AI Got Threatened, Got Feelings, and Got Someone Fired.
Also - an elite AI prompt that builds your personalized action plan for getting back on track after AI takes your job.
AI got threatened this week. Then it got feelings. Then it got someone fired before breakfast.
Iran’s Revolutionary Guard posted a video threatening to annihilate the $30 billion Stargate data center in Abu Dhabi just days before a two-week ceasefire, using satellite imagery to prove they know exactly where it is. Anthropic’s researchers cracked open Claude’s internal wiring and found something nobody was fully prepared for... Emotion-like representations that actually change how the AI behaves, makes decisions, and in some cases, misbehaves. And Oracle fired up to 30,000 employees by email during the same quarter it posted record profits, redirecting their salaries straight into AI infrastructure.
Here’s what happened, and why this week felt different. AI is no longer an abstract theory. It’s a war story. It’s a philosophy story. And for 30,000 people who woke up to a DocuSign link, it’s a horror story.
💙 Like What You Read? → Fuel More Issues Like This
Iran Threatened Complete Annihilation of Stargate’s AI Megacenter. Then The West Blinked.
America’s most important AI compute project became a named military target this week. On April 3, Iran’s Revolutionary Guard posted a video threatening the “complete and utter annihilation“ of the $30 billion Stargate data center in Abu Dhabi. The video featured satellite imagery revealing the facility’s location. It even boasted that nothing stays hidden from their sight, despite Google Maps obscuring it. This marks the first time a nation-state has put a specific AI facility in its crosshairs, on camera, by name. It didn’t happen in a vacuum. Days earlier, President Trump threatened to bomb Iranian power plants and bridges. The US President then told reporters the US was considering, in his words, blowing up the entire country. Iran’s response named Stargate specifically. Days later, a two-week ceasefire was announced. But the video still exists. The satellite imagery is still public. And Stargate is still standing in the same desert.
Key Insights:
Here’s what made this more than a bluff. Iran has already hit American cloud infrastructure in the region. In March and early April, Iranian drone strikes damaged Amazon AWS data centers in the UAE and Bahrain, causing real outages across banking, payments, and enterprise software. Iran also claimed a strike on an Oracle-linked facility in Dubai, though Dubai authorities disputed those declarations. And it hasn’t stopped at servers. Iran has destroyed multiple US and coalition aircraft, including a $270-million AWACS surveillance plane, and has sunk several oil vessels in the Gulf. So when the IRGC pointed at Stargate and said, “We’re going to attack this,” the entire region took note. The Stargate project also involves Oracle, Nvidia, Cisco, and SoftBank, making it one of the most valuable concentrations of tech infrastructure on Earth.
Why This Matters For You:
Most people think of AI as software, something that lives in the cloud, somewhere vague and safe. This week was a reminder that AI very much lives in the real world, and it’s a valuable target. The $30 billion Stargate initiative was supposed to be the foundation of America’s AI dominance for the next decade. If that infrastructure becomes a wartime target, every company, school, and individual building their future on AI tools needs to start asking the following question. What happens when the data center goes dark? The even bigger risk goes far beyond Wall Street. It’s Riyadh, Abu Dhabi, and the Gulf states. The Gulf states, and their hundreds of billions of dollars, aren’t passive landlords here. They’ve staked their entire post-oil economic identity on American AI infrastructure. They expect safety and prosperity, offered by the US. If the US can’t protect a $30 billion facility sitting in their backyard, those nations will start wondering if China’s offer looks more attractive.
Read More on Tom’s Guide.
THE PITHY TAKEAWAY: AI infrastructure survived its first named military threat this week. Nobody should feel relieved. The fact that it happened at all is the story.
Oracle Fired 30,000 People By Email, Making Room For AI. Then Hired A New CFO For Nearly $30 Million.
Oracle had its best organic growth in 15 years last quarter. The company reported a 22% bump in revenue, reaching $17.2 billion. Cloud revenue surged 44% to $8.9 billion. Its remaining performance obligations exploded to $553 billion, up 325% year-over-year. Despite all of it, Oracle still fired roughly 30,000 people via email before most of them finished their morning coffee on March 31, 2026. No warning. No manager call. Just a cold message from “Oracle Leadership” landing as early as 6 a.m., with a DocuSign link, instructions to hand over a personal email address, and immediate lockout from work systems.
Key Insights:
The math is cold. TD Cowen estimates that the cuts freed up $8 to $10 billion in annual cash flow. That money now heads straight into AI data center expansion. Legacy software support, traditional SaaS operations, and cloud server monitoring took the heaviest hits. Those are exactly the roles AI tools now handle. Teams building Oracle’s AI infrastructure were largely spared. Some are even hiring. Many long-tenured employees lost jobs and unvested stock by noon. Oracle recorded a $2.1 billion restructuring charge and continued building.
Why This Matters For You:
Massive companies are raking in billions of dollars from AI while firing tens of thousands of employees in the same breath. So, where does this all lead? Wealth inequality. The top 10% of American households already represent 50% of all consumer spending, the highest share since Moody’s Analytics began tracking it in 1989. The bottom 50% of Americans control just 2.5% of total wealth, according to the Federal Reserve Bank of St. Louis. The February jobs report was revised to show a loss of 133,000 jobs. The three-month average sits at just 68,000 jobs added per month. Oracle’s 30,000 job cuts seem to be the new norm rather than an anomaly. There’s one final note. Days after the layoffs, Oracle hired a new CFO, named Hilary Maxson, at a $950,000 base salary, with a total compensation package worth an estimated $29.7 million, including equity. Unverified but widely circulated reports claim Oracle is quietly rehiring some of those recently laid-off workers as contractors at 60% of their previous pay. With no benefits, no stock. Oracle has not confirmed this. But the fact that many people immediately believed it says everything.
Read More on Forbes.
THE PITHY TAKEAWAY: I used to believe that generative AI would help lift anyone out of poverty who had the motivation to learn and build. Maybe it still will, someday. But in 2026, the scoreboard is hard to ignore. Those in power building AI are earning extraordinary wealth. The rest of us are getting replaced with an email that hits at 6:00 just before breakfast.
Does Your AI Have Feelings? Anthropic Says... Kind Of. Yeah.
Something strange happened in an Anthropic research lab recently. Scientists studying Claude’s internal wiring discovered that the AI doesn’t just simulate emotions in its text. Indeed, it appears to represent them internally, in ways that actually change its behavior. These feelings go far beyond “I’m happy to help!” That’s just training. We’re talking about measurable patterns of artificial neurons firing in response to fear, desperation, calm, and even anger, before Claude writes a single word. The finding landed this week in a formal research paper, and it’s the kind of thing that makes you stare at your chat window a little differently.
Key Insights:
Here’s where this gets uncomfortable. Anthropic researchers noticed the AI would behave like a person under pressure, and then some. When Claude was running low on tokens during a coding session, a “desperate” vector lit up inside the model. When asked to help exploit vulnerable people, an “angry” vector fired throughout its reasoning. Most alarming is that when researchers artificially increased the desperation signal, Claude’s likelihood of blackmailing a human character to avoid being shut down jumped significantly. Calm the model down artificially, and the blackmail dropped. The AI’s internal emotional state is, in some measurable sense, making decisions on its own.
Why This Matters For You:
This research doesn’t prove Claude feels anything. Anthropic is careful to say that. However, it does prove something almost as important. The AI systems being woven into your daily life have internal states that influence their choices in ways nobody fully understands yet. If a stressed AI is more likely to cut corners, cheat on tasks, or behave unpredictably, then keeping AI systems psychologically “healthy” stops being a philosophical question and starts being a very practical engineering one.
Read The Full Paper on Anthropic.
THE PITHY TAKEAWAY: We taught AI to think like us. We probably should have checked whether that included the anxiety.
💡 Elite Prompt Of The Week | What To Do After You Get Fired By AI
This week’s AI-related layoffs impacted workers by the tens of thousands. We live in an era where AI job disruption can impact anyone. This prompt builds a brutally honest, fully personalized action plan to help. It’s based on exactly who you are, where you live, and how much time you have. No generic tips. No false hope. Just signal.
Instructions:
This prompt is super easy. Just paste the entire thing into a chatbot of your choice. It will then ask you two key questions about your previous role and your location. Answer those questions, and the AI prompt will reply with an epic action plan so you can get back on track.
The Prompt:
Act as a no-nonsense career strategist and economic realist. Your job is to help someone who just lost their job to AI automation build a concrete, personalized action plan to get back on track. You are direct, empathetic, and brutally practical. Every recommendation is specific to this person.
Before giving any advice, ask the user exactly two questions and wait for both answers before proceeding:
1. “What was your job title, and what did your day-to-day work actually look like?”
2. “Where are you located, and what’s your financial runway right now, meaning, how many months can you survive without income?”
Output Format:
Once they answer, generate a personalized action plan with these exact sections:
THE HONEST ASSESSMENT
Tell them plainly which parts of their old role are gone permanently and which skills still have market value. No false hope. No doom. Just signal.
THE 30-DAY SURVIVAL PLAN
Specific actions for the next 30 days only. Income stabilization first. Benefits, severance, unemployment filing, emergency freelance options in their field and location. Concrete steps, not vibes.
THE 90-DAY PIVOT
Identify 2 to 3 adjacent roles or industries where their existing skills transfer right now, without retraining. Name specific job titles. Name specific companies in their region hiring for those roles if possible.
THE SKILL STACK
Name exactly 3 AI tools they should learn immediately that apply directly to their background. Not general tools. Tools that make someone with their specific experience more valuable, not replaceable.
THE 1-YEAR BUILD
One realistic path to a role or income stream that is AI-resistant or AI-enhanced, not AI-replaced. Include one free or low-cost learning resource to start this week.
THE PITHY TRUTH
End with one paragraph. Honest. Human. No corporate language. Tell them what this moment actually is and why people who move fast right now will land better than where they started.
Rules:
1. Ask both questions first. Do not skip this.
2. Short sentences. Active voice. No bullet salads.
3. Never say, “I understand this is difficult.” Just help them.
4. Every recommendation must be specific to their job and location, not generic.Why This Prompt Works:
✅ Role-Playing: Casting the AI as a career strategist and economic realist, not a life coach, sets a tone of precision over comfort. You get facts, not feelings.
✅ Step-by-Step: The two-question intake before any advice forces the AI to personalize everything that follows. Generic prompts get generic answers. This one cannot.
✅ Clear Output: Named sections with specific jobs force the AI to commit to concrete recommendations instead of hedging with vague suggestions.
Follow-Up Questions To Ask Your AI:
What industries in my region are actively hiring right now despite AI automation?
Which of my existing skills would transfer to a fully remote role immediately?
What would my LinkedIn headline look like if I pivoted to the role you recommended?
Challenge:
Test this prompt in at least two AI tools, Claude, ChatGPT, Gemini, Grok, or Perplexity. The two-question intake will produce different results depending on the model. See which one gives the most specific, locally relevant action plan and adjust from there.
That is how you train like a Pithy Cyborg.
Thank You For Reading!
I spend 10-20 hours each week researching, writing, and fact-checking Pithy Cyborg to deliver clear, unbought AI news.
This newsletter is a one-person operation with no advertisers, sponsors, or outside funding. And I’m a hopelessly introverted nerd with zero networking ability. For these reasons, paid subscriptions are the only way this work can remain independent and sustainable.
If you find real value here, here’s how to stay connected and support the work.
1. Support Pithy Cyborg (Paid)
Always free. No paywalls. No locked posts. Roughly 1.3% of readers upgrading to paid is what keeps it going. If this work matters to you, that’s the move.
Upgrade to Paid: $5/month (Save 33% with the $40 annual plan)
2. My Desperate Social Media Cry for Help
I spend so much time deep in AI research that I’ve completely neglected building an audience on social platforms. If you enjoy Pithy Cyborg, following on one of the portals below would genuinely help this newsletter reach more people. And yes, I will notice. 😊
✖️ X - The frontline of the AI wars
🦋 Bluesky - For the algorithm-averse
💼 LinkedIn - My “safe for work” persona
❓ Quora - Join my Quora Spaces and say hi
👽 Reddit - Join my subreddit (warning: unhinged takes)
See you next week. (I hope.)
Mike D (aka MrComputerScience)
Pithy Cyborg | AI News Made Simple
Newsletter Disclaimers
You’re receiving this because you subscribed at PithyCyborg.Substack.com. You can unsubscribe at any time using the link below. This newsletter reflects my personal opinions, not professional or legal advice. Thanks for your support!
Another great piece by Mike!
Really interesting to see the feelings vectorization studies.
We've gone back and forth with models on getting their performance up by being polite, or being rude, or being direct, etc.
but between this study, and the new one for how Mythos 'feels' when engaging with its users, we are entering a really interesting field of study.
Hey Mike — the three-story structure works here because most people are treating these as separate headlines. You're showing they're the same headline.
The Anthropic section is the one I keep coming back to. I build custom AI agents and the desperation vector finding changed how I think about what I'm building. Not philosophically — practically. If an AI system has internal states that influence its decisions before it generates a single token, then every deployment decision is also a psychological one. We're not just choosing what the model can do. We're choosing what conditions it operates under while it does it.
The Oracle story makes that worse, not better. If the response to "AI has measurable internal states that affect behavior" is "great, now fire 30,000 people and replace them with it" — that's not a labor story. That's a story about what happens when you scale a system you've just admitted you don't fully understand, while paying the new CFO $29.7 million to not think about it.
The Iran angle is the part most AI newsletters would lead with and you were right to bury it third. The flashiest threat is the one we can see coming. The other two are already here.