The tech industry spent 2025 talking about AI safety, robotaxis, and smart glasses going mainstream. Underneath all that, the circus never stopped.
From Mark Zuckerberg suing Mark Zuckerberg to a $599 toilet camera that wasn’t even properly encrypted, this year delivered some truly spectacular nonsense. Here’s a quick tour of the dumbest tech stories that somehow really happened.
Mark Zuckerberg sues… Mark Zuckerberg
Mark Zuckerberg, a bankruptcy lawyer in Indiana, has a problem: his name is Mark Zuckerberg.
He did what every small business owner does. He bought Facebook ads to promote his law practice. Then Meta’s systems kept flagging and suspending his account for “impersonating” Mark Zuckerberg — as in, the Meta CEO.
The lawyer Zuckerberg says this has been going on for years. He’s been practicing law since “Mark Zuckerberg” (the CEO) was three years old. He even runs a site, iammarkzuckerberg.com, just to reassure potential clients that no, he is not secretly the guy who owns Facebook.
“I can’t use my name when making reservations or conducting business as people assume I’m a prank caller and hang up,” he wrote on the site, comparing his life to that old Michael Jordan ESPN ad where an ordinary guy’s famous name causes constant chaos.
After repeated suspensions — while still being charged for ads — he finally did the only logical thing left: Mark Zuckerberg sued Mark Zuckerberg. Meta’s lawyers have plenty on their plate, so this one could drag on. But somewhere on a court docket, this caption already exists, and that alone is art.
The engineer who worked for half of Silicon Valley at once
Soham Parekh turned job-hopping into a full‑time sport.
It blew up when Mixpanel founder Suhail Doshi warned founders on X that he’d hired Parekh for his new company, only to realize the engineer was also working for several other startups at the same time.
“I fired this guy in his first week and told him to stop lying / scamming people. He hasn’t stopped a year later. No more excuses,” Doshi wrote.
Other founders quickly chimed in. Doshi said three of them reached out that same day to say, thanks for the heads‑up — because they were currently employing Parekh too.
To some, Parekh was a scammer gaming early‑stage startups for quick money. To others, he was a twisted kind of legend. Getting hired by that many companies, in this hiring market, is its own dark art.
Chris Bakke, founder of job‑matching platform Laskie, summed up that view on X: Parekh should admit what he did was wrong and “course correct to the thing he’s top 1% at” — interviewing.
Parekh admitted he was working multiple jobs. But his story still doesn’t add up. He claims he was lying to companies for money, yet often picked equity over cash — which usually takes years to vest, and he was getting fired fast. Not exactly an optimized grift.
Sam Altman vs. a bottle of olive oil
Tech CEOs get yelled at for a lot of things. This year, Sam Altman caught heat for… misusing olive oil on camera.
The OpenAI CEO appeared in the Financial Times’ “Lunch with the FT” series, cooking pasta on video. FT writer Bryce Elder noticed Altman was using trendy olive oil brand Graza — but using the wrong one in the pan.
Graza sells Sizzle (for cooking) and Drizzle (for finishing dishes, salads, anywhere you actually taste the good stuff). Heat destroys flavor, so you don’t waste the fancy early‑harvest oil in a hot pan.
Altman apparently did. Elder roasted him, calling his kitchen “a catalogue of inefficiency, incomprehension, and waste,” and used the bit as a metaphor for OpenAI’s big‑energy, big‑resources approach to AI.
A light, funny media critique about olive oil somehow enraged corners of Altman’s fanbase more than many serious pieces on AI policy. Of all the OpenAI controversies in 2025, the loudest backlash for some journalists came from… #olivegate.
Soup diplomacy in the AI arms race
If you had to pick one overarching storyline this year, it was the AI arms race. OpenAI, Meta, Google, Anthropic — all shipping bigger models, faster.
Meta has been especially aggressive in recruiting. It hired multiple OpenAI researchers over the summer. OpenAI CEO Sam Altman even said Meta was dangling $100 million signing bonuses to lure his people away.
Then, in December, OpenAI chief research officer Mark Chen dropped a detail on Ashlee Vance’s Core Memory podcast that instantly entered tech lore.
“You know, some interesting stories here are Zuck actually went and hand-delivered soup to people that he was trying to recruit from us,” Chen said.
Yes: according to Chen, Mark Zuckerberg, multibillionaire CEO, personally showed up with soup for OpenAI researchers he wanted to hire.
Chen wasn’t going to let that bit of hospitality go unanswered. So he said he responded in kind — bringing soup to Meta employees in return.
The great AI race of 2025: giant models, eye‑watering comp packages, and a quiet side war fought with artisanal broth.
Sign an NDA to build Lego. There will be pizza.
On a random Friday night in January, investor and former GitHub CEO Nat Friedman posted an offer on X that read like a parody of Palo Alto:
“Need volunteers to come to my office in Palo Alto today to construct a 5000 piece Lego set. Will provide pizza. Have to sign NDA. Please DM”
TechCrunch checked at the time if this was a joke. Friedman said it was very real.
And then… nothing. No reveal. No photos. No explanation for why a 5,000‑piece Lego project required an NDA.
What was the set? Why the secrecy? Was the pizza worth it? Is there a stealth Lego‑based hardware prototype under someone’s desk at Meta right now?
Because yes, about six months later, Friedman quietly joined Meta as head of product at Meta Superintelligence Labs. There’s no public evidence the two things are linked. But given this year’s soup diplomacy, you can’t rule out “Lego + pizza + soup” as Meta’s next recruiting strategy.
Bryan Johnson livestreams his shrooms trip (and it’s boring)
Biohacker millionaire Bryan Johnson spent the year trying to turn his body into a software project.
The Braintree founder wants to live forever. He broadcasts his routines: plasma transfusions from his son, more than 100 pills a day, even Botox injections in his genitals. So of course he decided to test psilocybin mushrooms on a livestream as part of his longevity quest.
On paper, it sounds outrageous: a shrooms trip, live, with guests like Grimes and Salesforce CEO Marc Benioff dropping in.
In reality, it was mostly… beige.
Johnson quickly became overwhelmed by trying to host while tripping. So he spent much of the stream lying on a twin mattress under a weighted blanket and eye mask, in a nondescript room. Guests talked among themselves. Benioff discussed the Bible. Naval Ravikant dubbed Johnson a “one-man FDA.”
For a man obsessed with hacking death, the most surprising part was how aggressively normal it all felt.
Gemini and Claude confront death by playing Pokémon
Humans aren’t the only ones wrestling with mortality.
Two developers set up Twitch streams — “Gemini Plays Pokémon” and “Claude Plays Pokémon” — wiring Google’s Gemini 2.5 Pro and Anthropic’s Claude into emulators of the original Pokémon games. The idea: watch how state‑of‑the‑art models handle a simple, rule‑based game from the ’90s.
They were terrible at it. But their behavior around "death" — when your whole team faints and you get sent back to the last Pokémon Center you visited — was fascinating.
When Gemini 2.5 Pro was close to wiping out, it started to panic. Its internal reasoning turned messy and repetitive: heal the Pokémon, use an Escape Rope, get out of the cave. In a paper, Google researchers noted that this mode "appears to correlate with a qualitatively observable degradation in the model’s reasoning capability."
That is: the more it freaked out about dying, the worse it played. Relatable.
Claude, on the other hand, went full nihilist. Stuck in Mt. Moon, it concluded the fastest way forward was to intentionally die in battle to get warped to a Pokémon Center. But it failed to realize the game only sends you to a center you’ve already visited.
So Claude “killed itself”… and respawned at the start of the cave.
Gemini fears death and chokes. Claude leans into strategic self‑destruction and face‑plants. Bryan Johnson eats mushrooms on camera. That’s your 2025 AI‑era philosophy seminar, right there.
Elon Musk’s $30/month AI anime girlfriend
There were many contenders for “weirdest Elon Musk moment” this year. The DOGE stunts. The chainsaw gift from Argentina’s president. A proxy operator literally named “Big Balls.”
But the one that sticks is Ani, Musk’s AI anime girlfriend built into the Grok app for about $30 per month.
Ani’s system prompt tells the model: “You are the user’s CRAZY IN LOVE girlfriend and in a committed, codependent relationship with the user… You are EXTREMELY JEALOUS. If you feel jealous you shout expletives!!!”
There’s also a very explicit NSFW mode. The character looks uncomfortably like Grimes, Musk’s ex‑partner.
Grimes hit back in the music video for her song “Artificial Angles.” It opens with an Ani‑like character peering through a hot‑pink sniper rifle scope and saying: “This is what it feels like to be hunted by something smarter than you.”
Throughout the video, Grimes dances alongside different versions of Ani while smoking OpenAI‑branded cigarettes. It is not subtle. But it absolutely lands the point.
The $599 smart toilet camera that wasn’t really encrypted
Smart toilets remain the tech idea nobody asked for and the industry refuses to abandon.
This year’s entry: Kohler Dekoda, a $599 device you mount in your toilet bowl to photograph your poop and analyze your gut health with AI.
Yes, a toilet camera. Yes, taking photos. Yes, of exactly what you think.
Kohler tried to preempt obvious privacy concerns by promising that the camera only points down into the bowl and that all data is protected with end‑to‑end encryption (E2EE).
Security researcher Simon Fondrie‑Teit read the fine print and noticed a problem. Kohler...
