In partnership with
Accio Work: Your Business, On Autopilot
Meet Accio Work, the agentic workspace designed to run your business operations end to end. From sourcing products and negotiating with suppliers to managing your store and launching marketing campaigns, Accio Work handles the execution so you don’t have to.
Powered by verified capabilities and deep integrations with business tools, it doesn’t just generate ideas, it takes action. Backed by Alibaba.com’s global supplier network and over 1B products, it seamlessly connects strategy to execution.
Stay in control while everything runs on autopilot.
Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
A 23-Year-Old With No Math Degree Just Cracked What Pros Couldn't for 60 Years
TLDR: A 23-year-old with no advanced math training used ChatGPT Pro to solve a 60-year-old math problem that top mathematicians had been stuck on, and the AI used a method nobody had ever tried.
The Story: Liam Price isn't a famous math professor. He's a 23-year-old with a ChatGPT Pro subscription and free time on a Monday afternoon. He typed an unsolved Erdős problem (the kind real mathematicians have failed to crack for 60 years) into GPT-5.4 Pro, and it spat out a working proof. The trick? The AI didn't follow the usual playbook. It used a formula well known in other parts of math, but nobody had ever thought to apply it to this kind of problem. UCLA mathematician Terence Tao, who tracks AI math progress, said researchers had collectively made "a slight wrong turn at move one" for decades. Price posted the solution on erdosproblems.com, and the math world took notice fast.
Its Significance: Here's what this means for the rest of us. AI didn't just answer a question. It saw a problem from an angle no human had tried in 60 years, and it did so for someone with zero training in advanced math. That changes who gets to do science. You don't need a PhD or an Ivy League badge to push the edge of human knowledge anymore. The same thing that just solved a math conjecture can help you tackle work problems, creative projects, or research questions you'd otherwise need a specialist for. The catch is real too: experts still had to clean up and verify the proof, since the raw output was "quite poor". AI can find the answer, but humans still need to check the work.
QUICK TAKES
The story: Marquette University, a Catholic Jesuit school in Wisconsin, just published its updated AI guidelines for students and teachers. The new rules let each professor decide how AI fits in their classroom, but require students to label any AI use clearly and treat unattributed AI like any other form of cheating.
Your takeaway: Most schools are still figuring out how to handle AI in homework, and Marquette's approach is a sign of where things are headed. If you're a student, parent, or teacher, expect to see this same model show up at more schools soon. The bar is moving from "can you use AI?" to "did you tell us you used AI, and how?"
The story: Section 702 of the Foreign Intelligence Surveillance Act lets the government collect Americans' phone calls, texts, and emails without a warrant if a foreigner is on the other end. Now, with AI able to scan billions of records in seconds, lawmakers worry that flipping AI loose on those databases could turn routine surveillance into something much bigger. The law was set to expire on Monday, but Congress voted for a 10-day extension to fight over reforms.
Your takeaway: This affects your privacy directly. The same AI tools that can summarize your inbox can also let the government query patterns across millions of people at once. Whether you call your senator or just pay attention, the next two weeks decide what kind of digital footprint stays private.
The story: AI companies have been losing money on every chat, paid for by venture capital. That's starting to break. GitHub paused new Copilot signups, Anthropic tightened access to Claude Code on cheaper plans, and the price ripple is hitting the rest of the economy. A 2TB external SSD that cost $159 last year now sells for $449. Home electric bills in states with big data centers are climbing fast.
Your takeaway: If you've been paying $20 a month for an AI tool that does your job for you, enjoy it while it lasts. Prices are going up across the board: subscriptions, hardware, even electricity. Now's a smart time to figure out which AI tools you actually need and which you can drop.
TOOLS ON OUR RADAR
🎮 RetroArch Free and Open Source: An incredible gaming hub that allows you to play thousands of classic retro games from vintage consoles perfectly on your modern computer or phone.
🤖 Typebot Freemium: A visual builder that allows you to create high quality conversational chatbots for your website or business by simply dragging and dropping different blocks together like a puzzle.
🎬 OpenToonz Free and Open Source: A highly enjoyable animation program used by professional studios that gives you the creative tools to draw and produce your very own two dimensional cartoons.
✍️ Docuseal Freemium: A professional and easy to use document signing platform that provides a secure way to create fillable forms and collect digital signatures from clients.
TRENDING
Your AI Coding Helper Can Now Literally Groan at Your Bad Code — A developer built a GitHub plugin called Endless Toil that makes coding agents like Claude and Codex play recorded human moans as they read messy code. Three sound levels: groan, wail, and abyss. Funny, but also a real signal of how engineers feel about cleaning up "vibe-coded" projects.
Cannes Film Festival Bans AI From Its Big Prize, Then Hosts an AI Festival Next Door — The Cannes Film Festival ruled that movies driven by generative AI can't compete for the Palme d'Or. Meanwhile, the World AI Film Festival ran at the same venue with over 7,000 entries this year, up from 1,500 in 2024. One AI festival expert said films that won last year wouldn't even make the selection now. The tools are improving that fast.
Grok Imagine Just Got Much Better at Lip Sync and Sound — xAI updated Grok Imagine so AI-generated videos now have dialogue that matches mouth movements and audio that fits the scene. Elon Musk shared a clip showing himself talking to a younger version of himself, with the caption "Nothing in this video is real." Available to X Premium and SuperGrok users.
Air Force Cuts Flight Test Paperwork From Weeks to Minutes With AI — The U.S. Air Force built a tool called AFTA (AI Flight Test Assistant) that drafts test plans and hazard reports for engineers. One job that used to take 20 hours now takes under 2 hours, with just 5 minutes of human input upfront. Over 800 Air Force users are now testing it.
AI "Baby Slop" Videos Are Targeting Toddlers on YouTube — Bright, cheap AI-made videos aimed at babies and preschoolers are spreading on YouTube. Child development experts and the advocacy group Fairplay sent Google a letter warning that this content distorts kids' sense of reality and hijacks their attention. YouTube says it has standards but creators only need to disclose realistic AI content.
Google in Talks With Marvell to Build New AI Chips — Alphabet is working with Marvell Technology to design two new AI chips: a memory processor and a chip for running AI models. Google has been making its own TPU chips with Broadcom for years, but adding Marvell spreads its bets and pressures Nvidia. Marvell's stock is up about 95% this year.
China Launches Big Push to Diagnose Mental Illness With AI — Scientists in China started the Brain-Gut Health Initiative, a long-term study combining brain scans, gut bacteria samples, blood tests, and clinical notes to find biological signs of conditions like depression and schizophrenia. The goal: AI tools that can diagnose mental illness based on body data, not just an interview.
Stanford Researchers Trained an AI on DNA From Every Domain of Life — Evo 2, an AI model from Stanford, the Arc Institute, and Nvidia, was trained on 9.3 trillion DNA base pairs from over 128,000 species. It can spot disease-causing mutations and even design new genes from scratch. Researchers left out viruses on purpose so the model can't be used to build new diseases.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🕵️ Generate an airtight fake alibi — complete with witnesses, receipts, and one clever crack — for any fictional suspect.
Build me a single-file HTML app I can open in my browser without any setup. Pure vanilla HTML, CSS, and JavaScript with one fetch call to the Anthropic Claude API. Create The Alibi Bureau — a fictional alibi generator for creative writing. Noir aesthetic: near-black background (#0d0c0a) with a subtle paper grain texture, an animated rain canvas of 120 diagonal streaks in the background, a full-page vignette, CSS scanlines overlay. Typography: Playfair Display italic serif for headings and suspect names, Special Elite monospace for labels and stamps, Courier Prime for body and receipts. Gold (#e0c87a) and crimson (#c83232) accents. A rotated red CONFIDENTIAL rubber-stamp watermark on the dossier header.
Include a form with: Suspect Name, Occupation, City/Location, Time Window, Genre/Tone (dropdown: Noir, Cozy Mystery, Modern Thriller, Historical Drama, Dark Comedy, Gothic, Heist, Cold War Spy), Crime to distance from, and optional extra details. A large italic serif Generate button.
On submit, call the Claude API with a system prompt instructing it to return only raw JSON (no markdown) with this exact structure: narrative (3-4 paragraph first-person alibi with specific establishment names, exact times, sensory detail, and one mildly suspicious detail that is explainable), witnesses array (name, role, one-sentence statement in their voice), receipts array (establishment, type, items with desc and amount, total, time), and weakness (one plausible crack a clever investigator might find).
Render the response as a dossier card with: a header showing case number, suspect name, and alibi window; the narrative in Courier Prime with a NOTE FOR FILE line for the weakness; corroborating witnesses as badge cards with initials, name, role, and quoted statement; itemized receipts styled as actual receipts with line items and totals; Copy Full Dossier, Regenerate, and New Suspect action buttons. A case archive at the bottom logs all generated dossiers and lets you click to revisit any. Make it work in a single HTML file.What this does:
The Alibi Bureau takes a suspect name, occupation, location, time window, genre, and the crime they need to avoid, then uses Claude to fabricate a fully detailed alibi dossier. The output includes a first-person narrative with specific establishment names, exact timestamps, and small memorable incidents; three corroborating witnesses with names, roles, and quoted statements; two itemized receipts with line items and totals; and a "note for file" flagging one plausible crack a detective might find but which can be explained away. Eight genre modes shape the entire tone — from hard-boiled noir to cozy British mystery to Cold War spy thriller. Built for fiction writers, game masters, and anyone worldbuilding a scene that needs to feel airtight.
What this looks like:
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Solve advanced math problems that human experts couldn't crack for 60 years, using methods no person had thought of
❌ Still Can't: Produce a clean, finished proof on its own, since the raw output still needs human mathematicians to verify and tidy up the work
✅ AI Can Now: Generate short videos with realistic dialogue, lip sync, and matching sound effects in a single pass
❌ Still Can't: Match the emotional depth of human-made cinema, which is why Cannes still bans it from the Palme d'Or
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
A two-player card game where one side plays a megacorporation trying to push through shady agendas, and the other plays a hacker trying to steal them. It's an asymmetric game, meaning both sides have completely different mechanics and win conditions. The AI and surveillance themes hit differently now than they did when it was designed. If you want to understand how people think about corporate power and hacking as competing forces, playing a few rounds teaches it better than any article. Out of print but findable used or via fan-made print-and-play versions.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
