AI @ Work
Friday, May 22, 2026
Issue #007
This week: a blunt, practical case for why your current job is probably expiring — and why that might be the best news you’ve heard all year. Google rewrites Search in its biggest overhaul in 25 years. We now know the exact price Anthropic is paying SpaceX for compute. And Perplexity — the search tool you may have dismissed — turns out to be building something far more ambitious than search. All translated into what it means for your work.
★ This Week’s Big Story
Why this matters to you
You’ll probably lose your job in 2027. That’s actually good news — if you start preparing today.
Elena Verna, one of the sharpest growth strategists in tech, published a piece this week via Lenny’s Newsletter with a disarmingly direct title: “You’ll lose your job in 2027.”
Her argument isn’t that AI will necessarily eliminate your role — it’s that your current role is already approaching its expiration date, and the only thing worse than knowing that is pretending it isn’t happening. Intuit just cut 17% of its staff. Meta laid off employees it had previously used as AI training data. And even people who keep their seats are discovering that the responsibilities, tools, scope, and bar for performance are being rewritten continuously.
The uncomfortable truth she surfaces: the “stability” of full-time employment has always been an illusion — AI is just making the illusion harder to maintain. The productive response isn’t fear. It’s optionality. Start building career alternatives before you need them: freelance work, a small software product, a consulting niche, anything that gives you an income stream that isn’t tied to one company’s quarterly decisions. And get genuinely AI-fluent — not by having an LLM open in your browser, but by actually changing how you think about and approach your work. Next week in Tools & Workflow, we’ll break down her six-tier AI fluency framework with the specific prompts she uses at each level. | HR & People | Executive & Strategy | All Readers |
Source: Lenny’s Newsletter (cross-post from Elena’s Growth Scoop), May 21, 2026
AI Platforms
Claude / Anthropic
In Issue #005, we told you Anthropic signed a massive compute deal with SpaceX. This week we found out the price — and it reframes the entire AI infrastructure race.
Anthropic Is Paying SpaceX $1.25 Billion Per Month. That’s $45 Billion Over Three Years.
The number is now confirmed: Anthropic has committed $1.25 billion per month to SpaceX for access to the Colossus 1 data center in Texas — $45 billion over three years. To put that in context, $45 billion is more than the GDP of over 80 countries. It is the single largest compute commitment in AI history. What this tells you about the industry: the companies building the most capable AI are now making multi-decade infrastructure bets at a scale that makes most corporate capital expenditures look like rounding errors. The cost of staying competitive at the frontier of AI is not measured in millions. It is measured in the GDP of small nations. | Executive & Strategy | Finance | IT & Security |
Source: Techpresso, May 21, 2026
Gemini / Google
If you use Google Search every day — and you do — your search experience is about to change more than it has in the past quarter century.
Google’s Biggest Search Upgrade in 25 Years: Agents That Work in the Background While You Do Something Else
Google announced at I/O this week what it called the biggest upgrade to Search in 25 years. The search box now accepts text, images, files, videos, and Chrome tabs simultaneously. But the more significant shift is the agent layer. Information agents will operate in the background, monitoring the web for things you care about — apartments that match your criteria, competitors who change their pricing, news about topics you follow — and send you updates as they find them. Agentic booking and agentic coding are also coming to Search this summer. Personal Intelligence, which connects Search to your Gmail, Google Photos, and Calendar for tailored responses, is expanding to 200 countries across 98 languages. Google’s advantage in this race is distribution: billions of people already use Search every day. The risk is that AI-generated answers are harder to fact-check than a list of blue links. | Operations | Marketing | All Readers |
Source: Techpresso, May 20, 2026
ChatGPT / OpenAI
For anyone tracking OpenAI’s business — the company may be going public sooner than most people expected.
OpenAI’s IPO Could Come as Early as September, According to the Wall Street Journal
The Wall Street Journal reported this week that OpenAI could file for its IPO as early as September 2026. OpenAI is currently valued at roughly $300 billion, making it one of the most valuable private companies ever to go public. A successful IPO would give OpenAI access to public capital markets at the precise moment when AI infrastructure spending is accelerating fastest. For enterprise customers, an IPO also signals stability and long-term commitment. It would also put OpenAI’s finances under public scrutiny for the first time, which means we’d finally see exactly how much the company spends to run ChatGPT. | Executive & Strategy | Finance |
Source: The Deep View · Wall Street Journal, May 21, 2026
Microsoft
If your organization is building or deploying AI agents, Microsoft just gave your security team two free tools they’ll want to know about.
Microsoft Open-Sources RAMPART and Clarity — Two New Security Tools Built Specifically for AI Agents
Microsoft’s AI red team released two open-source security tools this week designed for organizations building AI agents. RAMPART embeds red-teaming techniques directly into the development workflow, letting teams continuously test how agents behave under normal and adversarial conditions. Clarity helps engineers reason through what they’re building before they write a single line of code — pressure-testing ideas and surfacing risks while they’re still cheap to fix. By open-sourcing both, Microsoft is making the argument that AI security needs to be an industry-wide discipline. As one Microsoft AI red team leader put it: the shift from AI that generates text to AI that does things in the world changes the safety equation entirely. | IT & Security | Operations | Executive & Strategy |
Source: The Deep View, May 21, 2026
Perplexity
Most people think of Perplexity as an AI search engine. That’s no longer the whole story — and next Wednesday’s deep-dive will tell it in full.
Perplexity Computer Now Connects to Snowflake. Here’s Why That Matters — and What Perplexity Computer Actually Is.
Perplexity announced this week that its Computer agent now connects to Snowflake, the enterprise data warehouse platform — meaning Perplexity Computer can now query your organization’s internal data, not just the public web. But to understand why that’s significant, you need to know what Perplexity Computer actually is. Launched in February, Computer is not a chatbot. It is a multi-model orchestration system that takes a high-level goal, breaks it into tasks, spins up specialized sub-agents to execute each one, and delivers a completed result — often after running for hours in the background. It uses Claude Opus as its core reasoning engine and routes subtasks to Gemini, Grok, and ChatGPT depending on what each does best. The Snowflake integration extends that orchestration into enterprise data infrastructure. Next Wednesday, we go deep on the full Perplexity story. | IT & Security | Finance | Operations |
Source: Perplexity Hub, May 2026
Tools & Workflow — Try These This Week
For anyone who uses AI search tools to find information — this changes how you should read the results.
AI Search Can Be Manipulated in 24 Hours. A BBC Journalist Proved It. Here’s What You Should Do About It.
A BBC investigation published this week found that AI chatbots — including ChatGPT, Gemini, Claude, and Google’s AI Overviews — can be manipulated into presenting false information simply by publishing a convincing-looking online article. BBC journalist Thomas Germain published a fake article claiming he was a world-champion hot-dog eater. Within 24 hours, multiple major AI tools were repeating the claim as fact. The same technique has reportedly been used around health supplements, financial advice, and retirement planning — areas where a wrong answer can cause real harm. The practical rule: treat AI search answers as a starting point, not a final answer. For anything that matters — health, money, legal, personnel decisions — verify with a primary source before acting. | All Readers | Legal & Compliance | IT & Security |
Source: Mindstream, May 21, 2026 · BBC Future
For anyone who spends hours each week on tasks that feel like they shouldn’t take that long.
Seven Prompts That Kill the Busywork Loop: Stop Managing Tasks and Start Finishing Them
AI Business Insights published a useful framework this week built around a single premise: most AI prompts ask for help thinking, when what you actually need is help finishing. The seven prompts cover the tasks that consume the most unproductive time in a typical workday — drafting the follow-up you’ve been putting off, summarizing the document you don’t have time to read, converting meeting notes into action items, generating a first draft of the report that’s been sitting blank, writing the job posting that takes three hours instead of twenty minutes, turning a data set into plain language for a non-technical audience, and creating a structured agenda from a disorganized list of talking points. None of these require special AI skills. They require stopping and asking. | Operations | HR & People | All Readers |
Source: AI Business Insights, May 21, 2026
Worth Watching
GitHub confirmed a breach of approximately 3,800 internal code repositories — traced to a poisoned VS Code extension that infected an employee’s device. A developer tool that millions of engineers use every day was turned into a delivery mechanism for an intrusion. If your team uses VS Code extensions, now is a good time to audit which ones are installed and where they came from. | IT & Security |
Intuit is cutting 17% of its global workforce and citing AI as the reason. The maker of TurboTax and QuickBooks says AI tools have made certain roles redundant. Intuit is not a startup — it is a 40-year-old enterprise software company. When companies like Intuit cite AI efficiency as a restructuring rationale, the trend is past the early-adopter stage. | HR & People | Executive & Strategy |
Nvidia reported quarterly revenue of $82 billion — up 85% year over year. Nvidia’s revenue is essentially a proxy for how much money the AI industry is spending on compute infrastructure right now. $82 billion in one quarter means the spending is not slowing down. | Finance | Executive & Strategy |
Meta used its own employees as AI training data — then laid them off. Meta trained AI systems on work produced by staff who were subsequently cut in a round of layoffs. Who owns the outputs of an employee’s work, and can a company train AI on those outputs and then eliminate the role? Expect this to become a more common labor dispute as AI adoption accelerates. | HR & People | Legal & Compliance |
Published by Independent Intelligence — an independent AI newsletter for people who use AI at work, not just read about it. Five newsletters, one Substack. Friday general edition + Wednesday deep-dives on Claude, ChatGPT, Gemini, and Perplexity. Forward freely.