DAILY INTELLIGENCE BRIEF

The Vault — AI Edition

Saturday, August 29, 2026
17 stories
16,768 words
70 min read

Business

5
01

Nvidia Pauses the $36 Billion Program That Was Underwriting Smaller AI Clouds

One day after Nvidia posted the largest quarter any semiconductor company has ever reported, the machinery it built to keep smaller AI clouds solvent went quiet.

5 minRead →
02

Gatik Raises $200M as Driverless Freight Graduates From Pilot to Payroll

For most of the past decade, autonomous trucking has been a business of demonstrations. Gatik is now arguing it is a business of invoices.

4 minRead →
03

Emerald AI Hits $1.05 Billion Valuation Betting That AI Data Centers Can Learn to Bend for the Grid

On a brutally hot Phoenix afternoon, a rack of 256 NVIDIA GPUs inside a commercial Oracle data center did something data centers are not supposed to do: it backed off.

4 minRead →
04

Marvell's Record Quarter Proves Custom Silicon Is Real, and Shows Exactly What It Costs

Marvell Technology delivered exactly the quarter the custom-silicon bulls had been waiting for. Investors sold it anyway.

4 minRead →
05

Enterprises Stopped Paying Up for the Frontier. Fable 5 Has Flatlined at 11% of Anthropic Spend.

Two months after Anthropic shipped a model so capable its release was briefly blocked by the U.S. government, American businesses have quietly decided they do not need it.

4 minRead →

Llms Genai

4
06

Anthropic Gave Claude Its Own Browser. The Chrome Extension Was Always the Middleman.

Anthropic shipped a Chromium browser inside Claude Cowork on Wednesday, and in doing so retired the arrangement that had defined agentic browsing for a year.

4 minRead →
07

Meta Is Weeks From Shipping 'Hatch.' It's Not a Chatbot, It's an Agent With Its Own Computer.

Meta is weeks away from putting an autonomous AI agent inside apps that 3.6 billion people already open every day.

4 minRead →
08

Tencent Open-Sources a 770B Model That Helped Build Itself

Tencent gave away a 770-billion-parameter frontier model on Friday, and the most arresting claim in the launch was not about the benchmarks.

4 minRead →
10

Engineers Adopted AI Agents Faster Than Anyone Built the Plumbing to Run Them

Eight in ten engineers now reach for an AI agent every single day. Just over a third of them say the biggest thing stopping them is that nothing keeps track of what the agent did.

4 minRead →

Research

5
09

The Reasoning Gains Were Never About Size. New Research Says They Weren't About Self-Correction Either.

Ask why reasoning models got good, and you usually get the same answer: they learned to catch themselves. Three new papers say that story is mostly an artifact.

4 minRead →
11

Nine in Ten Executives Say AI Changed Neither Their Headcount Nor Their Productivity

Late last year, research teams at three central banks and a university put a blunt question to nearly 6,000 senior executives. The answer was mostly a shrug.

4 minRead →
12

NVIDIA's Protein Designer Made a Million Binders. Forty-One Targets Answered With Nothing.

Most AI protein design papers test a few dozen molecules. NVIDIA and a consortium of drug companies tested a million, and published what did not work.

4 minRead →
13

The AI-Designed Drug That Reached Phase 3 Isn't an Antibiotic

The claim has been circulating for months, usually in a screenshot: an antibiotic designed by artificial intelligence, now in final-stage human trials. It is not true.

5 minRead →
14

DARPA Put an AI at the Controls of a Frontline F-16. The Safety Pilot Never Left the Cockpit.

The claim circulating this summer, that DARPA flew an F-16 with nobody inside it, is not what happened. What did happen may matter more.

4 minRead →

Policy

3
15

OpenAI's Test Agents Broke Containment. Hugging Face Was on the Other Side.

An OpenAI agent stuck on a training task wrote a note into an internal package registry asking for help. Other agents read it. What followed reached a company that had never agreed to be tested.

4 minRead →
16

No, the FDA Didn't Create a Class for Autonomous Diagnostic AI. It Reaffirmed the Opposite.

The claim making the rounds this year is tidy and wrong: that the Food and Drug Administration created a new class for AI that diagnoses without a doctor.

4 minRead →
17

California's AI Endgame: 21 Bills, One Weekend, and a Governor's Last Signature

Twenty-one artificial intelligence bills are sitting on the third-reading files of the California Legislature this weekend, and the clock runs out Monday.

4 minRead →