The IT sector lost 13,000 jobs in a single month. Unemployment in tech rose from 3.6% to 3.8%. The same week, Coinbase announced 700 layoffs while restructuring around what it called "AI-native operating models."

This is not a forecast anymore. It's a labor report.

Sam Altman warned this week about "AI-washing": companies blaming AI for cuts they would have made anyway. He's not wrong, but it's also not the full picture. AI is genuinely changing who gets hired and who gets let go. The people staying employed have a specific pattern: they use AI to do the work of multiple specialists at once.

This week's workflow shows you exactly how. 3 Claude sub-agents that replace what would normally take a $200,000 hire to do. Researcher. Editor. Analyst. Three markdown files. One folder. Set up in 20 minutes.

If you're worried about your job, the answer is not to compete against AI. The answer is to run a team of AI specialists from your laptop.

The Old Way:

You hire a research analyst, an editor, and a data analyst. Combined salary: about $200,000 a year. They sit in three different roles, on three different teams, with three different managers.

You need a competitive briefing before Friday's meeting. You email the analyst. They get back to you in 4 days with a 12-page document you don't have time to read.

You finished a draft last night. You send it to the editor. They get to it next Tuesday because they have 14 other drafts ahead of yours.

You need to know what last quarter's sales numbers actually say. You request a report from the analyst. They send back a 40-row spreadsheet and tell you to "look at row 23."

None of this is anyone's fault. Specialists are slow because specialists have queues. Their work is valuable. It just doesn't move at the speed you need it to.

That's the old way. Three people. Three queues. Three weeks before you have the inputs to make Friday's decision.

The Replacement:

3 sub-agents. Each one is a markdown file. Each one has a defined role, its own context window, and a specific output format. Claude can call any of them on demand without polluting your main thread.

A sub-agent is not magic. It's a job description in a markdown file. The structure looks like this:

You save it as a .md file. Claude finds it automatically and calls it when the task matches.

The shift is from "Claude is my assistant" to "Claude is my team."

The 3 sub-agents in this issue handle the three most common knowledge-work tasks: research, editing, and data analysis. The combined cost of hiring all three roles is roughly $200,000 per year (Research Analyst at $90K, Senior Editor at $85K, Data Analyst at $120K, source: BLS Occupational Employment Statistics 2026).

The combined cost of running these three sub-agents is whatever you already pay for Claude.

Setup time: 20 minutes for all three. Time saved per week: 8-15 hours depending on how much research, writing, or analysis your role involves.

A regular prompt is a one-time instruction. You write it. Claude responds. You move on.

A sub-agent is a persistent specialist. You define it once. Claude calls it whenever the task matches. The behavior stays consistent across sessions, devices, and projects.

This matters because the value of AI compounds. The first time you use the Researcher, it saves you an hour. The hundredth time, you've reclaimed weeks. The pattern is the same for the Editor and the Analyst.

Anthropic published its sub-agent documentation in late 2025. Most Claude users still haven't read it. The ones who have are operating at a different level of productivity than the ones who haven't.

This week's workflow closes that gap in 20 minutes.

Each one is standalone. Install one, two, or all three based on your work.

The Researcher: Pulls primary sources, returns structured briefs, flags contradictions. Use when you need a real briefing before a meeting, pitch, or decision.

The Editor: Cuts drafts by 30%, rewrites hooks, flags weak claims. Use when you've finished a draft and you know it's bloated but you're too close to see it.

The Analyst: Reads any data and returns the headline finding plus the one chart you should build. Use when you've exported a report and don't know what to say in the meeting.

The 3 Sub-Agent Pack: The Researcher pulls primary sources and returns briefs in 5 minutes that would take an analyst a full day. The Editor cuts your drafts by 30% and flags weak claims. The Analyst reads any data and returns the chart you should actually build. → 3 complete markdown files copy-paste ready, plus install instructions for Claude Code, Claude.ai, and Claude Desktop

The Big Story Breakdown: Which 4 IT job categories are getting hit first, which 3 skills are getting hired more right now, and the exact action to take in your role this month. → Full analysis with specific implications and one decision to make this week

Monthly resource: The MCP Server Cheatsheet: Every MCP server worth installing in Claude and ChatGPT. What each does, install instructions, and 3 prompts that demonstrate value. Updated monthly.

🔒 If you want all 3 sub-agents running by Friday afternoon, the complete setup is below. Each one is copy-paste ready. Start your free 7-day trial to unlock the full pack.

Anthropic announced a new agent capability called Dreaming on May 6 at their Code with Claude event. It lets autonomous agents review their own past sessions, identify patterns in what worked and what failed, and apply those lessons to future runs without human intervention.

Why this week: Currently in research preview, but it changes what "agent" means. Until now, every agent session started fresh. The agent might be brilliant in one run and incompetent in the next because it had no memory of what worked before. Dreaming fixes this. The agent literally improves between sessions, the way an employee gets better at a job over time.

For most readers, this is a "watch and wait" tool rather than a "deploy this Friday" one. It signals where agent infrastructure is going. The companies using AI as static assistants today will be competing against companies using AI that gets better every week.

🔒 Implementation checklist is for pro members. Get a breakdown plan to implement AI into your workflow.

📉 IT sector loses 13,000 jobs in April. US Department of Labor data shows the IT unemployment rate rose from 3.6% in March to 3.8% in April. The sector shed 13,000 jobs in a single month amid widespread AI restructuring. Wall Street Journal analysis covered the pattern. Why this matters for you: This is the first month where the labor statistics confirm what tech Twitter has been saying for six months. If you work in IT or hire IT staff, this changes the conversation about headcount. (Wall Street Journal)

💰 OpenAI raises $122B at $852B valuation. The largest funding round in tech history. Anchored by Amazon, Nvidia, SoftBank, and Microsoft. Same week, Anthropic took an additional $40B from Google and $5B from Amazon, packaged with $100B of AWS spend, plus chip deals reportedly worth hundreds of billions. Why this matters for you: AI is no longer "an emerging technology." It's infrastructure on the scale of telecoms or energy. The companies that integrate it into core operations now will compete on a different cost curve than the ones that wait. (OpenAI)

🐉 4 Chinese open-source models released in 12 days. Z.ai's GLM-5.1, MiniMax M2.7, Moonshot's Kimi K2.6, and DeepSeek V4 all landed at roughly the same capability ceiling on agentic engineering as Claude Opus 4.7 and GPT-5.5. None costs more than a third of the Western frontier per token. Zhipu's stock closed up 15.92% on GLM-5.1 launch day. Why this matters for you: The "Claude vs ChatGPT" framing your team uses to evaluate AI tools is missing two-thirds of the options. The price-performance curve on these Chinese models is significantly better for high-volume agentic work. Worth a vendor review this quarter. (Abhs)

On May 7, the Bureau of Labor Statistics released April employment data. The IT sector lost 13,000 jobs. Unemployment in the sector rose from 3.6% to 3.8%. The Wall Street Journal connected the move to ongoing AI restructuring across major tech companies.

The same week, Coinbase announced 700 layoffs while explicitly restructuring around "AI-native operating models." Earlier in 2026, Snap had cut 1,000 jobs citing AI productivity gains. Oracle eliminated 20,000-30,000 positions. Block cut 40% of its workforce.

Sam Altman, the CEO of OpenAI, warned this week about companies "AI-washing" layoffs. He's not wrong. Some of these cuts would have happened regardless of AI. But the pattern across companies, industries, and roles is consistent enough that the labor statistics are now confirming what everyone has been arguing about on social media.

This is the first month where the data caught up with the narrative.

The cuts are not evenly distributed across IT. Three patterns are visible in the public layoff announcements:

Pattern 1: Companies are cutting middle layers (mid-level engineers, mid-level analysts, mid-level project managers) while keeping junior hires (cheap labor that learns AI faster) and senior staff (architectural judgment that AI can't replicate yet).

Pattern 2: Companies are restructuring teams around AI deployment rather than cutting overall headcount. The same number of people, doing different work.

Pattern 3: Companies are hiring for AI-specific roles at premium rates while letting traditional roles attrition out. Job postings for "AI Operations" and similar titles are up 340% year-over-year.

The honest read on the data: AI is not replacing IT workers wholesale. It's reshaping which IT workers get hired, at what level, and for what work.

The question your readers should be asking is not "will AI take my job?" The question is "which of these 3 patterns is my employer following, and what does that mean for my next 12 months?"

🔒 The full breakdown (the 4 job categories getting hit first, the 3 skills hiring is shifting toward right now, and the specific action to take in your role this month) is for paid members. Start your free 7-day trial.

New This Week:

Prompt Library Updates:

Resources This Month:

The labor data caught up to the narrative this month. 13,000 IT jobs gone in April. The people who keep their jobs through the rest of 2026 are not the ones competing against AI. They are the ones running it.

If you do one thing before next Thursday, install at least one of the three sub-agents. The Editor takes five minutes. The first time you use it on a real draft, you'll understand why this is the workflow shift that matters.

Every Thursday you get:

A complete deployment system with copy-paste prompts and setup steps (not a tutorial, an actual implementation). A breakdown of the week's biggest story with specific implications for your work and one action to take this month. A new monthly resource that lives in your library forever.

Pro members deploy AI in their work an average of 3-5x more often than free readers (based on reply data from past issues). The difference is having the exact setup, not the concept.

Your toolkit (built over the last 3 issues):

MCP Server Cheatsheet (Monthly resource): Every major MCP server worth installing in Claude Desktop. What each does, install instructions, and 3 prompts that demonstrate value. Updated monthly with new MCP releases.

3-GPT Setup Pack: Brand Voice GPT, Company Knowledge GPT, and Competitor Intelligence GPT. Complete setups with custom instructions, knowledge file lists, conversation starters, and troubleshooting.

AI Operations Toolkit: 5 skills for deploying AI in any business: Revenue Leak Finder, Email Auto-Responder, Quote Generator, Lead Qualifier, Weekly Snapshot. Includes guides for using internally at your job, in your own business, or as freelance services.

Claude Design Starter Kit: 5 standalone prompts for pitch decks, landing pages, dashboards, one-pagers, and design systems using Claude Design and Cowork Live Artifacts.

AI Job Exposure Audit: Notion guide that assesses your specific role's automation risk based on the Anthropic methodology that mapped 1,100+ occupations. Especially relevant given this week's IT layoffs data.

Claude Can Now Use Your Computer: Notion guide to computer use and Dispatch setup.

Prompt Library: 60+ workflows across writing, content, analysis, sales, operations, and creative work. Updated weekly with new additions from each issue.

Each resource lives permanently in your Pro account. Use them whenever you need them.

🎉 Try free for 7 days. Limited time.