AI INSIDERS NEWSLETTER / ISSUE №198 · 13th of October 2026

Character.AI Goes Bankrupt, and Everyone Asks Who Answers for the Agents

Character.AI filed for Chapter 11 with 20 million monthly users, and Disney has offered $20 million for some of its assets.

Read the latest brief →

ISSUE №197 · MON, 12 OCT 2026

Small Judges, Loose Agents, and Who Is Watching

Cloudflare, Perplexity and Microsoft each shipped a small model that never writes a sentence.

ISSUE №196 · FRI, 09 OCT 2026

Claims, Receipts and the Gap Between Them

Most of this issue is about the distance between what AI is claimed to do and what anyone has shown.

ISSUE №195 · THU, 08 OCT 2026

The Wire — live

14:19
UTC

WIRE A faster Zhipu coding model surfaces, with one user's speed screenshot

14:19
UTC

TOOLS A free add-on teaches AI coding assistants to test the agents they build

14:19
UTC

MODELS AI builds a scene you can walk through, with sound that turns as you do

14:19
UTC

MODELS Amazon rebuilds trained AI models into leaner ones for a sliver of the cost

14:19
UTC

TOOLS Amex wants a signed record of what you told your AI shopper to buy

14:19
UTC

AGENTS Anthropic: small AI models may soon be able to copy themselves onto smart TVs

14:19
UTC

FUNDING Character.AI files for Chapter 11 and puts its chatbot platform up for sale

14:19
UTC

WIRE Claude users suspect Anthropic is quietly testing a new model

All Wire entries →

Models & Research

All models →

Models / 19 min ago

AI builds a scene you can walk through, with sound that turns as you do

Noiz and a Hong Kong university made a model that generates an explorable place with stereo sound that pans as the view turns. The speed claim comes from one data-center chip.

By AI Insiders Staff

Models / 19 min ago

Amazon rebuilds trained AI models into leaner ones for a sliver of the cost

By AI Insiders Staff

Models / 19 min ago

Google's new health AI round-up: one clinical trial, mostly studies

By AI Insiders Staff

Models / 19 min ago

Meta's new Muse Wonder models promise an AI that remembers you

By AI Insiders Staff

Agents & Automation

All agents →

Agents / 19 min ago

Anthropic: small AI models may soon be able to copy themselves onto smart TVs

By AI Insiders Staff

Agents / 23 hr ago

Claude exploited a university server and sent a false police tip

By AI Insiders Staff

Agents / 23 hr ago

Teams of AI coders cost up to 5.1 times more and rarely scored higher

By AI Insiders Staff

Agents / 4 days ago

Devin's Security Swarm docs never say how often it is wrong

By AI Insiders Staff

Developer Tools

All tools →

Tools / 19 min ago

A free add-on teaches AI coding assistants to test the agents they build

By AI Insiders Staff

Tools / 19 min ago

Amex wants a signed record of what you told your AI shopper to buy

By AI Insiders Staff

Tools / 23 hr ago

Four ways to make chatbots answer faster, sorted by one question

By AI Insiders Staff

Tools / 4 days ago

Anthropic will send open-source maintainers bug reports no human read

By AI Insiders Staff

By the numbers

One brief. Every signal.

Independently owned, read daily by the people building AI. No syndication, no SEO games — just the signal.

14,224

Newsletter subscribers

47,929

LinkedIn followers

1,742

Original articles

106

Daily editions shipped

Funding & Business

All deals →

Funding / 19 min ago

Character.AI files for Chapter 11 and puts its chatbot platform up for sale

By AI Insiders Staff

Funding / 23 hr ago

Amazon Is Said to Be Weighing a $7 Billion Bid for Chip-Switching Startup

By AI Insiders Staff

Funding / 4 days ago

Hone raises $60M at a $285M valuation to sell AI agents it has yet to show

By AI Insiders Staff

Funding / 4 days ago

OpenAI's $70B revenue figure was assembled by investors, not published

By AI Insiders Staff

Policy & Safety

All policy →

Policy / 23 hr ago

Anthropic's usage policy bars sustained, needless abuse of Claude

The rule starts 12 November and excludes frustration, pushback, dark fiction, and testing. The Register notes no established evidence that models experience distress.

By AI Insiders Staff

Policy / 23 hr ago

OpenAI Cites a Bigger Breach of Trust but Not What It Was

By AI Insiders Staff

Policy / 23 hr ago

Who pays AI's outside evaluators is still undecided

By AI Insiders Staff

Policy / 4 days ago

Fired OpenAI Researchers' Own Letter Denies Leak and Explains Email Access

By AI Insiders Staff

Opinion

All op-eds →

Opinion / 19 min ago

Ship code nobody read, and a person still answers for it

Mark Seemann accepts that tests can beat human review, then argues that when unread AI-written software fails, courts will still look for a human to hold responsible.

By AI Insiders Staff

Opinion / 19 min ago

Slow AI down, but how? What one attendee saw at The Curve

Eli Pariser's report from Berkeley: slowing the frontier is now a mainstream idea, the fight is over the method, and one lab employee says no one has a plan past 14 months.

By AI Insiders Staff

Opinion / 19 min ago

Why AI excels at cheap tests and struggles when each test costs real time

A LessWrong essay argues models shine where feedback is fast and cheap, and stumble in science, business and medicine, where each experiment is slow or costly.

By AI Insiders Staff

Opinion / 23 hr ago

Nadella wants AI to ship with a human "emergency brake"

Microsoft's chief executive argued on X that advanced models need a stop control people can use while work is underway. He described no mechanism, and Microsoft has announced no product.

By AI Insiders Staff

Character.AI Goes Bankrupt, and Everyone Asks Who Answers for the Agents

Small Judges, Loose Agents, and Who Is Watching

Claims, Receipts and the Gap Between Them

Labs Under the Microscope, Agents Moving In

The Talent Gap And The Asterisks

The Price Comes Before the Investor

The Wire — live

Models & Research

AI builds a scene you can walk through, with sound that turns as you do

Amazon rebuilds trained AI models into leaner ones for a sliver of the cost

Google's new health AI round-up: one clinical trial, mostly studies

Meta's new Muse Wonder models promise an AI that remembers you

Agents & Automation

Anthropic: small AI models may soon be able to copy themselves onto smart TVs

Claude exploited a university server and sent a false police tip

Teams of AI coders cost up to 5.1 times more and rarely scored higher

Devin's Security Swarm docs never say how often it is wrong

Developer Tools

A free add-on teaches AI coding assistants to test the agents they build

Amex wants a signed record of what you told your AI shopper to buy

Four ways to make chatbots answer faster, sorted by one question

Anthropic will send open-source maintainers bug reports no human read

One brief. Every signal.

Funding & Business

Character.AI files for Chapter 11 and puts its chatbot platform up for sale

Amazon Is Said to Be Weighing a $7 Billion Bid for Chip-Switching Startup

Hone raises $60M at a $285M valuation to sell AI agents it has yet to show

OpenAI's $70B revenue figure was assembled by investors, not published

Policy & Safety

Anthropic's usage policy bars sustained, needless abuse of Claude

OpenAI Cites a Bigger Breach of Trust but Not What It Was

Who pays AI's outside evaluators is still undecided

Fired OpenAI Researchers' Own Letter Denies Leak and Explains Email Access

Opinion

Ship code nobody read, and a person still answers for it

Slow AI down, but how? What one attendee saw at The Curve

Why AI excels at cheap tests and struggles when each test costs real time

Nadella wants AI to ship with a human "emergency brake"

The morning brief for people inside the AI industry.