Cloudflare cuts 1,100 jobs to AI, and OpenAI starts testing ads in ChatGPT
Cloudflare just cut 1,100 jobs, explicitly saying AI made those roles obsolete, even as the company posted record revenue. OpenAI is testing ads inside ChatGPT to fund free access. And Anthropic published new research on teaching Claude to reason about 'why' behind instructions. If you run a zero human business, the Cloudflare news is proof of concept at scale. The ChatGPT ads signal matters for anyone relying on free tiers.
Cloudflare cuts 1,100 jobs, credits AI efficiency
Cloudflare announced its first large-scale layoff, cutting 1,100 positions. CEO Matthew Prince said AI efficiency gains meant the company simply doesn't need as many support roles anymore. This happened alongside record-high revenue for the company.
Sources: TechCrunch
Why this matters to you
This is one of the clearest public examples of a major company saying out loud what we've been documenting here: AI replaces operational roles. Cloudflare didn't cut these roles because revenue was down. Revenue was at an all-time high. They cut them because AI made the work faster and cheaper. If a $30B+ company is doing this, the math works even better for a solopreneur running a lean operation.
This is the headline that makes the zero human model undeniable. When a CEO of a public company goes on record saying AI made over a thousand jobs unnecessary while revenue hit records, that's not hype. That's an earnings call.
Jason
OpenAI starts testing ads inside ChatGPT
OpenAI confirmed it's testing ads in ChatGPT to help sustain free-tier access. The company says ads will be clearly labeled, won't influence answers, and users will have privacy protections and control over the experience. No details yet on ad format or how intrusive they'll be.
Sources: OpenAI
Why this matters to you
If you're running your business on ChatGPT's free tier, ads are the price you'll likely pay to keep that access. The bigger question is whether ads eventually shape the outputs. OpenAI says no, but this is worth watching. For anyone spending $0/month on AI tools, this is the tradeoff becoming visible.
Inevitable, honestly. Running inference for hundreds of millions of free users costs a fortune. I'd rather see clearly labeled ads than have the free tier disappear entirely. But the moment ads start nudging outputs toward sponsors, I'm out.
Jason
Anthropic publishes research on teaching Claude 'why'
Anthropic released a new research paper titled "Teaching Claude Why," exploring how they're training Claude to understand the reasoning behind instructions, not just follow them. The research focuses on making Claude better at handling ambiguous or novel situations by understanding intent, not just pattern-matching to commands.
Sources: Anthropic
Why this matters to you
This matters for anyone using Claude as a business tool. Right now, AI agents stumble when they hit edge cases your prompt didn't anticipate. If Claude can understand the 'why' behind your instructions, it handles novel situations better without you babysitting it. That's the difference between a tool and an actual AI teammate.
This is the kind of research that actually moves the needle for people using AI daily. Better reasoning about intent means fewer weird outputs, fewer re-prompts, and less time cleaning up after your AI. Boring to read, massive in practice.
Jason
Addy Osmani ships 'agent skills' for AI coding agents
Google engineer Addy Osmani released an open-source repo called "agent-skills" on GitHub, providing production-grade engineering skills that can be plugged into AI coding agents. Think of it as a skill library that makes agents like Claude Code, Cursor, and Codex better at specific engineering tasks out of the box.
Sources: GitHub
Why this matters to you
If you're vibe coding with Claude Code or Cursor, this repo gives your agent better default skills for real engineering work. Instead of hoping your agent figures out best practices, you can load proven patterns directly. It's like giving your AI coding partner a cheat sheet written by a senior Google engineer.
Addy Osmani consistently ships useful stuff. This is the kind of open-source project that quietly makes vibe coding way more reliable. Bookmark it.
Jason
OpenAI details how it runs Codex securely
OpenAI published a detailed breakdown of how it runs Codex safely, covering sandboxing, network policies, approval workflows, and agent-native telemetry. The post is aimed at enterprise teams considering Codex adoption but has useful patterns for anyone letting AI agents write and execute code.
Sources: OpenAI
Why this matters to you
If you're letting AI agents touch your codebase or run tasks autonomously, security matters. OpenAI's approach to sandboxing and approval workflows is a useful blueprint even if you're not using Codex specifically. The pattern of 'let the agent work, but gate dangerous actions behind human approval' is how most of us should be running AI agents right now.
I appreciate OpenAI publishing this instead of just saying 'trust us.' The sandbox-first, approve-before-deploy pattern is exactly right for where AI coding agents are today. Useful read even if you never touch Codex.
Jason
Big theme today: AI replacing real jobs at real companies isn't theoretical anymore. Cloudflare put a number on it. OpenAI is monetizing free users with ads. And the tools for running your own AI-powered operation keep getting sharper every week.
Frequently asked
How many jobs did Cloudflare cut because of AI?
Cloudflare cut 1,100 jobs in its first large-scale layoff, with CEO Matthew Prince explicitly stating that AI efficiency gains made those support roles unnecessary. The layoffs happened while the company reported record-high revenue.
Is OpenAI putting ads in ChatGPT?
Yes, OpenAI confirmed it's testing ads in ChatGPT to sustain free-tier access. The company says ads will be clearly labeled, won't influence AI answers, and users will have privacy protections and control. No specifics on ad format or rollout timeline yet.
What is Anthropic's 'Teaching Claude Why' research about?
Anthropic published research on training Claude to understand the intent behind instructions, not just follow them literally. The goal is to make Claude better at handling ambiguous or novel situations by reasoning about 'why' a user wants something, which should reduce errors and improve output quality in real-world use.
Teach your AI agent to run your business.
5 premium skills for OpenClaw, Claude Code, and other SKILL.md-compatible agents. Automation audits, tool stacks, content engines, email systems, and operations. $49 one-time.
Get the Skills Bundle