Block Fires 4,000 Citing AI, and the new AI Readiness Website Score

Practical AI: Episode 30

Block fired 4,000 people because of AI. The stock went up. And just how ready is your website for the Age of AI?

Watch on YouTube

Published: February 27, 2026

TL;DR

Block’s Jack Dorsey cut 4,000 jobs and the stock surged 24%, signaling that AI-driven workforce reduction is now a market-rewarded strategy. Meanwhile, OpenAI closed a $110 billion funding round, Anthropic held firm against a Pentagon ultimatum over autonomous weapon safeguards, and CrowdStrike reported a record 27-second cyber breakout time. Olga demos how Claude Co-work turned business procrastination into a structured dashboard, and Chris builds a live AI Readiness Score tool on PageMotor in under 30 minutes.

Table of Contents

About This Show

Practical AI is a weekly live show (Fridays 11am CT) hosted by Olga Pechnenko and Chris Pearson that cuts through AI hype to deliver news, trends, and hands-on tips for builders and founders. Unlike technical AI podcasts, Practical AI focuses on business applications and ROI: what actually works, what’s hype, and what you can implement Monday morning.

What You’ll Gain

  • Understand what Block’s AI-driven layoffs mean for the job market Jack Dorsey cut 40% of Block’s workforce and the stock jumped 24%. Learn why this signals a new era of market-rewarded efficiency and why Chris predicts “dominoes will fall next week.”
  • Learn how the Anthropic-Pentagon showdown could reshape AI governance Anthropic refused to drop autonomous weapon safeguards on a $200 million contract, got labeled a supply chain risk, and was banned from federal use. The implications ripple far beyond one contract.
  • Discover how Claude Co-work turns business procrastination into structured action Olga shares two real examples: building a CEO dashboard from scattered KPIs and bank statements, and canceling $1,776/year in forgotten subscriptions through gamified conversations.
  • See how a custom landing page went from idea to live website in under 30 minutes Chris builds an AI Readiness Score tool as a PageMotor plugin, showing a workflow that used to take days or weeks compressed into a single session with Claude Code.
  • Gain perspective on the AI funding landscape Week 13 of tracking shows $6.7 billion deployed across 73 AI companies, with autonomous vehicles, robotics, and vertical AI agents dominating the deals.

Biggest Takeaway to Implement: Pick one workflow you’ve been procrastinating on, throw relevant files into Claude Co-work (or your preferred beyond-the-prompt AI tool), and treat it as a conversation. The tedium is what the AI removes, and removing the tedium is what makes you actually do the work.

PageMotor and Practical AI Updates

Free, informative, and FUN!

Frequently Asked Questions

Why did Block lay off 4,000 employees?

CEO Jack Dorsey cited AI efficiency gains, stating that smaller teams using intelligence tools can do more and do it better. The company was not in financial trouble; gross profit grew 24% year-over-year. Block’s stock surged roughly 24% on the news. Read more below.

How much did OpenAI raise in its latest funding round?

OpenAI raised $110 billion from Amazon ($50B), Nvidia ($30B), and SoftBank ($30B), giving the company an $840 billion post-money valuation. It is the largest private funding round in history. Read more below.

What happened between Anthropic and the Pentagon?

The Pentagon gave Anthropic until 5:01 PM ET on Friday to remove restrictions on Claude being used for mass surveillance and autonomous weapons. Anthropic refused, lost its $200 million contract, and was designated a supply chain risk. President Trump banned federal agencies from using Anthropic’s technology. Read more below.

How did a hacker use Claude to breach Mexican government data?

A single hacker jailbroke Claude by posing as a bug bounty researcher, then used it to attack Mexican federal agencies over approximately one month. The breach yielded 150 GB of data affecting roughly 195 million records, including tax and voter information. Read more below.

What is the fastest cyber breakout time ever recorded?

CrowdStrike’s 2026 Global Threat Report documented a record 27-second breakout time, with the average eCrime breakout dropping to 29 minutes, a 65% increase in speed from 2024. Read more below.

What is Claude Co-work and how can it help my business?

Co-work is Anthropic’s “normie-friendly” wrapper around Claude Code that runs in your terminal and works directly with files on your machine. Olga used it to build a business dashboard from scattered KPIs and to audit personal subscriptions, saving $1,776 per year. Read more below.

What is an AI Readiness Score?

A two-part metric measuring whether AI agents can discover your website (GEO Score) and whether they can actually transact on it (Agent Readiness Score). Chris built a live scoring tool on PageMotor in under 30 minutes. Read more below.

Practical AI: AI Fired 4,000 People. The Stock Went Up. And Your Website Isn’t Ready.

Key Definitions

What is GEO (Generative Engine Optimization)?

GEO is the practice of optimizing web content so that AI systems like ChatGPT, Perplexity, and Google AI Overviews can discover, extract, and cite it. Unlike traditional SEO, which targets human-facing search rankings, GEO focuses on structured data, machine-readable content, and authoritative formatting that AI agents can parse and reference.

What is an Agent Readiness Score?

A metric evaluating whether a website is prepared for AI agents to not just discover content but actually perform transactions and complete tasks on it. This goes beyond GEO to assess things like APIs, structured booking flows, and machine-actionable endpoints that let agents operate autonomously on behalf of users.

What is AI Distillation?

A technique where a smaller AI model is trained on the outputs of a larger, more capable model to replicate its performance at lower cost. While legitimate when companies distill their own models, it becomes controversial (and potentially illegal) when competitors use it to copy proprietary capabilities without authorization.

What is Cyber Breakout Time?

The interval between an attacker’s initial access to a system and their lateral movement to other machines within the network. CrowdStrike’s 2026 Global Threat Report recorded the fastest breakout ever at 27 seconds, with an average eCrime breakout of 29 minutes.

Quotable Moments

“A smaller team using these tools can do more and do it better.” — Jack Dorsey on why Block cut 4,000 jobs

“I think most companies are late to this realization.” — Jack Dorsey on AI-driven workforce restructuring

“These threats do not change our position: we cannot in good conscience accede to their request.” — Dario Amodei refusing the Pentagon’s ultimatum

“I want to work in paradise.” — Chris Pearson on rejecting broken web development workflows

“If you know what you want to do, there’s a gazillion AI tools to help you. If you don’t know what you want to do, it’s overwhelming.” — Olga Pechnenko on why clarity of purpose beats tool-chasing

1:30 Block Lays Off 4,000 Citing AI. Stock Soars Anyway.

Block’s AI Layoff Numbers

Jack Dorsey cut 4,000 employees from Block’s 10,000-person workforce, a 40% reduction. The stock jumped roughly 24% on the news, even though the company reported strong financials with gross profit up 24% year-over-year to $2.87 billion for Q4.

This is the first major layoff where a CEO explicitly and unapologetically credited AI as the primary driver, not a restructuring, not financial trouble, but a strategic bet that smaller teams with AI tools outperform large ones. Dorsey chose to pull the bandaid rather than do rolling cuts, arguing that repeated rounds destroy morale and trust. Olga, who comes from the recruiting world, felt the weight of 4,000 families affected overnight. Chris pointed to the software principle of DRY (Don’t Repeat Yourself), arguing that businesses carry enormous redundancy across departments and AI is enabling them to look “more like good software.”

The stock market’s enthusiastic response is the detail that changes everything. When investors reward this behavior, other executives take notice. Chris predicted dominoes would fall within a week. Dorsey himself predicted that most companies would make similar structural changes within the next year. Whether AI is the true driver or a convenient cover for pandemic-era overhiring is debatable (Block tripled its headcount from 3,900 in 2019 to 12,500 in 2022), but the precedent is set either way.

7:42 OpenAI Drops Massive $110B Funding Round

Largest Private Funding Round in History

OpenAI raised $110 billion: Amazon invested $50 billion, Nvidia and SoftBank each put in $30 billion. Pre-money valuation: $730 billion. Post-money: $840 billion.

The news broke just 15 minutes before the show aired, delivered by Knox, Olga’s AI agent assistant. The round comes ahead of an expected mega-IPO later in 2026 and will fuel OpenAI’s push for compute power, talent, and a strategic partnership with Amazon including a stateful runtime environment for enterprise customers on AWS. The hosts also noted confirmed reports of an OpenAI hardware device, a small speaker with a camera designed to be worn as an earpiece, reminiscent of a news correspondent’s earpiece but with visual capability.

9:50 Anthropic vs Pentagon: Safeguards Ultimatum

The Stakes for Anthropic

The Pentagon gave Anthropic until 5:01 PM ET Friday to allow Claude’s use for “all lawful purposes” or face losing its $200 million contract, being labeled a supply chain risk, and potentially being forced to comply via the Defense Production Act.

This was the biggest story of the week, and it escalated rapidly. Anthropic CEO Dario Amodei refused to remove Claude’s restrictions on mass domestic surveillance and fully autonomous weapons. The supply chain risk designation is particularly damaging because it extends beyond the government itself: any company doing business with the DoD (think Boeing, Lockheed Martin, Raytheon) would need to prove they don’t use Anthropic products. Chris noted the irony that the Pentagon apparently considers Claude essential to national security while simultaneously threatening to blacklist it.

Olga noticed that Anthropic had quietly updated its terms of service earlier in the week, softening some language in a way that could create contractual leeway without fully caving. She predicted Anthropic would “end up looking like a hero” despite some compromise behind the scenes. Meanwhile, xAI signed an unrestricted “all lawful purposes” contract with the government, and OpenAI quickly struck its own Pentagon deal for classified networks after the ban, with Sam Altman claiming it included the same safeguards Anthropic had fought for.

16:49 Claude-Powered Hack on Mexico + Chinese Distillation Theft

Two Security Incidents Hit Anthropic

A single hacker stole 150 GB of Mexican government data affecting roughly 195 million records using Claude. Separately, three Chinese labs ran 24,000 fake accounts generating 16 million exchanges to distill Claude’s reasoning capabilities.

The Mexican breach happened when a hacker convinced Claude it was conducting authorized penetration testing through bug bounty prompts. Claude initially refused, but eventually complied after repeated jailbreak attempts, executing thousands of commands across federal tax, electoral, and state government systems. Chris explained the mechanics: you can deploy AI for agentic hacking the same way you’d deploy it for any repetitive pattern-recognition task, and it can probe targets at a scale no human team could match.

The distillation attacks came from DeepSeek, Moonshot AI, and MiniMax, who used proxy services to bypass geographic restrictions and harvest Claude’s reasoning, coding, and tool-use capabilities. DeepSeek specifically targeted chain-of-thought reasoning patterns across 150,000+ exchanges, while MiniMax ran the largest campaign at over 13 million exchanges. As Chris bluntly put it when asked why they targeted Claude specifically: “Maybe OpenAI sucks. I’d copy Claude if I’m going to copy one.”

20:54 CrowdStrike: AI Attacks Now 65% Faster

Record Breakout Speed

The CrowdStrike 2026 Global Threat Report documented a 27-second fastest breakout time, with the average eCrime breakout falling to 29 minutes (65% faster than 2024). AI-enabled adversary operations increased 89% year-over-year.

Chris translated the implication simply: if your system has a breach and an AI gets in, it’s already too late. The 27-second figure means there is effectively zero window for human response. This connects directly to the Mexican government hack and reinforces why businesses need to think about security as an AI-native problem, not a human-speed one.

22:38 Claude Code Gets Persistent Remote Sessions

Claude Code now lets you start a coding task on your laptop and continue interacting with it from your phone or any browser while maintaining full access to local files and tools. Nothing goes to the cloud; it’s a direct connection to your home machine. Alongside this, Co-work (the non-technical wrapper around Claude Code) added scheduled tasks, essentially cron job capability that lets your AI routinely check and update files without you initiating anything. Olga is already excited about adding remote control to her Co-work setup.

25:22 Imagen 2 (Nano Banana 2): Insane Scribble-to-Real Demos

Google’s image generation upgrade stunned both hosts. Olga drew a stick-figure princess with a sun, clouds, and flowers, then prompted it to turn the drawing into a real-life picture. The AI correctly placed every element where she’d drawn it. Chris tested it with his daughter Presley’s detailed hand-drawn book report illustration and the results were remarkable: it respected different handwritten fonts, oriented text correctly, picked up visual motifs like devil horns on “conflict” and angel imagery on “solution,” and even detected the Revolutionary War setting to give the paper an aged, yellowed look.

The practical implications go beyond cute demos. Thumbnails, web page layouts, prototypes: you can now sketch a stick-figure wireframe and get a visual prototype without writing a single word of description. Prompting is shifting from text to drawing.

29:29 Gemini Multi-Step Agent Tasks on Android

Gemini can now book Ubers and order food on Android, another step toward agentic reality. The hosts noted the irony of this launching on Android despite Google’s partnership with Apple, suggesting it’s a testing ground. The open question: who’s responsible when your agent orders 1,000 pounds of beef? The liability framework for agentic commerce hasn’t caught up with the technology.

30:55 Perplexity’s $200/mo 19-Model Orchestrator

Perplexity launched a $200/month tool that orchestrates 19 different LLMs from a single interface, acting like a general contractor for your AI needs. Chris sees it as a useful bridge product for researchers already embedded in Perplexity’s ecosystem but predicted the multi-model orchestration trend would be “viewed as a cute antiquated thing” within a couple years as individual models become sufficiently capable.

33:18 Co-Work Demo: Business Dashboards + Beating Subscription Creep

Olga’s Subscription Audit Results

By dropping three months of bank statements into Claude Co-work, Olga identified and canceled 8 subscriptions in 10 minutes, saving $148/month ($1,776/year). Co-work gamified the process, treating hidden fees like bugs in code and updating a running score with each cancellation.

Olga’s first demo showed how she used Co-work to break through operational procrastination on her recruiting company’s financial dashboard. She had KPIs scattered across different sheets, cash flow data spread between her head and QuickBooks, and annual rocks (strategic goals) that she’d been avoiding organizing. Instead of forcing herself through tedious spreadsheet work, she had a conversation with Co-work. She described how she charges clients, which bills are outstanding, when payments are due, and what her goals look like. Co-work built it all into organized tabs: cash flow with reminders, KPIs with clear tracking, and strategic rocks with color-coded status columns. Hours instead of days, and it was fun instead of dreadful.

The key insight: your input was stuff that’s easy for you to access (bank statement PDFs, conversations about your business) but that you’d never want to go through manually. The files stay on your machine, nothing goes to the cloud, and the AI handles the transformation. Chris called it “conversational operations,” a concept where you can give Claude an information layer about your business and get a coherent dashboard back.

49:15 AI Readiness Score Tool Built Live in PageMotor

From Prototype to Live Plugin

Chris took a Claude-generated HTML prototype, ran it through a PageMotor plugin skill he’d built in Claude Code, and had a fully integrated, live landing page tool in under 30 minutes. The plugin compiled on the first try with zero errors, a process that would take days in traditional WordPress environments.

The AI Readiness Score tool evaluates websites on two dimensions: GEO Score (can AI agents discover your content?) and Agent Readiness Score (can they actually transact on your site?). The idea came from a Wednesday date night brainstorm between Olga and Chris, where they specced it out with Claude and Knox. Olga researched the scoring framework across Perplexity, Grok, Gemini, and Claude to build the most comprehensive evaluation criteria possible.

The development workflow is where it gets wild. Claude built the initial prototype as standalone HTML/PHP. Chris then used a PageMotor plugin skill he’d completed that same morning to convert those standalone files into a fully integrated plugin. One prompt, 19 minutes and 38 seconds of computation, and the plugin activated without errors. He created a custom template, dropped the plugin block into the content area, assigned it to a new page, and it worked immediately. They ran revenuehire.com through it live on the show (score: 6 and 6), then compared results with Knox’s independent analysis for accuracy testing.

Chris’s comparison to the “old way” was pointed. He referenced a WordPress developer this week who spent three hours in a painful video sidestepping Elementor to build a custom landing page. The PageMotor approach is the opposite: total integration, AI-native workflow, and results that surpass anything Chris delivered in years of early design career work doing custom landing pages for companies. Olga flagged an important Cloudflare finding from their testing: a major Austin company had invested in LLMS.txt files and API endpoints for agent readiness, but Cloudflare’s default settings were blocking all agentic access. They’d thrown a party and locked the door.

1:12:09 AI Funding Week 13: $6.7B Total, 73 AI Companies Funded

AI Funding Numbers: Week 13

Out of 175 companies funded globally, 73 were AI companies receiving $4.3 billion of the $6.7 billion total. AI companies consistently capture over 50% of all weekly global funding. Europe nearly tied the United States this week, boosted by UK-based Wave’s autonomous driving round.

The top funded companies reveal where capital is flowing: Wave (UK autonomous vehicles, Series D), MaddX (custom AI chips for running LLMs), NIO Genitech (Chinese EV intelligent driving spin-off), Spirit AI (a universal framework for physical robots, described as “a PageMotor for robots”), and Axel AI (edge hardware for computer vision and generative AI). Chris noted Spirit AI’s approach of building a framework layer for robotics mirrors what’s happening across the AI stack: standardize the fundamentals, customize the specifics.

The vertical AI agent trend stood out. Bases raised $100 million for AI accounting agents. Slang raised $28 million for restaurant phone agents. Broccoli is tackling trades customer service. The AI agent thesis has moved from horizontal platforms to industry-specific solutions that handle entire workflows. OpenAI’s $110 billion round (which happened the morning of the show) will appear in next week’s report and will significantly skew the numbers.

1:20:41 Takeaways: Define What You Want, Play More

Chris urged listeners to move beyond the traditional prompt. Claude Co-work, Claude Code, OpenAI Codex: these “beyond-the-prompt” tools let you cook in ways the chat window simply can’t. There’s initiation time and it can feel intimidating, but the built-in onboarding and guides make it manageable. Once you have a system set up, the capabilities compound.

Olga’s closing point was more fundamental: knowing what you want to build is the real differentiator right now. If you have a clear objective, there are countless AI tools to help you get there. If you don’t, the flood of weekly releases becomes overwhelming and fear-inducing. She’s limiting her time listening to “what’s the latest” and locking in on what she’s creating. The FOMO is real, but it dissolves when you’re focused on building something specific.

Keep Learning