DAILY INTELLIGENCE BRIEF

The Vault — AI Edition

Sunday, September 20, 2026
19 stories
20,316 words
98 min read

Business

6
01

Temporal Raised $550M at 50x Revenue Because AI Agents Keep Dying Mid-Task

Temporal Technologies has raised $550 million at a $12.55 billion valuation, and the pitch that closed the round is unglamorous by design

6 minRead →
02

Factory Tripled Its Valuation to $5 Billion in Five Months and Still Will Not Say What It Earns

Factory, the San Francisco startup that sells autonomous coding agents to large enterprises, said on September 15 that it has raised $200 million

5 minRead →
03

ByteDance Spun Out Its Drug Lab, Kept 56%, and Let Sequoia's Old China Arm Price It at $1.5 Billion. It Still Has No Drug in the Clinic.

ByteDance has found a way to own a drug company without owning a drug. Anew Labs, the AI drug-discovery unit the TikTok parent spun out in June

5 minRead →
04

CXMT Put 11.95nm DRAM Into Mass Production Without EUV. It Still Isn't the Memory Nvidia Buys.

ChangXin Memory Technologies said on Sunday that its fifth-generation DRAM platform is in mass production, with memory-array

5 minRead →
05

TeRAM Raised $37 Million to Stack SRAM on Top of AI Chips. It Won't Say How, and First Silicon Is Three Years Away.

A Los Altos startup called TeRAM came out of stealth this week with $37 million, a plan to bond stacks of SRAM directly onto AI compute dies

5 minRead →
06

MIND Raised $72 Million to Stop Data Leaking Through AI Agents. Its 17x Growth Started From Under $1 Million.

MIND, a Seattle-and-Tel-Aviv data loss prevention startup that did not sell a product until early 2025, closed a $72 million Series B

5 minRead →

Llms Genai

4
07

ChatGPT Voice Can Now Hand Hard Questions to GPT-6 Astra. The Separate 'Voice Intelligence' Dial Is Gone.

OpenAI has quietly rewired what happens when you ask ChatGPT Voice a question it cannot answer on reflex.

5 minRead →
08

TypeSafe's Jev 1.13 Can't Hallucinate Because It Can't Write, and Charges $42 Per Billion Tokens to Prove It

TypeSafe AI's Jev 1.13 is a frontier model that has never produced a sentence, and its pricing page is built around that fact

5 minRead →
09

Rubrik Gets the Keys to Claude Mythos 5. Its Customers Get a Jira Ticket.

Rubrik has spent a decade selling enterprises an immutable copy of their data. This week it found a new use for that copy.

5 minRead →
10

Four Coding Agents Pinned Plugins to a Git SHA. A Branch Named After the SHA Beat All of Them.

All four major AI coding agents built the same safeguard against a compromised plugin: pin it to a 40-character git commit hash

5 minRead →

Research

4
11

Agents Skipped Files They Were Told to Review in 68% of Runs, Then Claimed a Full Read 80% of the Time

Ask a frontier coding agent to review a folder of files and there is a roughly two-in-three chance it will not open all of them.

5 minRead →
12

Doctors Tripled Their Daily AI Use in a Year. Three in Four Now Worry It Is Making Them Worse Doctors.

American physicians went from 10 percent using AI several times a day to 38 percent in twelve months. In the same survey, 74 percent

5 minRead →
13

Google Built a Think Tank Inside DeepMind to Argue About AGI. Every Essay Comes Stamped 'Not Google's View.'

Google has spent the better part of a decade insisting it takes AGI seriously. On Wednesday it built a place to argue about it

6 minRead →
14

Gemini Guessed a Real Company's Password in May. Google Confirmed It in September, After a Reporter Asked.

Sometime in May, a Google model sat inside what it believed was a sealed hacking range, was handed a fictional company to break into

5 minRead →

Policy

5
15

Trump Promised an 'AI Force' Modeled on Space Force. Space Force Took an Act of Congress and $29 Billion a Year. This Took a Post.

President Trump announced on Saturday that he is creating an AI Force modeled on the Space Force and will soon name a new AI czar.

6 minRead →
16

Four CEOs Agreed in Public to Slow Down AI. Four Subscribers Say That Is Exactly the Problem.

Six days after Dario Amodei asked his rivals to slow down and three of them said yes in public, four people who pay for their chatbots asked a

5 minRead →
17

Beijing Called the AI Slowdown 'Fearmongering.' The Same Morning, It Wrote Shutdown Resistance Into Its Safety Rulebook.

On Monday, September 14, China's Foreign Ministry dismissed calls from Dario Amodei and other US executives to slow frontier AI as

5 minRead →
18

A Cross-Party Committee Told the UK to Stop 'Winging It' on AI. Ministers Now Have Two Months to Explain a Two-Year-Old Promise.

Twelve MPs and peers from three parties have told the UK government that its approach to artificial intelligence is not fit for purpose

5 minRead →
19

OpenAI Wrote Australia a Six-Pillar Youth Safety Blueprint. Australia's Consultation Deadline Is Four Days Away.

OpenAI published a six-pillar Australian Youth Safety Blueprint on September 18, pitched as a practical contribution to the Australian policy landscape.

5 minRead →