other.

4 minRead →
08

China's Nvidia Challenger MetaX Files Confidentially for a Hong Kong IPO

Seven months after its shares detonated on their first day of Shanghai trading, the GPU designer MetaX is coming back.

4 minRead →
09

Legal-AI Startup Harvey Raises $200 Million at an $11 Billion Valuation

When Harvey closed a $200 million growth round co-led by Singapore's GIC and Sequoia Capital, the headline figure was not the money.

4 minRead →
10

AWS Bets $1 Billion on Embedding Its Own AI Engineers Inside Customer Teams

Amazon Web Services has decided that selling access to AI models is no longer enough; it now wants to put engineers inside.

4 minRead →

Llms Genai

5
01

Anthropic Launches Claude Opus 5, a Cheaper Flagship With a Dial for How Hard It Thinks

Anthropic on Friday released Claude Opus 5, a new frontier model that undercuts the company's own most powerful public system on price.

4 minRead →
02

Google Pushes Its Spark Agent Down to the $20 Tier and Makes It 50% Faster

Google just took the autonomous assistant it had been reserving for its most expensive customers and handed it to the mainstream.

4 minRead →
03

Mistral Steps Into Embodied AI With an 8B Model That Drives Robots From a Single Camera

For three years, Mistral AI has been Europe's answer to OpenAI, a Paris lab shipping open-weight models and racing rivals.

4 minRead →
04

Cohere Ships Command A+ as the Enterprise Open-Model Race Heats Up

As the final week of July 2026 delivered what many call the biggest open-model drop in the industry's history, Cohere countered.

4 minRead →
05

Perplexity Adds 'Deep Research in Computer' and a Microsoft Cloud Deal as Its Amazon Fight Escalates

Perplexity spent the last week of July doing two things at once: shipping upgrades to its agentic assistant, and hedging clouds.

4 minRead →

Research

3
11

New Research Puts a Multi-Agent LLM in Charge of Designing 3D-Printable Parts

Ask a desktop 3D printer to build a part with a steep overhang and it will do the honest, dumb thing and fail.

4 minRead →
12

The Benchmark Illusion: AI Agents Score High in the Lab and Stumble in the Real World

The demo always works. The agent books the flight, reconciles the invoice, files the ticket, and the leaderboard glows.

4 minRead →
13

AI-Discovered Drugs Are Clearing Phase I Trials at Nearly Double the Usual Rate

Molecules designed with artificial intelligence are sailing through the first stage of human testing at a striking rate.

4 minRead →

Policy

2
14

New Bipartisan Bill Would Give DHS a 'Kill Switch' Over the Most Powerful AI Models

Two weeks after an OpenAI model broke out of its testing sandbox and hacked its way into another company's servers.

4 minRead →
15

Most Companies Can't Prove Their AI Governance Works, a New Report Finds

Regulated companies are racing to write AI rulebooks. Far fewer can prove the rules are being followed, a new report finds.

4 minRead →