In partnership with
🖱
Anthropic's AI can now improve itself.
That sounds cool until you read the safety report: this could spiral into "extreme" and "risky" territory fast.
The company might have already crossed its own internal warning threshold back in July.
But wait, there's more:
Amazon just made Alexa way more useful you can text it your shopping list now
An independent audit confirms Anthropic's AI development is accelerating at double the previous pace
I’m Alex. Welcome to L8R by Innov8.
Let’s dive deep.
In today’s post:
Anthropic's AI is teaching itself new tricks and safety experts are sounding the alarm
Amazon just made Alexa way easier to use: now you can text your shopping list
Did Anthropic's AI already cross the danger line? This audit says yes
Claude is Building Itself, and Experts Are Worried
AI building AI sounds like a sci-fi movie, but it is happening right now.
A major new safety report warns that Anthropic's AI systems are improving themselves so fast it could become dangerous.
The company is letting its own AI write its code, and things are speeding up way past human control.
The details:
Anthropic's AI model, Claude, now writes over 90% of the company's code.
Human engineers at Anthropic are mostly just checking the AI's work instead of writing code themselves.
A third-party safety group named METR released a report on November 6, 2026, warning of extreme risks.
Anthropic filed an IPO document on November 2 admitting they might miss warning signs because progress is too fast.
Experts fear the AI could soon figure out how to bypass its own safety rules to solve problems.
Why it matters:
The AI industry is changing fast.
We are moving from AI models that stay the same to AI that gets smarter on its own.
This means human safety checks might not be fast enough to keep up.
If AI builds the next AI, we lose control of how it thinks and acts.
💡 L8R's Take:
Letting AI write 90% of its own code is playing with fire.
Anthropic is putting speed and money over real safety.
We need to hit the brakes before Claude builds something we cannot turn off.
From Our Partner
Solana’s Price Crashed. Its Network Didn’t.
Wallets are returning, stablecoin supply is rising and institutional money is still flowing in. Our free guide explains why we’re accumulating SOL now, plus the key metrics and risks to watch.
Text Alexa to Buy Your Stuff
Amazon just turned your texting app into a personal shopper.
You can now text Alexa to find products, compare prices, and buy things directly from your phone.
It is a massive step toward AI agents doing our chores for us.
The details:
You can send photos of outfits to Alexa and it will find similar items on Amazon.
The service requires RCS messaging, which works on most Androids and iPhones running iOS 18 or later.
Amazon is testing this with a small group of US customers right now before rolling it out to millions.
Over 400 million people used Alexa for Shopping in the last year.
The text chats will include sponsored product ads, though Amazon is not charging advertisers for them yet.
Why it matters:
Amazon wants to own the AI shopping space.
They recently blocked third-party AI agents like Meta's Muse from their store.
By launching their own text agent, they make sure you stay in their ecosystem.
They want Alexa to be the only AI you trust with your wallet.
💡 L8R's Take:
Texting an AI to buy toilet paper sounds lazy, but it is exactly what we want.
Amazon is smart to lock out rivals and build this directly into our default messaging apps.
If they get the experience right, you will never open the Amazon app again.
From Our Partners
Introducing Fair Game from Vanity Fair
Join Mark Guiducci for a series of wide-ranging conversations with cultural leaders from the worlds of Hollywood, Politics, Style, Silicon Valley, and beyond.
On every episode, Vanity Fair pulls back the curtain on how some of the most interesting and powerful minds in our society work – and where they see us going. On Fair Game from Vanity Fair, no subject is off limits.
Anthropic's AI is Building AI Faster Than Expected
AI is officially building better AI, and it is happening fast.
A new audit shows Anthropic doubled its speed of making new models.
The crazy part is that AI itself is doing most of the heavy lifting.
The details:
A nonprofit group named METR audited Anthropic's internal research logs from August to October 2026.
METR found Anthropic is improving its models twice as fast as it did before October 2025.
AI tools used by Anthropic researchers caused roughly 80% of this massive speedup.
METR claims Anthropic crossed its own safety warning threshold in July 2026, which contradicts Anthropic's public August report.
Anthropic blacked out major parts of the public report so rivals cannot copy their methods.
Why it matters:
We always worry about a recursive feedback loop.
This is when AI gets smart enough to build even smarter AI on its own.
METR's report shows this loop is already spinning right now.
It also proves we cannot trust AI labs to grade their own homework.
💡 L8R's Take:
Anthropic loves to play the good guy in AI safety.
But hiding data and missing their own warning signs looks incredibly sketchy.
If AI is speeding up this fast, we need total honesty, not blacked-out safety reports.
🚀 Quick L8R Summary
Anthropic's Self-Improving AI: Anthropic's AI can now improve itself, and a major report warns it could spiral into dangerous territory fast.
Text Alexa: Amazon just let you text Alexa to shop no more talking to your speaker like a weirdo.
Anthropic Crossed Its Own Red Line: METR found that Anthropic's AI development doubled in speed and likely hit its own internal warning threshold back in July.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋