AI Insider Weekly | Issue 001 | 10 May 2026
Today’s Headline
Companies are printing record profits. Then firing 20% of their staff. And blaming AI.
Cloudflare fired 1,100 people this week — its first mass layoff in 16 years — the same quarter it posted record revenue. PayPal is cutting 4,700 jobs. Upwork, the platform that connects companies to freelancers, laid off 24% of its own workforce. Coinbase: 14%. All in the same week. All citing AI.
Meanwhile, Anthropic’s secret weapon just scared the Pentagon enough to reopen blacklisted negotiations. The US government is now testing AI models before public release. And an open-source “zero-human company” hit 30,000 stars on GitHub in days.
Five stories. Eight minutes. Let’s go.
What We Cover Today
- The Great AI Layoff Wave — Cloudflare, PayPal, Upwork, Coinbase
- Anthropic’s Secret Model That Scared the Pentagon
- US Government Will Now Test AI Before Public Release
- The Open Source “Zero-Human Company” Taking Off
- Only 17.8% of the World’s Working Population Uses AI — India’s Opportunity
01. The Great AI Layoff Wave
Record profits. Record layoffs. Same week.
Quick Take
In a single week in May 2026, Cloudflare cut 1,100 jobs (20% of workforce), PayPal announced 4,760 cuts (20% over two years), Upwork slashed 24% of its team, Coinbase cut 14%, and payments firm BILL eliminated 30% of headcount. Every single CEO used a variation of the same sentence: we are restructuring for the AI era. Total tech layoffs in 2026 have now crossed 128,000 people.
What You Need to Know
The Cloudflare number is the one to watch. Their internal AI usage went up 600% in three months. Revenue hit a record high. Then they fired 1,100 people anyway — the first mass layoff in the company’s 16-year history. CEO Matthew Prince said it plainly: “There are roles at Cloudflare that are not the roles we need for the future.”
Upwork is the darkest irony. A platform built to connect companies to human freelancers just laid off a quarter of its own humans. Their press release said the cuts were to build “a more efficient operating model as the nature of work evolves.” Translation: AI agents are replacing the people who were supposed to connect you to work.
PayPal’s CEO was even more direct: “First, we will remove duplication. Second, we will accelerate AI adoption across our operations.”
The Pattern
This is not a cost-cutting story. Revenue is rising. This is a structural story. The companies doing best with AI are not using it to grow headcount — they are using it to grow output per person and then reducing the number of persons. The roles going first: support, operations, middle management, and anything that can be described as “process.”
What This Means for India
India has one of the largest concentrations of IT support, BPO, and operations talent in the world. These are precisely the roles disappearing first. The window to upskill is not three years away. It is now.
02. Anthropic’s Secret Model That Scared the Pentagon
Claude Mythos: the AI that finds security holes humans never found in decades.
Quick Take
Anthropic unveiled a model called Claude Mythos Preview — described as “far ahead” of any other model on cybersecurity. It can identify critical vulnerabilities in legacy software systems, including weaknesses in banking infrastructure and government networks that went undetected for decades. Anthropic restricted public access immediately and briefed US government officials directly. Dario Amodei met with the White House Chief of Staff. The Pentagon — which had blacklisted Anthropic just weeks earlier — reopened negotiations.
What You Need to Know
The reason Anthropic restricted Mythos is the same reason it is significant: it works both ways. A model that can find security holes can also show a hacker exactly where to attack. Anthropic launched a special initiative called Project Glasswing to limit access to approved organisations only.
The Pentagon situation is extraordinary. The Trump administration had declared Anthropic a “supply chain risk” — a label previously reserved for Chinese companies — after Anthropic refused to let the military use Claude for autonomous weapons without safety guardrails. Then Mythos landed, and suddenly both sides found reasons to talk again.
My Read
Mythos is the most important AI story of May that most people are not paying attention to. When an AI model is powerful enough to scare governments into changing geopolitical positions, we have left the product era and entered the infrastructure era. This is not a chatbot. It is a capability that changes what governments can and cannot do.
03. The US Government Will Now Test AI Before Public Release
The “move fast and break things” era for AI just ended.
Quick Take
The US Center for AI Standards and Innovation (CAISI) announced agreements with Google DeepMind, Microsoft, and xAI to evaluate AI models before public launch. The government will test for national security risks, cybersecurity implications, and public safety concerns — before the model ships. This builds on earlier agreements with OpenAI and Anthropic. The White House is also exploring an executive order to formalise the process.
What You Need to Know
This is the regulatory moment AI has been building toward. For five years, labs shipped first and let regulators figure it out later. That is no longer the deal. The Mythos situation — a model powerful enough to brief the White House before launch — is exactly what pushed this from discussion to action.
The EU AI Omnibus also reached final text agreement this week, with simplified compliance rules and extended deadlines. Two major jurisdictions moved toward harder regulatory frameworks in the same week.
My Read
For Indian businesses building on AI APIs — this matters. Compliance timelines are about to become a real variable in product planning. If you are selling into the US or EU market in 2027, budget for evaluation lead times. The free ride is ending.
04. The Open Source “Zero-Human Company” at 30,000 Stars
One person. One dashboard. Many AI agents. Under ₹1,700 a month.
Quick Take
A tool called Paperclip hit 30,000 GitHub stars in days. It is an open-source, self-hosted platform that lets you run teams of AI agents like a real company — assign high-level goals, agents figure out the how, you get the output. Full deployments are running for under $20 a month on cheap cloud hosting. MIT licensed. No vendor lock-in.
What You Need to Know
The cost structure is the story. Two years ago, running a multi-agent AI workflow required enterprise-level API spend. Now it costs less than a Swiggy order per day. The reason: model inference has become commodity. The “zero-human company” is no longer a thought experiment — solo operators are running content pipelines, research agents, and business workflows on this stack right now.
Do This Today
If you run a side project or small business: search “Paperclip AI GitHub” and spend one evening exploring what it can automate for you. The point is not to replace yourself. The point is to feel how cheap automation has become — and recalibrate what you think is worth your time.
05. Only 17.8% of the World Uses AI — India’s Real Opportunity
The gap between who knows and who doesn’t is still enormous.
Quick Take
Microsoft’s Global AI Diffusion Report for Q1 2026 shows global AI usage at 17.8% of the working-age population — up from 16.3% last quarter. The UAE leads at 70%. The US sits at 31.3%. India is not in the top tier. South Korea, Japan, and Thailand saw the biggest jumps this quarter, driven by improved AI capabilities in Asian languages.
What You Need to Know
The gap between AI adopters and non-adopters is not a skills gap — it is an awareness gap. Most people in India who could benefit from AI tools have never tried them seriously. Meanwhile, the layoff data above shows what happens to people and companies that wait too long.
The opportunity is not in being the best AI researcher. It is in being the most AI-fluent professional in your office, your city, your industry. That bar is still surprisingly low in most Indian sectors.
My Read
AIInsider exists for exactly this reason. The people who learn this in 2026 will not be catching up in 2027 — they will be teaching it. That gap between knowing and not knowing is your edge. Do not waste it.
The Editorial — One Thing to Take From This Week
Five stories this week. They look unconnected. They are not.
Companies are posting record profits and firing people in the same breath. Governments are testing AI before it ships. A tool that scared the Pentagon is restricted to a chosen few. Solo operators are running AI companies for ₹1,700 a month. And most of the world — including most of India — has still never seriously used any of this.
The pattern is the same in every story: the advantage is going to whoever moves first.
Not the biggest company. Not the most funded startup. The person or team that actually deploys AI into their real work, today, before the window closes.
Cloudflare’s CEO said something worth remembering: “The people embracing these tools are so much more productive than we have ever seen before.”
He then fired the people who weren’t.
Which side of that line you are on is still your choice. For now.
— AI Insider Team | aiinsider.in
Forward this to one person who needs to read it. That is how AI Insider grows.
