The Big Story

A doctor in China recently used ChatGPT to solve a math problem that had gone unsolved for decades. The story shows how a regular person with no special coding skills can turn to an AI chat tool for help on something genuinely difficult. Think of it like having a super-patient tutor who never gets tired and can look at a problem from every angle at once. The doctor kept asking questions and refining ideas until the AI helped land on a solution that worked. This matters because it shows AI isn't just for writing essays or making pictures — it can team up with human curiosity on hard problems in science and math. For students, it hints at a future where homework help or project ideas could come from chatting with an AI instead of struggling alone. If you're someone who freezes up on tough questions, this example proves you can break them down step by step with an AI partner. You don't need to be an expert to start; you just need to keep asking and checking the answers. Try opening ChatGPT or Claude on your phone or computer and paste in a math or science question that's been bugging you. Ask it to explain its thinking one step at a time and see where the conversation takes you. Start with something from your current class, like a geometry proof or a physics equation, and watch how the back-and-forth builds toward an answer. Many teens already use these tools for quick facts, but treating them like a study partner changes the experience into something more like a real conversation that builds understanding over multiple turns. The key is to treat the AI response as a starting point you verify, not the final word, which keeps the process honest and educational. This approach works on any device with a browser, so you can experiment during a study break without needing special software. Source: uniladtech.com

Explain Like I'm 14

You know how when you play a video game, you sometimes keep a notebook of what worked and what didn't so you don't make the same mistake twice? Now picture an AI agent doing the exact same thing, except instead of a notebook it builds its own wiki page filled with both wins and fails. Every time the agent tries a task, it writes down what went right, what went wrong, and why. Later, when it faces a similar task, it checks that wiki first instead of starting from zero. Bigger models get even better at using the notes, but even smaller models suddenly perform closer to the big ones because they're not forgetting everything after each try. The result feels like the AI is slowly learning from experience the way a person does after practicing something many times. It turns out the secret isn't magic — it's just saving the lessons so they can be reused. The wiki structure lets the agent store details in plain language that any future run can read, turning isolated attempts into a growing library of knowledge. Over repeated tasks the entries become more useful because the agent adds context about why one approach succeeded where another failed. This method helps close the gap between small and large models by giving the smaller ones a way to carry forward what they learned instead of resetting completely each time. In practice it means the AI spends less time repeating errors and more time building on earlier progress, which is exactly how people improve at skills like playing an instrument or solving puzzles. Source: the-decoder.com

Cool Stuff & Try This

Run your own private chatbot at home Imagine having an AI helper that lives only on your computer and never sends your chats to any company. That's what running a local language model lets you do. It keeps everything private and works even without internet once it's set up. This is perfect if you want to experiment with AI without worrying about data leaving your device. Head to the Wired article linked below and follow the simple steps they give for installing a model on a regular laptop. Once it's running, try asking it to write a short story about your favorite game or explain a homework topic in plain language. The setup usually starts with downloading a small model file that fits on most modern laptops, then using a free app to load and chat with it. Because the model runs locally, your questions stay on your machine, which removes concerns about privacy that come with cloud services. After the first successful chat you can test it on creative tasks like generating character ideas for a story you're writing or summarizing a long article you found online. The Wired guide walks through choosing a beginner-friendly model size so the process stays manageable even if you've never installed software like this before.

ChatGPT now remembers things even in temporary chats You can now turn on memory in a temporary chat so the AI still recalls details you saved earlier without keeping a permanent record of the whole conversation. It feels like chatting with someone who knows your preferences but forgets the rest once you close the window. This is handy when you want helpful context without leaving a long trail. Open ChatGPT, start a temporary chat, and ask it to use one of your saved memories on purpose — like reminding you of a project you're working on. See how natural it feels compared to starting fresh every time. The feature lets you test how memory changes the flow of a conversation while still keeping the session private overall. For example, you could ask it to reference a book you mentioned in a previous saved memory and then watch it weave that detail into the current discussion without saving the new exchange. This balance gives you the benefit of continuity for short tasks without the full history that regular chats store. Source: wired.com Source: notebookcheck.net

Quick Bits

AI chatbots vs search engines for spotting propaganda New research suggests chatbots can sometimes do a better job than regular search engines at helping people recognize foreign propaganda. The key difference is that chatbots can explain why something might be misleading instead of just showing links. It's a small but useful shift in how we might check information online. The NPR report highlights that the explanatory style of chatbots gives users more context to judge claims on their own. This could matter for anyone scrolling social media who wants a quick way to understand why a headline feels off. Trying the difference yourself is as simple as asking the same question to both a search engine and a chatbot and comparing the responses side by side.

An anti-AI art shop written entirely by ChatGPT Someone made a whole online storefront arguing against AI art, but used ChatGPT to write all the text and format it. The result looks exactly like typical AI output, which makes the whole thing oddly funny. It shows how easy it is to spot when AI writes something even when the topic is about avoiding AI. The Reddit post includes screenshots of the markdown formatting and emoji bullet points that gave the page away immediately. Reading the actual storefront text reveals how the AI's usual style can clash with the message it's delivering, creating an unintentional loop. Checking out the linked Reddit thread lets you see the full example and decide for yourself how obvious the AI writing feels. Source: npr.org Source: reddit.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

Forwarded this email? Subscribe here — it's free.