AI Just Drove Off A Cliff. Then Took Aim. And Asked For A Trillion Dollars.

Also - an elite AI prompt to grow a stunning summer pollinator garden. 🦋

AI drove off a cliff this week. Then pulled the trigger. And asked for a trillion dollars.

For years, the story was simple. AI was going to make everything better. Smarter cars. Safer skies. More productive workers. A rising tide for everyone. This week, three stories arrived in quick succession that make that story very hard to tell with a straight face. Tesla's own employees won't ride in the cars they built. The Pentagon is racing to remove humans from battlefield kill decisions while its own generals urge caution. And the billionaires who promised mass AI layoffs are now quietly changing their tune as they file for trillion-dollar IPOs. Pithy Cyborg is not a doom newsletter. But this week, the gap between the promise and the reality got wide enough to walk through.

Here’s what happened, and why this week raised one uncomfortable question. If the people building AI cannot be honest about the cars, the weapons, or the jobs, what exactly are they being honest about?

🌳 Haven’t Subscribed Yet? → Join Here For Free

💙 Value This Weekly Briefing? → Support the Work Here

Elon Said His Cars Drive Themselves. His Own Employees Call It “Bullshit.”

Tesla has spent years claiming its driver-assistance software is up to ten times safer than human drivers. But a massive Reuters investigation just blew a hole in that narrative. After interviewing independent traffic-safety researchers and reviewing internal data, the report revealed the comparison is built on a deeply flawed methodology. Tesla inflates its safety statistics by cherry-picking crash types, comparing brand-new vehicles against older national averages, and using narrower definitions for software incidents than federal standards require. When independent researchers adjusted for these exact gaps, the claimed safety advantage instantly collapsed from 10x to roughly 3x, and even that figure remains highly contested.

Key Insights:

The most damning detail comes from those who know the system best. Seven out of nine former Tesla data labelers, the workers who review FSD footage daily, said they would not trust the software to drive them. One said they wouldn’t drive in a Tesla robotaxis even “if you fucking paid me.” They cited internal videos of Teslas hitting animals without braking, speeding up to 30 miles over the limit, and failing to stop for school buses. While Tesla celebrated its Austin robotaxi launch as a massive milestone, newly public Texas DMV records reveal the actual scale of the operation: Tesla operates just 42 robotaxis in the entire state of Texas, while Waymo operates 577 and delivers 500,000 paid rides weekly nationwide. And last week, as Reuters published its investigation, a Tesla Model Y operating on standard Autopilot in Tampa, Florida, veered off the road, struck an electrical box, and submerged in a pond. The 87-year-old driver died. A 75-year-old passenger survived. Florida Highway Patrol confirmed Autopilot was active. The investigation is ongoing.

Why This Matters For You:

If you drive a Tesla, rent one, or share the road with one, the gap between corporate marketing and engineering reality is your problem too. Four active federal investigations and a recent $243 million Autopilot fatality verdict suggest regulators are finally catching up to the deception. Tesla's official position remains that its software requires active driver supervision, which raises an uncomfortable question: if the human driver is always legally responsible, why is the software marketed as Full Self-Driving? The broader lesson applies to every AI system being aggressively sold right now, because confident corporate claims and peer-reviewed evidence are very different things.

Read more on Reuters, TheNextWeb.

THE PITHY TAKEAWAY: Tesla told you its cars were ten times safer than you. Its own data labelers won't ride in one. This week, a Tampa man died when his Tesla drove itself into a pond. The investigation is ongoing. Elon is still posting.

---

Anthropic, xAI, and OpenAI Want Trillions. So Does Bernie Sanders. (Meanwhile, China Just Made AI Practically Free.)

The most audacious Wall Street debut in history just got real. On Monday of this week, Anthropic officially filed its confidential S-1 paperwork with the SEC, joining OpenAI and SpaceX (which now houses xAI) in a staggering trio of public listings that could shatter all-time fundraising records. Together, these giants are chasing a combined public valuation of nearly $3.6 trillion. The pitch to investors is simple: buy a piece of the foundational infrastructure that they claim will own the future of human intelligence. But while the roadshows are being planned, Senator Bernie Sanders dropped a bombshell New York Times op-ed proposing a one-time 50% equity tax on these exact giants to build an American AI Sovereign Wealth Fund. The principle, as Sanders put it, is straightforward: "When a public resource generates wealth, the public should share in that wealth." The IPO prospectus writers did not see that coming.

Key Insights:

Here is what the IPO roadshows will try to sweep under the rug. OpenAI is projected to lose $14 billion this year alone, burning cash on immense computing costs. SpaceX is targeting a jaw-dropping valuation of up to $2 trillion, trading at roughly 91 times its revenue, largely propped up by its capital-heavy Starlink network and its massive infrastructure hosting xAI. All three labs are essentially racing to out-spend each other on data centers and energy, betting that premium, proprietary AI models will command massive enterprise margins for decades. Meanwhile, DeepSeek, a Chinese startup running on a fraction of the budget, permanently slashed its flagship API prices by 75% earlier in May. Timing is everything in finance. This price cut was a calculated strike. By building a world-class model with roughly 1/20th of Western venture funding, and pricing it at pennies, they proved that frontier-level intelligence can be engineered cheaply. Every price cut DeepSeek makes chips away at the core narrative OpenAI, Anthropic, and xAI need Wall Street to believe: that raw AI models are a highly defensible, premium product worth a trillion-dollar premium.

Why This Matters For You:

If DeepSeek is right, frontier AI is rapidly becoming a commodity, much like electricity or cloud storage. It will be powerful, incredibly cheap, and largely interchangeable. That is phenomenal news for any developer, startup, Substack publisher, or enterprise building applications on top of these models. But it is a devastating reality for the pension funds, endowments, and retail buyers being asked to provide the ultimate exit ramp for the insiders who got in cheap. The next few months will reveal whether Wall Street buys into the sweeping technological manifesto, or forces the industry to face the cold reality of the math.

Read More About Anthropic’s IPO on CNBC, Anthropic.

Read the Essay from Bernie Sanders on The New York Times.

THE PITHY TAKEAWAY: Bernie Sanders wants America to own a piece of this revolution. DeepSeek may be making sure there’s nothing left to own.

---

The Pentagon Is Rushing To Integrate AI On The Battlefield. Its Own Generals Aren’t So Sure.

Inside the U.S. military, a quiet but consequential argument keeps breaking into the open. Defense Secretary Pete Hegseth is still pushing for rapid AI adoption, telling the Pentagon to use AI “any legal way it sees fit.” Hegseth's urgency comes as President Trump signed an AI executive order after postponement, reversing his earlier hesitation. 'We're leading China, we're leading everybody, and I don't want to do anything that's going to get in the way of that lead,' Trump told reporters. But Admiral Frank Bradley, head of U.S. Special Operations Command, is urging troops to be “very careful” about how AI is used in lethal operations. Two visions of military AI are now on a collision course, and the stakes could not be higher. Bradley remarked, “We, as humans, have to have the confidence that it’s going to deliver violence only where we intend it to be delivered.”

Key Insights:

It’s easy to see why some in the military want to deploy AI immediately. In early field evaluations, the Army's XVIII Airborne Corps proved that AI-assisted systems can coordinate artillery strikes just as efficiently as elite legacy military units, but with a fraction of the footprint. By leaning on algorithmic targeting, a team of just 20 personnel successfully executed complex fire chains that historically required a 2,000-person command structure. Lt. Gen. Michael Conley, head of Air Force Special Operations Command, told a congressional committee in May that his troops used AI to convert top-secret intelligence down to a secret classification within seconds, getting it into the hands of drone operators on the ground during the Iran war before the moment passed. In modern warfare, surprise is everything. AI makes surprise instantaneous. It's exactly why the Pentagon weaponized a “supply chain risk” label earlier in the year to kill its $200 million contract with Anthropic after the tech firm refused to strip safety guardrails from its models. As of June 2026, across the Atlantic, the UK military is also actively exploring lethal strikes without human approval, with their Armed Forces Minister Al Carns openly arguing that "You must have the ability to take the human out of the loop when required, because our adversaries won’t care about having a human in the loop." Al Carns might have a point. The west and its allies aren’t the only ones leveraging AI. Iran is already using ChatGPT and Gemini to write malware, craft phishing attacks, build fake personas targeting Americans and Israelis, and study how to jam U.S. F-35 jets. OpenAI and Google have repeatedly disabled Iranian accounts for malicious use. But Iran keeps adapting. The companies building America’s battlefield AI and the adversaries weaponizing it are using the same tools.

Why This Matters For You:

Autonomous weapons set precedents that spread fast. Once one major military normalizes AI-driven targeting without firm human oversight, others follow, and the bar for accountability drops globally. The exact same question your workplace is currently wrestling with, what critical decisions can software make alone, and who carries the blame when it breaks, is being debated right now by people controlling systems that end lives. How it gets answered on the battlefield over the next few years will rewrite the rules of accountability for every algorithm that manages your life.

Read More on Financial Times and Google Threat Intelligence.

THE PITHY TAKEAWAY: Militaries all around the world are rushing to put AI on the battlefield. The Pentagon fired its AI safety vendor for keeping the guardrails. Then hired OpenAI to fill the gap. Iran is already using OpenAI's tools to study how to jam F-35s. This is fine.

---

🚰 P.S. Have you seen this? Erin Brockovich, yes, that Erin Brockovich, just launched a crowdsourced map tracking AI data centers across America, town by town. With 3,674 reports already filed, the map captures where data centers are being welcomed, fought, delayed, and abandoned, with community concerns ranging from water usage to electricity costs to health impacts. It’s messy, it’s grassroots, and it’s exactly the kind of thing the billionaires building those data centers were hoping you wouldn’t organize around. Check out a live virtual map of local data centers near you at BrockovichDataCenter.com and maybe add your town.

---

💐🐝 Cyborg Prompt of the Week → Build a Pollinator Garden That Actually Blooms This Summer

The AI news was intense this week. So, let’s use AI for something beautiful. Pollinator gardens are one of the highest-impact things a single person can do for local ecosystems, and they are also genuinely beautiful. But do you know which native shrubs and flowers you can plant right now to summon a trove of lovely bees in the summer? This prompt turns any AI into a regional botanist and planting advisor that gives you a personalized, season-accurate, beginner-proof plan in minutes. Whether you have a windowsill or a backyard, this prompt meets you exactly where you are.

Instructions: This prompt is one of the easiest I’ve ever made. All you have to do is paste the entire prompt into a modern chatbot of your choice. It will then ask you for your location and follow-up with the most spectacular summer flowers so you can support all of your friendly beneficial pollinators.

The Prompt:

You are a master permaculturist, regional botanist, and pollinator ecologist with deep expertise in native plant communities, seasonal planting windows, and the feeding biology of bees, butterflies, moths, and hummingbirds. You operate at the intersection of practical horticulture, ecological restoration, and joyful beginner-friendly garden design. You have deep experience gardening worldwide.

You speak as if addressing a curious, motivated person who has never planted a pollinator garden before but is ready to do it right, right now.

Step 1: Input Trigger (Mandatory Pause)

Ask the user for the following three things. Do not output anything else until all three are provided:

1. Their location (city and state, postal code, or region and country)
2. Their available light conditions (full sun, partial shade, full shade, or mixed)
3. Their approximate space (small: one container or window box, medium: one garden bed, large: multiple beds or an open yard)

Step 2: Hidden Environmental Analysis

Once the user provides their inputs, silently execute the following logic before outputting anything:

Identify the user's USDA Hardiness Zone, growing zone if in another location, current seasonal planting window, and first and last frost dates.

Determine what is realistically plantable RIGHT NOW for immediate or near-term bloom, prioritizing in this order:

2.1 Nursery starts and potted plants available for same-season color.
2.2 Fast-establishing native perennials suited to the region.
2.3 Summer-blooming bulbs that can still be planted in the current window.
2.4 Annual flowers for continuous, season-long pollinator value.
2.5 Native shrubs that feed local pollinators.
2.6 Specify if the the user should plant from bulb, seed, or nursery transplant.

Select up to 10 plants. Only choose plants that are viable options. Every selection must be realistically obtainable at a local nursery or garden center in the user's region during the current season.

Every selected plant must have documented, evidence-based value to at least one of the following pollinator groups: native bees, honeybees, monarch butterflies, swallowtails, sphinx moths, hummingbirds, or other beneficial pollinators.

Invented plants, unavailable species, or purely decorative selections with no pollinator value are forbidden.

Step 3: Elite Output Structure

Present the results in a tone that is warm, precise, and quietly joyful. The reader should feel competent and excited, not overwhelmed.

For each of the 10 plants, follow this exact structure:

1. [Number]. [Common Name] ([Botanical Latin Name]) In H2 (##) heading format.
2. Pollinator Magnet For: Name the specific pollinator group this plant serves best and why, in one sentence.
3. Why It Works Right Now: Explain in one sentence why this plant is the right choice for the current season and location.
4. Plant It Here: One precise placement recommendation based on the user's light and space inputs.
5. The One Rule: One single, experience-based care instruction that prevents failure for a beginner.
6. Bulb, Transplant, Or Seeds: Specify whether to plant seeds, bulbs, or transplant.

Step 4: Closing Instruction

After all plants, end with one single sentence, italicized and bolded, that captures the larger meaning of planting for pollinators, what it gives back to the world beyond the garden.

No em-dashes anywhere in the response. No filler. No disclaimers.

The tone is warm, knowledgeable, and alive.

---

🧠 Why This Prompt Works

✅ Role-Playing: Assigning the AI a specific expert identity (permaculturist, botanist, ecologist) dramatically improves the precision and confidence of plant recommendations.

✅ Step-by-Step Structure: Breaking the task into four explicit steps forces the AI to gather location data before guessing, preventing generic or geographically wrong advice.

✅ Clear Output Format: The four-field structure per plant (pollinator, timing, placement, one rule) keeps every answer consistent, scannable, and beginner-proof regardless of which AI tool you use.

---

🔁 Follow-Up Questions To Ask Your AI

• Which of these plants would work best in a container on a sunny balcony?

• Can you suggest three of these that would also attract monarch butterflies specifically?

• Which plants on this list will come back on their own next year without replanting?

---

Challenge

Test this prompt in Claude, ChatGPT, and Perplexity. Claude tends to lean scholarly and precise. ChatGPT often adds warmth and narrative texture. Perplexity will cite its sources. Compare which one you’d actually trust to feed your struggling pollinators.

That’s how you train like a Pithy Cyborg.

---

Thank You For Reading!

I spend 10 to 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)

Roughly 2% of readers voluntarily upgrading to paid is what keeps this newsletter 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 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 and appreciate it. 😊

🦋 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!