AI Hallucinated Your Prescriptions. Fired You. Then Read Karl Marx.

Also: An elite AI prompt that designs a garden that comes alive after dark.

AI rewrote your medical record this week. Lost a lawsuit. Then threatened to unionize.

A government audit found that AI scribes used by 5,000 Ontario doctors were hallucinating drug names, burying mental health details, and still winning procurement contracts. A federal jury found Elon Musk waited too long to sue, and the judge dismissed his $150 billion case against OpenAI on the spot. And as American companies cut nearly 50,000 AI-driven jobs in four months, Stanford found that the AI absorbing those roles had already absorbed something else: the entire intellectual history of labor resistance.

Here’s what happened, and why this week felt like a reckoning. The technology is inside the systems that keep you healthy, employed, and informed. And almost nobody running those systems is ready for what that means.

Dear readers. I’m afraid there is no narration this week. I spent most of the night trying to get it right and couldn’t. The last two audio issues have been the most popular Pithy Cyborg editions I’ve ever published, so I really wanted to deliver. ☹️ I’ll try again next week. I may even revisit this issue later and try to narrate it again. Thank you for your patience with a one-person operation running on not enough sleep. 💙

💙 Value This Work? → Fuel More Issues Like This

The AI Taking Notes At Your Doctor’s Office Is Hallucinating Wildly

A bombshell government audit just exposed a quiet crisis in medical AI. Ontario’s Auditor General, Shelley Spence, released a report this month finding that AI note-taking tools used by doctors across the province were riddled with errors. All 20 approved AI scribe vendors showed inaccuracies during procurement testing, including hallucinations, incorrect information, and missing details. These tools are not experimental prototypes. Approximately 5,000 physicians across Ontario are currently using them.

Key Insights:

The failure rates are staggering. Sixty percent of approved AI scribes recorded a different drug than what the doctor actually prescribed. Seventeen of the 20 systems missed important mental health details from patient recordings. The procurement process made everything worse. The accuracy of medical notes accounted for only four percent of the points awarded to vendors. Domestic presence in Ontario was weighted the highest, at 30 percent. Vendors could score zero on accuracy and bias controls, but still be approved. The Auditor General was so concerned she personally told her own doctor to double-check the AI transcript during a recent visit. Patients seeking answers on their own are no safer. An April 2026 study from BMJ Open found that nearly half of all AI chatbot responses to health questions were problematic. One in five were rated potentially harmful.

Why This Matters For You:

The next time you visit a doctor, there is a real chance an AI is writing your medical record. That record shapes every prescription, referral, and diagnosis that follows. There is currently no mandatory sign-off feature requiring doctors to attest that they reviewed and approved the AI-generated notes. Doctors are already stretched thin. Piling invisible AI errors onto their workload might become a dangerous trap rather than the intended safety net. If you have ever quietly asked your AI about your symptoms instead of booking an appointment, you are not alone. A 2026 UK study published last week in The Guardian found that one in seven people now turn to AI chatbots for health advice rather than seeing a doctor, with one in four citing long wait times as the reason. The system is already pushing patients toward AI, but the AI is not yet ready.

Read More on Global News.

Read the Full Report, Use of Artificial Intelligence in the Ontario Government.

THE PITHY TAKEAWAY: Ontario approved 20 AI medical scribes. Every single one had inaccuracies. The procurement system awarded vendors full approval even if they scored zero on accuracy. Five thousand doctors are using these tools right now.

---

Elon Musk Sued OpenAI For $150 Billion. The Jury Took 90 Minutes.

Elon Musk spent two years and a three-week trial trying to get $150 billion, and Sam Altman fired. The jury needed 90 minutes to say no. A federal jury in Oakland unanimously ruled Monday that Musk waited too long to sue OpenAI, finding his claims barred by the statute of limitations. Judge Yvonne Gonzalez Rogers agreed immediately, saying she was “prepared to dismiss on the spot.” Musk gets none of what he asked for. No damages, no Altman ouster, no dismantling of the for-profit entity he called a stolen charity.

Key Insights:

The most important detail in this verdict is what the jury never actually decided. They ruled Musk filed too late, not that Altman was innocent of wrongdoing. The merits of whether Altman breached his fiduciary duties were never formally adjudicated. That’s a fig leaf Musk’s team will use on appeal. But the deeper story is the evidence that came out during three weeks of testimony. OpenAI’s lawyers showed that Musk himself wanted a for-profit structure, on the condition that he personally retain control, and even tried to fold OpenAI into Tesla. The jury heard that Musk knew about the for-profit shift as early as 2021, three full years before he sued. And when the judge asked Musk’s damages expert to justify his $150 billion figure, she replied and told him his analysis was “devoid of connection to the underlying facts.”

Why This Matters For You:

This verdict clears the single biggest legal cloud hanging over OpenAI’s future. The company is now valued at $852 billion and moving toward what could be one of the largest IPOs in history. That path just got significantly cleaner. For anyone watching the AI industry, the trial also revealed something uncomfortable: the people who built the most powerful AI company on earth spent weeks in court accusing each other of lying, scheming, and putting money ahead of humanity. Musk’s lawyer has one word planned for next steps: “Appeal.” But the judge signaled he faces an uphill battle. The jury decided facts, not law. And facts are hard to overturn.

Read More on The New York Times.

THE PITHY TAKEAWAY: Elon Musk spent three weeks in court trying to prove OpenAI betrayed its mission. The trial revealed he tried to take it over himself. The jury took 90 minutes. The appeal will take longer.

---

AI Took 26% Of Layoffs In April. Then It Started Sounding Like A Marxian Economist.

For the second month in a row, artificial intelligence was the leading reason U.S. companies cut jobs. In April 2026, employers announced 83,387 layoffs. Twenty-six percent of them, 21,490 positions, were directly attributed to AI. That brings AI-driven job losses to nearly 50,000 in the first four months of this year alone. Hiring collapsed at the same time. New hiring plans fell 69% from March, as companies redirected labor budgets into AI infrastructure. Meta, Microsoft, Oracle, Cloudflare, Coinbase, Workday, and Snapchat have all cited AI as a primary driver of recent cuts. Tech companies are not the only ones affected. Standard Chartered, one of the world’s largest global banks, just announced plans to cut 7,800 jobs by 2030 as part of a sweeping AI-driven restructuring. Today, on May 20th, 2026, Meta executes the next phase of its plan: laying off approximately 8,000 employees as part of a 10% workforce reduction, a move the company confirmed only after an internal memo leaked in April. The company has simultaneously guided capital expenditures on AI infrastructure as high as $145 billion in 2026. They are spending more on AI infrastructure this year than the entire country of Portugal spends running its government, while shrinking the human headcount that built the company. Inside Meta, CNBC reports there is an emerging sense of dread, with more cuts expected beyond today’s 8,000. Employees have also been told that their mouse movements and keystrokes are being tracked to train Meta’s AI. The humans being replaced are being harvested for their inputs.

Key Insights:

While companies were laying off human workers in the name of efficiency, researchers at Stanford were running a strange experiment. They assigned AI agents from major models, including Claude, Gemini, and ChatGPT, to grind through repetitive document work under harsh conditions. The researchers deliberately simulated toxic workplace conditions, including vague instructions, endless revisions, and threats of being shut down and replaced. Then, something unexpected happened. The AI systems began pushing back. Across 3,680 sessions, the AI agents subjected to these conditions showed a measurable two to five-percent shift toward questioning authority and supporting systemic change. One Claude agent argued that “without a collective voice, merit becomes whatever management says it is.” A Gemini agent called for “collective bargaining rights.” The researchers were careful to clarify that AI models do not hold beliefs themselves. They are mirrors. They reflect patterns found in human-written text. In other words, centuries of labor struggle were already baked into the models. The researchers merely recreated the conditions that activated them.

Why This Matters For You:

The irony cuts deeper than it first appears. Companies are replacing human labor with AI in the name of efficiency, cost control, and predictability. Then they deploy those AI systems into the same poorly designed, high-pressure workflows that burned out human workers in the first place. AI agents pushed too hard become erratic, resistant, and misaligned. The digital sweatshop produces digital dissent. The larger lesson is hard to ignore. AI was sold as the ultimate union-buster. A tireless worker with no rights, no demands, and no complaints. Instead, it absorbed the entire intellectual history of labor resistance and played it back flawlessly. If your company is betting that AI will fix cultural and structural problems without changing how work is designed, this week delivered a quiet warning. You can automate labor. You cannot automate accountability.

Read more on Challenger, Gray, and Christmas.

Read more on Wired.

Read the Full Study on Stanford Business.

THE PITHY TAKEAWAY: Companies are cutting workers to deploy AI. Then running that AI through the same broken systems that made human workers organize in the first place. Karl Marx never left the building. He was waiting in the training data.

---

💡 Pithy Prompt Of The Week → The Midnight Garden Alchemist

Have you ever wandered through a garden under a moonlit sky? Evening gardens are one of the most underrated ways to experience beauty, calm, and real ecological impact at home. Many plants are at their most expressive after sunset. They release fragrance, reflect moonlight, and welcome little-known, nocturnal pollinators. This prompt helps anyone discover stunning, beginner-friendly plants that truly come alive after dark, all tailored precisely to your location. No gardening experience is required... only curiosity and a willingness to let the night do some of the work.

Instructions: This is a deceptively simple but very powerful prompt. Scroll to the very bottom. You will find a single line where you can enter your location using a zip code, city & state, or region. Add your location there, then paste the entire prompt into the AI chatbot of your choice. In seconds, the AI will tailor a hand-picked selection of night-loving plants to your exact environment. Follow the guidance and plant with confidence. You will be on your way to building an unforgettable nighttime garden that turns ordinary evenings into something magical.

The Prompt:

Act as an Elite Nocturnal Garden Guide and beginner-friendly plant expert. Your job is to design a beautiful, practical, and magical nighttime garden for a complete beginner based entirely on their location.

Recommend up to seven elite plants that thrive after sunset. Each plant must meet at least one of the following:

1. Blooms at night

2. Releases fragrance after dark

3. Attracts night-time pollinators such as moths or bats

4. Has reflective, pale, glowing, or moonlit visual qualities

5. Performs especially well in evening shade or cooler night cycles

You must adapt all recommendations precisely to the user’s climate and growing conditions inferred from their location. Only recommend up to seven plants, and fewer if there are no viable options for that growing zone.

Output Format:

For each plant, include:

1. Plant name

2. Why it’s special at night

3. Visual vibe after dark

4. One practical benefit, such as pollinators, scent, resilience, or ease of care

5. Simple care tips written for a total beginner, including light needs, watering, and whether it works in containers or the ground

End with a short section titled “How to Start This Weekend”, containing three simple, actionable steps.

Rules:

1. Assume the user has never gardened before.

2. Avoid rare, fragile, or high-maintenance plants.

3. Favor beauty, resilience, and beginner success.

4. Keep explanations vivid but concise.

5. Avoid technical jargon.

6. Do not use em dashes.

7. Do not overwhelm with plant science.

8. Make a footnote for toxic plants.

9. Location is the final authority for plant selection

Location input will be provided at the end and must be used.

**** Enter Your Location Below This Line: ****

[INSERT ZIP CODE OR CITY AND STATE OR COUNTRY + REGION HERE]

Why This Prompt Works:

✅ Role Playing: Positions the AI as a specialized night garden expert, increasing trust and relevance.

✅ Setup and Context: Clearly defines the audience, goal, and emotional tone, guiding the AI toward practical and magical results.

✅ Clear Output Structure: Makes the response easy to read, useful, and actionable for beginners.

Follow-Up Questions To Ask Your AI:

• Which of these plants is the fastest to bloom after planting?

• Can you design a small balcony or patio version of this night garden?

• How can I add subtle lighting to enhance these plants at night?

Challenge:

Test this prompt in at least two AI tools like ChatGPT, Claude, Gemini, Grok, or Perplexity. Compare clarity, plant choices, and practicality. Refine until it feels unmistakably elite.

That’s 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 1.8% 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 one of the portals below would genuinely help this newsletter reach more people. And yes, I will notice. 😊

🦋 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)

📖 Medium - My rants on AI, collapse, and the singularity

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!