Actually Tested AI

Things AI Is Genuinely Bad at That Nobody in Tech Wants to Admit I spent eleven minutes last Tuesday watching ChatGPT confidently explain how to fix a plumbing issue that would have flooded my bathroom. The instructions were clear, well-formatted, and completely wrong about which valve to turn first. This wasn't a hallucination in the traditional sense — the AI wasn't making up facts. It just had no idea what "wrong" meant in a context where water pressure and pipe age actually matter. The Confidence Problem Is Worse Than the Accuracy Problem Everyone in AI circles talks about hallucinations. Models making up citations, inventing court cases, fabricating statistics. That's real, and it's a problem. But here's what two years of daily testing taught me that nobody mentions at conferences: the stuff AI gets subtly wrong is far more dangerous than the stuff it gets obviously wrong. When ChatGPT invents a fake Supreme Court case, you catch it. W...

ElevenLabs AI Voice Tool: What It Actually Is and When to Use It I spent three hours last Tuesday making my dead grandfather's voice read a birthday card to my mom. That's not the intended use case for ElevenLabs, but it worked, and it completely broke my brain about what this tool actually is. What ElevenLabs Actually Does (And Doesn't Do) It's a text-to-speech tool. That's the short version. You type words, it speaks them in voices that don't sound like robots. But that description undersells what's happening here. The voices have breath. They pause in weird places sometimes, like a person thinking. They emphasize words you didn't ask them to emphasize. It's unsettling the first time you hear it because your brain keeps trying to find the robot tell. With most of the voices, you won't find it. What it doesn't do: edit audio, generate music, transcribe speech, or work offline. It's specifically speech synthesis. You...

How to Use AI to Write a Cover Letter That Doesn't Sound Desperate I submitted 23 job applications last month using AI-generated cover letters, and the first batch was so cringeworthy that I deleted them from my sent folder out of embarrassment. The problem wasn't that ChatGPT wrote bad cover letters. The problem was that it wrote cover letters that sounded like a golden retriever begging for a treat. You know the tone. "I would be absolutely thrilled to contribute my skills to your dynamic team." "This opportunity perfectly aligns with my lifelong passion." "I am confident I would be an exceptional addition." Reading them back felt like watching myself grovel. I started testing different prompts to see if I could get AI to sound like a competent adult instead of someone who really, really needs this job. The Default AI Tone Is the Problem Here's what I figured out after about 40 minutes of iteration: ChatGPT's default cover...

How to Use AI to Write Cold Emails That Don't Sound Like AI Wrote Them Last month I sent 47 cold emails. The first 23 were written with ChatGPT using the obvious approach — paste in context, ask for an email, copy the output. I got two responses, both from people politely asking me to stop emailing them. The next 24 used a completely different method I stumbled into by accident. Nine responses. Three actual calls booked. Something changed between batch one and batch two, and it wasn't the recipients or the timing. It was how I was using the AI. I'd been doing it backwards the whole time. The Default Approach Is a Trap Here's what I did wrong for those first 23 emails. I gave ChatGPT exactly what every tutorial tells you to give it: the recipient's company, their role, what I was selling, what I wanted them to do. Then I asked for a cold email. The outputs were technically good. Professional tone. Clear value proposition. Proper structure with a hoo...

I Used AI to Track My Spending for 90 Days — Here's What I Found On day 47, ChatGPT told me I'd spent $312 on "convenience purchases" that month. I had no idea what that meant until I scrolled back through our conversation and realized it had created its own category for all the times I paid extra to avoid minor inconveniences. Uber instead of the bus. DoorDash instead of cooking. The $4 oat milk latte instead of making coffee at home. Nobody asked it to do that. It just noticed a pattern I'd been blind to for years. The Setup That Actually Worked I tried three different approaches before landing on something sustainable. First attempt: copying and pasting my bank statements directly into ChatGPT. Lasted four days. Too tedious, and I kept forgetting. Second attempt: taking screenshots of my banking app and uploading them to GPT-4 with vision. Better, but it kept misreading amounts. A $47.23 dinner became $4,723 once, which made my weekly summary lo...

AI Tools That Work Well When English Is Not Your First Language My friend Priya sent me a screenshot last month of ChatGPT completely misunderstanding her question because she'd structured it the way you'd phrase something in Hindi. The AI gave her a confident, detailed answer to a question she didn't ask. She's fluent in English — works at a tech company in Bangalore — but the way native speakers phrase prompts isn't how everyone naturally thinks. I've been testing AI tools with non-native speakers for about three months now. Not in a lab. Just helping friends, watching over shoulders, collecting the moments where things break. What I found surprised me: the best tools for non-native speakers aren't always the ones with the best "multilingual support" badges. The Tools That Actually Handle Different Phrasing Claude turned out to be the unexpected winner here. I tested it with prompts that had grammar mistakes, word order issues, an...

Using AI to Declutter Your Home Room by Room: What Worked and What Did Not I was standing in my garage surrounded by seventeen years of accumulated whatever when I thought: maybe ChatGPT can just tell me what to throw away. Not in some philosophical "spark joy" sense, but literally look at this mess and give me a plan. So I spent three weekends testing whether AI could actually help someone declutter their home, room by room, with real results. Some of it surprised me. Most of it didn't. The Setup: What I Actually Asked ChatGPT to Do I started with photos. Took pictures of every room, every closet, every drawer that made me feel vaguely ashamed. Uploaded them to ChatGPT-4 with Vision and asked for specific decluttering recommendations. Then I tried a different approach: no photos, just detailed descriptions of what I owned and how I used each space. I wanted to know which method gave better advice. The photo approach was immediately more useful. When I des...

How I Used AI to Negotiate a Higher Salary and What Actually Worked The HR rep said "we're prepared to offer you $78,000" and I had about three seconds to respond before the silence got weird. Except I'd spent the previous two weeks having ChatGPT tear apart every possible version of this conversation, so I already knew exactly what to say. I ended up at $87,000. Not some mythical 40% jump, but $9,000 more than their first number. I'm going to walk through exactly how I used AI to prepare, what bombed completely, and the one thing that actually made the difference when I was sitting across from a real human who controlled my financial future. What I Thought Would Work vs. What Actually Happened I went in thinking AI would be most useful for research. Market rates, company financials, competitive analysis. The stuff you'd normally spend hours googling. Wrong. That part was fine but not special. Anyone with 20 minutes can find salary ranges on ...

Grok AI vs ChatGPT: I Tested Both for a Full Week — Here's the Real Difference Tuesday night, 11pm, I asked both Grok and ChatGPT to explain why my sourdough starter smelled like acetone. ChatGPT gave me a thoughtful, organized response about acetic acid and feeding schedules. Grok told me my starter was "going through its rebellious phase" and then actually explained the chemistry. That moment basically summarizes the entire week I spent switching between these two. The Setup: How I Actually Ran This Test I didn't do the typical "ask both the same 10 prompts and compare" thing. That's useless. Instead, I used both as my primary AI assistant for seven days straight — alternating which one I opened first each morning. Real tasks only. Writing help, research, code debugging, random questions that popped into my head at 2am. I kept a running doc. Not formal notes, just quick reactions. "Grok faster here." "ChatGPT understood c...

I Tried Using AI to Manage My Anxiety for 30 Days — Here's What Actually Happened Day three, 2 AM, and I'm having a full spiral about a client email I sent six hours earlier. Instead of texting my friend who's definitely asleep, I opened ChatGPT and typed: "I think I ruined a professional relationship and I can't stop thinking about it." What happened next surprised me more than I expected. Why I Even Tried This I've been in therapy on and off for years. It works, but it's expensive and my therapist doesn't text back at 2 AM. I'd seen people online talking about using AI as a kind of mental health companion, and my immediate reaction was skepticism bordering on dismissal. Seemed like a tech bro solution to a deeply human problem. But I was curious. And honestly, my anxiety had been spiking badly since taking on more freelance work. More unpredictability, more waiting for responses, more catastrophizing about every email. I figur...

How to Use AI to Write Your LinkedIn Bio Without Sounding Like Everyone Else I spent three hours last Tuesday generating LinkedIn bios with ChatGPT, and about 90% of them made me want to delete my entire professional existence. "Passionate leader driving innovation" showed up so many times I started keeping a tally. By attempt 47, I finally figured out why most AI-generated bios sound identical — and more importantly, how to make yours not. The Default AI Bio Problem Here's what happens when you type "write me a LinkedIn bio" into ChatGPT with no other context. You get something like: "Results-driven marketing professional with 10+ years of experience leveraging data-driven strategies to deliver impactful solutions." I actually generated that exact sentence structure four times in a row using slightly different prompts. The AI isn't broken. It's doing exactly what it's trained to do — synthesize the most common patterns from...

What I Learned Using AI to Reply to Every Work Email for Two Weeks Monday morning, 47 unread emails, and I'm staring at my inbox thinking about how much of my life I've spent typing "Thanks for the update, I'll circle back on this." So I did something stupid: I committed to letting ChatGPT draft every single work email response for the next two weeks. Every. Single. One. I thought I'd either become incredibly efficient or get fired. Neither happened. But something weirder did. The Setup Wasn't As Simple As I Expected First problem: you can't just paste an email into ChatGPT and say "reply to this." Well, you can, but you'll get something that sounds like a customer service bot wrote it. I spent the first day refining my prompt until I had something that actually worked. Here's what I landed on: I'd paste the original email, add two sentences about my actual position on the topic, and tell it my relationship to th...