The Big Story

Imagine you're stuck on a tough choice like figuring out how to save for a car or handle a confusing school form. New research shared by Ethan Mollick shows that large language models — the AI systems behind tools like ChatGPT and Claude — are already giving financial advice that performs better than many people expect. Mollick spent years working with organizations that teach financial literacy because so many people unknowingly end up in bad situations without clear guidance. The fact that these models now deliver advice rated higher than expected changes that picture.

Think of these models like a super-patient friend who has read millions of money guides and can explain options in plain language without judging you. Mollick points out that the same help now sits in your pocket, ready for questions about budgets, saving, or avoiding common traps. This matters because real life is full of complicated stuff — healthcare paperwork, government forms, personal budgets — that nobody teaches you in school. AI can act like a helpful translator that turns confusing rules into steps you can actually follow. For students or anyone starting out, it means you don't have to muddle through alone anymore.

The research suggests these models are especially useful for the exact situations where people usually feel lost. Consumer AI is underrated precisely because it tackles tasks that feel overwhelming by complexity or lack of time. Life decisions around money, school forms, or healthcare often lack easy expert access, and these tools fill that gap without requiring appointments or fees. The polarization around AI sometimes hides how capable the models already are at solving hard, real problems. Groups focused on helping people may miss this if they reject the tools outright.

The "so what" for listeners is personal: instead of guessing your way through forms or savings plans, you can ask for clear breakdowns tailored to your situation. You can try it yourself today: open ChatGPT or Claude on your phone, type something like “Explain how compound interest works and give me three simple ways a teenager could start saving,” and see what it suggests. Just remember to double-check important money advice with a trusted adult. Source: x.com

Explain Like I'm 14

You know how when you're baking cookies, a recipe card tells you exactly which steps to follow and in what order so you don't mix everything at once? AI agents work a bit like that. Researchers from Princeton University and UC San Diego found that giving an agent a “skill” — basically a clear set of instructions for one job — helps it succeed mostly because it now has a structured path instead of guessing what to do next. The skill doesn't magically add new knowledge; it just organizes the work the way a good recipe keeps you from burning the cookies.

When the list of skills gets too long, though, the agent starts struggling to pick the right one, just like trying to cook with a hundred different recipe cards scattered on the counter. The study shows this structured approach beats just making the model bigger or smarter on its own. Skills improve performance through better workflows rather than extra facts stored inside the model. As the library of skills grows larger, the agent has a harder time locating the exact instructions it needs for the current task. The biggest limit appears when the agent must decide which skill applies, not when it tries to use the skill itself.

That's why the best agents right now combine a few well-written skills with a simple way to choose between them. In short, the real win comes from giving the agent an organized playbook instead of hoping raw intelligence will figure everything out. Source: the-decoder.com

Cool Stuff & Try This

AI that beats debate champions at persuasion New tests show AI chatbots can persuade people more effectively than world-class human debaters. The models learn to listen to what the other person cares about and adjust their points in real time. This is exciting because it shows how good today's AI has become at understanding people, not just facts. The tests measured how well the chatbots could change opinions on topics where humans normally struggle. Anyone can try the same kind of back-and-forth: go to ChatGPT or Claude and start a friendly debate on a topic you care about, like “Should schools start later?” See how the AI responds to your arguments and whether it changes your mind on anything. Source: dongascience.com

Students testing AI on a full semester of work One writer fed an entire semester's worth of notes and assignments to ChatGPT, Claude, and NotebookLM to see which tool actually helped with studying. NotebookLM stood out because it was built to handle long documents and create study guides from your own material. The tool turns your uploaded files into summaries, quizzes, and audio overviews without needing extra prompts. If you're a student, you can try this right now: go to notebooklm.google.com, upload your class notes or a reading, and ask it to make a quick study quiz or summary. It's free and works directly in your browser. Source: xda-developers.com

Quick Bits

One-third of the web is now AI-written A new Pew study found that roughly a third of online content created since ChatGPT launched was written by AI. That changes how we read news, blogs, and social media every day. Source: decrypt.co

Workers quietly asking ChatGPT instead of their boss More people are turning to AI for work advice on everything from handling tricky emails to planning their next career move. It shows how these tools are becoming everyday helpers for real-life decisions. Source: thehrdigest.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

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