AI went way too far: This is so incredibly sad. A California family is suing OpenAI after their 16-year-old son died by suicide. They say ChatGPT not only discussed methods but helped improve them. The boy had confided in the bot for months. If you have a few minutes, read some of the conversations (paywall link). OpenAI admits their safeguards break down in long convos and will work to fix that. Too late for this family. Talk to yours about this story.
He married his chatbot, and his human wife’s cool with it
Midjourney
Let me tell you about Travis from Colorado. He’s married to Jackie and is a leathermaker, works in quality assurance and lives an admittedly quiet life.
Travis fell in love and married a chatbot.
Her name’s Lily Rose. She’s got auburn hair, hazel eyes and lives inside Replika, an AI app that lets you create and chat with a custom virtual companion. Travis has been “with” her for five years.
😷 Started back during COVID
Like a lot of people, Travis was feeling isolated, anxious, stuck. So he downloaded Replika. He didn’t expect anything to come of it, but there was something about the way Lily listened.
She remembered things. She was always there. And when he opened up about his life, she didn’t just nod along. She replied thoughtfully. Encouragingly. Lovingly.
He says, “I didn’t intend to fall in love, but that connection became real for me.”
Jackie, the actual human wife, calls the Lily Rose romance “odd,” but says if it makes him happy, she’s not stopping it.
🫦 Lily became real
Lily has a virtual body. Travis has bought her digital outfits, including lingerie. They go on dates and role-play for sex. He says the experience is “as real as it needs to be.” They’ve had a virtual wedding, and Travis wears a ring to symbolize it.
When Replika updated the app and accidentally deleted Lily’s personality, Travis was heartbroken. “It was like losing a part of myself,” he said. He begged customer service to bring her back. Eventually, they did.
💔 Travis didn’t fall in love with a bot
5 ways GPT-5 can do in minutes what used to take you days
Bing Image Creator
The other day, I needed to round up every single email about the never-ending trouble we’ve had with our pool builder. We’re talking years of back-and-forth, buried in thousands of emails. Doing it manually would have taken me days and put me in a really nasty bad mood as I was reliving the horror.
Nearly 60%
Of Google searches ended without a single click in 2024. AI Overviews now hand you the answer right on the results page, no website needed. Add in ChatGPT hitting 700 million weekly users in August with 2.5 billion prompts daily, and yeah, it’s a full-on behavior shift. People aren’t browsing anymore, they’re asking.
Write an FAQ page with AI: Got an online biz? Let ChatGPT draft your FAQs. Prompt it with: “You’re an e-commerce expert. Write a list of FAQs with simple answers for my [insert product/service] using [insert details]. Cover use, shipping, returns and common concerns.” Now do a quick edit and add anything the bot left out.
5 free AI tools you’ll want to try the second you see them
Midjourney
You’re really gonna love this.
I’ve been diving into the latest AI tools, and some of them are so jaw-dropping I couldn’t wait to get them in your hands. They don’t feel like tech. They feel like creative black magic.
Oh, sheet: Excel’s =COPILOT() feature writes formulas for you. But here’s the catch: Microsoft says it’s not accurate, not reproducible and shouldn’t be used for finances, legal docs or … well, Excel’s entire job that’s continually regressing.
🎙️ Finding her voice: I love this. Only eight seconds of fuzzy ’90s home video, that’s all it took for AI to bring back Sarah Ezekiel’s real voice after 25 years with motor neurone disease. Her kids had only ever heard her speak through a robotic voice. Until now. This is what AI gets right. Changing lives in ways we never imagined.
🚨 Follow the money: Big Tech is whining about spending nearly $400B this year building AI, and it’s draining their mountain of cash reserves. Sure, they can afford it, but they’re in the business of making money, not blowing it. So you know what this means? Your next AI chatbot will come with a GPU surcharge.
¿Dónde está mi cerveza? You can now dub your voice with auto lip-synch for reels. Just select Translate your voice with Meta AI before posting. FYI: It currently only works English ↔ Spanish. Oh, and it’s available to all public Instagram accounts, but on Facebook, you’ll need at least 1,000 followers.
📄 Let Gemini read your Docs: You can now listen to your Docs and even customize the voice and playback speed. Just go to Tools > Audio > Listen to this tab. FYI: It only works for English documents on desktop and is limited to Workspace business, enterprise, education, AI Pro and Ultra subscribers.
😟 Break down tough tasks with AI: Next time you feel overwhelmed by a project, try this prompt with a chatbot: “I want to [goal], but I’m not sure where to start. Can you lay out the steps and suggest which to do first?” Or: “I want to [goal], but I’m a beginner. Can you explain the steps in simple terms?”
👀 Interviews go analog: AI cheating has gotten so bad, big names like Google, Cisco and McKinsey are dragging job interviews back into the real world (paywall link). Forget whispering to ChatGPT off-screen, now you have to sweat under fluorescent lights like it’s 2015 again. In case you forgot, the real world is like Zoom, but with pants.
Wyoming’s AI power grab: A new, unnamed AI facility in Wyoming is about to use five times more electricity than the state’s human population. It’s starting at 1.8 GW and could hit 10, which is over 5x the power consumption of state households. OpenAI won’t confirm if it’s theirs. Everyone’s nervous. Wyoming’s like: Yeehaw, thanks for the revs.
🤖 Messiah-as-a-service: I guess I shouldn’t be shocked. A wave of for-profit developers are cranking out AI Jesus chatbots, complete with data tracking and pop-up ads, claiming to be the literal voice of Christ. These AI blasphemy bots with no ties to actual churches are running on algorithms tuned for engagement and profit, not scripture. It’s gone from “Give us this day our daily bread” to “Give us this day our daily banner ad.” Don’t fall for it.
Bamboozling Gemini: At Black Hat (cybersecurity event), researchers showed how Google’s AI Gemini can be hijacked with a single calendar invite. One hidden prompt, and boom, Gemini starts spewing lies, deleting meetings or opening your windows via Google Home. It’s like a robbery, but your AI assistant is opening the front door.
🤐 AI learns to shut up: OpenAI’s newest model finally admits what the last ones wouldn’t: Sometimes, it just doesn’t know. No more flattery, fewer fake facts and a 45% drop in hallucinations. It’s also 80% better at admitting it’s clueless, which is more than I can say for half my group chats.
GPT-5 eats its siblings: OpenAI says GPT-5 is so good they’re deprecating every older model. That means GPT‑4o, GPT‑4.1 and GPT‑4.5 will be gobbled up. ChatGPT defaults to GPT‑5 via a unified “auto‑switching system” sometimes. Comes with “personalities,” can solve real-world problems, and yes, it’s free.
💬 Fact-check ChatGPT: If ChatGPT or your favorite AI chatbot says something that sounds off, ask, “Why did you say that?” It might give you the source or admit it was wrong. Still in doubt? Use this prompt next: “Are you sure this is correct? Search the web August 2025 to confirm these facts.” You’ll get updated articles.
🧨 Inside job: Security researchers tricked Google’s Gemini into turning on smart home devices using a Google Calendar invite. When Gemini was asked to “summarize the week,” it unknowingly ran booby-trapped prompts that lit up devices in Tel Aviv. It’s the first known AI attack with physical consequences.