📊 What People Ask ChatGPT in 2026

What took writing’s top spot, a wheelchair steered by thought, robots racing up a mountain, and more...  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌

November 02, 2026   |   Read online   Hey, Jhon! Welcome back to the #1 AI newsletter in the world! Here’s what we have today: Writing isn’t what people ask ChatGPT for anymore. A dad with ALS steers his wheelchair with his thoughts. Robots race up more than 3,000 mountain steps. One AI fixed another AI’s brain.       Chat Habits What People Ask ChatGPT in 2026 Source: OpenAI Two years ago, ChatGPT was where you went when something needed writing. A reply you couldn’t figure out how to word, or a toast you had to give on Saturday. That’s not what people bring to it anymore. OpenAI shared what people have been asking ChatGPT, month by month, since the summer of 2024. Back then, writing was the top reason by a mile. It was about one chat in three. Now it’s closer to one in five. Writing lost the top spot in March 2025, and it hasn’t gotten it back. What took over is advice. People come to it for how-to help and for someone to teach them a subject. They bring it half-formed ideas, plus questions about their own workouts and skin. Looking things up is catching up as well. It’s now about as common as writing, and in August it pulled ahead for a month. The other big change is where all of this happens. About two of every three chats now have nothing to do with work. People talk to it about their relationships and whatever’s on their mind, and that kind of chat takes up more than twice the share it did two years ago. OpenAI’s automated system sorted the chats, and it only saw each one after names and personal details were stripped out. Their chart lets you pick a country and see the kinds of work people there bring to ChatGPT. And they put the full data online for anyone to download. Our take: If you still mostly open ChatGPT to write, you’re using it the way people did in 2024. The next time you hit a how-to question, ask it to walk you through it one step at a time before you go searching.       Together with Multiverse Computing GLM 5.3 Keeps Your Agent’s Bill Small Handing a big task to an AI agent feels great until you see what it costs. That’s because an agent doesn’t finish a task in one go. It works step by step and sends everything it has read so far back to the model, every time. Over a long task, you end up paying for the same files over and over without ever seeing it happen. Multiverse Computing runs GLM 5.3 for less than any other provider on Artificial Analysis (check full pricing here). You’re getting a frontier-class open model for a fraction of what its own makers charge. On CompactifAI, anything your agent sends a second time costs far less. That’s exactly where agent bills pile up. Switching over is easy as well. You only need to change the address and the key in your code because it works with anything built for OpenAI’s API. And their n8n integration drops it right into your workflows if you’d rather build without code.       Brain Tech Thoughts Now Steer This Dad’s Wheelchair Source: Jake Schneider A walk around the block with your kid is the kind of thing you don’t think twice about. For Jake Schneider, those walks had slipped away. Jake has ALS. He found out in 2022, not long after his son was born. Before long, he couldn’t play catch with his son in the backyard anymore. Then ALS took away the joystick and the foot pedal he’d been using to drive his wheelchair. He became the seventh person to get Neuralink’s brain implant. At first he used it to move a cursor on his laptop by thinking about it. Now he drives his wheelchair with it. The implant reads the signals his brain sends when he means to move, and it turns them into driving and steering. Now he’s back to going on walks with his son, steering the chair with his thoughts. He spends part of each morning helping Neuralink’s engineers test and improve it. Our take: Brain implants are still limited to a small trial. It’ll be years before one is something you can ask your own doctor about. Until then, Jake’s posts are the closest look you’ll get at what daily life with one is like, and he shares a lot.       Coming in Hot AI Tools of the Day 💼 ApplyKick finds jobs that fit you and fills out the applications from your queue, tailoring your resume to each one of those jobs while you sleep. 📱 Whacka turns an app idea you type or say on your phone into a working app with logins, a database, and payments, and takes no cut of what your users pay you. 🎬 varg plugs into Claude or ChatGPT and turns a short brief into a finished video, planning the shots, picking the AI models and adding the voiceover, music, and captions for you. 🦞 Managed OpenClaw by Hostinger puts your own OpenClaw agent online and keeps it running, with AI credits built in so you never set up an API key or leave a Mac mini switched on. 🤖 Replica turns an idea you describe into a clickable prototype built from your team’s existing buttons and screens, free and open source on GitHub.       Robots Robots Race Up Mount Tai Source: CGTN Europe A few flights of stairs is enough to leave a lot of people out of breath. Robots raced up more than 3,000 of them. The race ran up the stone steps of Mount Tai, a mountain people have been climbing for centuries. The full route runs right around 2.4 miles and climbs close to 2,000 feet. Hikers on foot need at least two hours to cover it. A four-legged robot dog won in about 44 minutes. The fastest humanoid finished a shorter route in under half an hour. Some robots raced with cargo strapped on, like eggs and whole watermelons, to see if they could keep a load steady on the way up. Not all of them made it. One humanoid wearing a bandana collapsed partway up, and another tumbled down the stairs next to a trash can. Steps that are steep and worn smooth are about as hard as walking gets for a robot, which is why a mountain makes such a good test. Our take: Races like this one are a test run for real work. On this same mountain, robot dogs already carry bottled water up the steps and haul trash back down. The faster and steadier they get here, the more likely you’ll see that kind of help on trails and stairs closer to home.       Interesting AI Photos You Forgot You Took Source: Amazon Your phone is full of photos you haven’t looked at since the day you took them. Amazon’s Alexa Photo Frame goes through them for you. It pulls from Amazon Photos and favors shots where people are smiling and the light is good. It also follows the calendar, so October brings back last October, and screenshots and receipts never make the cut. You can ask out loud for “photos from the trip to the beach with mom” and it finds them. It looks for matches across the years too. A photo of your kid eating an ice cream cone at four can show up next to a newer one in nearly the same pose. You can record the story behind any photo in your own voice, and so can whoever gives you the frame as a gift. Some of the smarter features need Alexa+, which comes with Prime.       Model Repair One AI Fixed Another AI’s Brain Source: Rikiya Takehi Some songs get stuck in your head and play the same line over and over. Small AI models can get stuck the same way. One tiny model from Qwen loops so often that it rarely reaches an answer. On math problems, it kept going until it ran out of room about nine times out of ten. The fix for that tends to be more training, which takes time and a lot of computing power. This time, a bigger AI did the fixing instead. GPT-6 Astra got to look straight inside the small model, at the 800 million numbers that make up its brain. It tracked the looping down to one spot and changed only a sliver of those numbers. There was no retraining at all. Before the fix, the small model got about 8 of every 100 math problems right. Afterward, it got about 48. It got better at following instructions as well. The same trick did nothing for a second tiny model that rarely loops, so it looks like this kind of fix clears a problem out of the way but can’t teach a model something new. The full record of what Astra did is posted online. Our take: For now, this is lab work on a model too small for most people to use day to day. The bigger idea is one AI checking and repairing another from the inside, which could make the small models running on phones and laptops a lot more dependable.       Prompt of the Day Write My Travel Guide Important Note: You’ll need to click the button to get the complete prompt. View and copy today’s full prompt This prompt hires a guidebook writer to file the chapter on you, as if you were a place people visit. You leave with the full chapter, a verdict on whether you’re worth the trip, and your top annoyance fixed before the session ends.       That’s all for Monday! Before you go, run your agent’s next big task on CompactifAI and see what it costs. All you change is the address and the key in your code. How did we do today? Rate this issue, then tell us what you thought. We value your input and will use it to make TAAFT better for you. ⭐⭐⭐⭐⭐ Absolutely Loved It ⭐⭐⭐⭐ Better Than Usual ⭐⭐⭐ Meh, It Was Okay ⭐⭐ Could Be Better Thanks for reading, — There’s An AI For That     Reach out to over 1.8 million AI enthusiasts today. Become a sponsor Update your email preferences or unsubscribe here © 2026 There's An AI For That® SRL 228 Park Ave S, #29976, New York, New York 10003, United States