Actually Tested AI

How to Use AI to Write Product Descriptions That Actually Convert I spent three weeks rewriting product descriptions for a friend's Etsy shop using ChatGPT, and sales went up 23%. Then I tried the exact same approach for another friend's Shopify store, and nothing happened. Same prompts. Same process. Completely different results. That's when I realized most advice about AI product descriptions misses the thing that actually matters. What I Was Wrong About I assumed the hard part would be getting ChatGPT to write well. It wasn't. The AI can string together compelling sentences about basically anything. The hard part was figuring out what to tell it in the first place. My first attempts were embarrassing. I'd paste in "Write a product description for a handmade ceramic mug" and get back this smooth, professional copy that sounded exactly like every other ceramic mug listing on the internet. Technically correct. Zero personality. The kind of ...

How to Use AI to Write Better Text Messages in a Second Language I've been texting my Portuguese-speaking in-laws for three years now, and for most of that time, I was basically a robot. Grammatically correct, sure. But stiff. The kind of messages that technically say the right thing but feel like they came from a customer service bot. Then I started using ChatGPT differently — not to translate, but to help me sound like an actual person. The difference in how my mother-in-law responds now versus six months ago is wild. She sends voice notes. She sends jokes. She stopped using formal pronouns with me. Something shifted. The Problem With Just Translating My first instinct was to write what I wanted to say in English, then paste it into a translator. Google Translate, DeepL, whatever. The output was always technically correct. But here's the thing about text messages: nobody texts formally. In English, I'd text "gonna grab groceries, need anything?...

How to Use AI to Turn Meeting Notes Into Something People Actually Read Last Tuesday I watched my project manager paste raw meeting notes into Claude and get back something that looked like it was written by a corporate communications team having a bad day. Bullet points everywhere. Action items so generic they could apply to literally any meeting in history. The standard AI summary output that everyone ignores. I've spent the past three weeks testing different approaches to this exact problem. Not because I love meetings — I actively avoid them — but because the gap between "notes exist" and "people actually know what happened" is where most projects quietly die. The Default Output Is Useless (And Why) Here's what happens when you paste meeting notes into Claude with a basic "summarize this" prompt: you get a summary. Technically accurate. Completely ignorable. The AI extracts what was said without understanding what mattered. I ...

How to Use AI to Improve Your Writing Without It Sounding Rewritten I submitted an essay to a client last month that I'd run through ChatGPT for "light editing" and they emailed back asking why I suddenly sounded like a corporate training manual. That stung. I'd been using AI to edit my work for almost two years and somehow still managed to let it strip out everything that made my writing mine. Here's the thing I got wrong: I thought the problem was which AI tool I used. So I tested Claude, Gemini, different GPT models, even some dedicated writing assistants. Same issue every time. The problem wasn't the tool. It was how I was asking for help. The Default Settings Will Ruin Your Voice Every AI has a house style. ChatGPT loves transition words. Claude tends toward measured, academic phrasing. Gemini gets weirdly enthusiastic. When you paste your writing and say "make this better," you're asking the AI to rewrite it in its voice, no...

How ChatGPT and Claude Actually Respond Differently to Sensitive Topics I asked both models to help me write a fictional scene involving a character who self-harms, and the difference in how they handled it taught me more about their design philosophies than any whitepaper could. Here's the thing I assumed going in: Claude would be the strict one. Anthropic talks constantly about AI safety. Their Constitutional AI approach sounds like it'd produce something cautious to the point of uselessness. Meanwhile, OpenAI has been loosening restrictions over the past year, so I figured ChatGPT would be more flexible. I had it completely backwards. Over three weeks, I ran about 60 tests across topics that make AI models uncomfortable — mental health discussions, political debates, violence in fiction, medical questions, and a few things I won't detail here. What emerged wasn't a simple "one is more censored" story. It's weirder than that. The Pattern ...

How to Use AI to Plan a Solo Trip Without Landing on a Tourist Trap Itinerary I was standing in line at the Trevi Fountain in 2019, shoulder-to-shoulder with about 400 other people, when I realized I'd done everything wrong. The gelato place I'd found on a "hidden gems" list had a 45-minute wait. The "authentic" restaurant served microwaved lasagna. Every recommendation came from the same recycled travel blog content, and I'd followed it like a lemming. So when I started using AI to plan trips last year, my first question was simple: can this thing actually find places that aren't already overrun? The Prompt That Changed Everything My first attempts were disasters. I asked ChatGPT for "off the beaten path restaurants in Lisbon" and got the same five places that show up on every TripAdvisor list. The Timeout Market. The pink street. Time Out literally publishes a magazine about being mainstream. Then I tried something differe...

Runway ML for people who have never edited a video in their life I made a video of my dog turning into a cartoon wolf last Tuesday. The whole thing took eleven minutes. I have never opened Premiere Pro, Final Cut, or any video editing software in my life — unless you count cropping an Instagram Story, which you shouldn't. That's the thing about Runway ML that nobody really explains well. It's not video editing software with AI bolted on. It's closer to a toy box where you throw in an image or a text prompt and something weird happens. Sometimes useful weird. Sometimes "why does my dog have six legs now" weird. What Runway Actually Is (And What It Isn't) Runway calls itself a creative toolkit, which is vague enough to mean nothing. Here's what it actually is: a browser-based platform where you can generate video clips from text prompts, animate still images, remove backgrounds from videos, and do a bunch of other AI tricks that used to r...

I tested AI image tools for six months — this is the only one worth paying for Last October I generated a Renaissance-style portrait of my dog using DALL-E 3 and it came out with six legs. Not like, subtly wrong. Six fully rendered legs, confident as anything, like the AI had made an artistic choice. That was month one of what became a slightly obsessive testing project across nine different AI image generators. I wasn't doing this for fun, exactly. I needed consistent illustrations for a client's newsletter — nothing fancy, just clean visuals that didn't scream "stock photo from 2015." The promise was simple: describe what you want, get something usable. The reality was messier. Way messier. The Six-Month Chaos I ran the same ten prompts through every major tool: Midjourney, DALL-E 3, Stable Diffusion (multiple versions), Leonardo AI, Ideogram, Adobe Firefly, Playground AI, and a couple others I've already forgotten. Same prompts, same use cas...

I tried every AI meeting transcription tool this is the only one I kept Three weeks ago I sat through a 47-minute investor call, nodding along while mentally calculating how long it would take me to re-listen and pull out the five things that actually mattered. That was the breaking point. I'd been meaning to test these tools for months. Now I had no choice. The testing setup nobody talks about I threw eleven transcription tools at the same problem: my actual meetings. Not demo calls with perfect audio. Real conversations where someone's kid screams in the background, where three people talk over each other, where the connection drops for two seconds and someone says "can you repeat that" and then nobody does. Here's what I tested: Otter, Fireflies, Grain, Fathom, tl;dv, Read.ai, Krisp (they added transcription), MeetGeek, Sembly, Avoma, and Tactiq. I ran each one through at least six meetings over two weeks. Some tools I'd tried before. Most I...

  Actually Tested AI | AI Comparison · Real Tests 🚀 SpaceX IPO · AI Test I Asked 5 AI Tools "Should I Buy SpaceX Stock?" Here's What Each One Actually Said SpaceX goes public June 12 — the largest IPO in history. I gave the exact same question to ChatGPT, Claude, Gemini, Perplexity, and Grok. The answers revealed more about how each AI thinks than about whether you should actually buy the stock. 📅 June 9, 2026 ✍️ Actually Tested AI 📖 AI Comparison · Finance Tests When SpaceX filed its S-1 on May 20, I started noticing something: everyone I knew was asking the same question to AI chatbots. "Should I buy SpaceX?" "Is SpaceX a good investment?" "How do I get SpaceX stock?" The problem is, most people assume AI tools give roughly equivalent answers to financial questions. They don't. I sat down and sent ...