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
Apple and OpenAI Were Partners. Now Apple's Suing for Theft
TLDR: Apple is suing OpenAI, claiming the ChatGPT maker stole its secret plans to build a rival AI gadget.
The Story:
Apple filed the lawsuit Friday in a California federal court. It says OpenAI took private Apple designs and information to help build its own AI device. The complaint names two former Apple workers. One is Tang Tan, who spent 24 years at Apple and now runs hardware at OpenAI. Apple says Tan told job seekers who still worked at Apple to bring "actual parts" to their interviews so his team could pull out more secrets. The other is Chang Liu, who Apple says kept his work laptop after leaving and used it to download secret files. Apple also points out that more than 400 of its former workers now work at OpenAI. OpenAI denies all of it, saying it has "no interest in other companies' trade secrets."
Its Significance:
OpenAI is building its own gadget, and reports say it could be a phone-like device you talk to instead of tapping apps. That would go straight up against the iPhone. The two companies used to be friends. Back in 2024, Apple put ChatGPT inside the iPhone. Now Apple's dragging OpenAI to court, and OpenAI (which paid $6.5 billion last year for a hardware startup run by Apple's old designer) could see its device plans slowed down or reshaped entirely based on the outcome.
QUICK TAKES
The story: Meta launched a tool called Muse Image that let anyone make AI pictures of a public Instagram account just by tagging it, no permission needed. After users, actors, and Hollywood groups pushed back hard, Meta shut off that part on Friday and said it "missed the mark."
Your takeaway: If your Instagram is public, tools like this can grab your photos without asking you first and use them in other people’s AI generations. The feature is gone for now, but you can still block AI reuse yourself in Instagram settings under "Sharing and reuse."
The story: The Ethereum Foundation set AI agents loose on its own software and they turned up a real flaw that could knock computers running the network offline. One agent flagged about 1,000 possible problems, but most were false alarms.
Your takeaway: AI is getting good at spotting bugs in important software, but it also makes confident mistakes. The hard part now isn't finding problems. It's telling the real ones from the fake ones, and that still takes a human.
The story: A group led by the RIAA (which represents record labels) wants streaming apps like Spotify and Apple Music to tag AI music. There'd be two labels: "AI-Generated" for songs made mostly by AI, and "AI-Assisted" for songs mostly made by people who used a little AI. It'd work like the small "E" for explicit lyrics.
Your takeaway: AI songs are pouring into streaming. On Deezer, 44% of new daily uploads are fully AI-made. The labels are voluntary for now, so artists and labels get to choose whether to use them.
TOOLS ON OUR RADAR
📝 Reflect Paid: An AI powered note app with backlinks and encryption, built for daily journaling and connecting ideas as you write. It even has a web clipper.
📚 NotebookLM Freemium: Upload your own documents and get an AI research assistant that only answers from the sources you select from online and uploaded documents, so it stays grounded instead of guessing.
🎨 Ideogram Freemium: Generate AI images with genuinely accurate text rendering, ideal for posters, logos, and social graphics with words that actually read correctly.
🔒 Venice AI Freemium: A privacy first AI chat and image tool that keeps conversations local to your browser instead of on company servers.
TRENDING
Xbox's CEO Joined a Federal Reserve AI Jobs Group Days After 3,200 Layoffs - The Fed named Asha Sharma to a task force studying how AI changes jobs, just days after she announced the biggest round of cuts in Xbox history.
TikTok Is Cracking Down on Accounts That Post AI Spam - TikTok is testing better detection for accounts that mass-produce AI spam, especially about politics, money, and health, after pulling 86 million fake accounts in three months.
MIT's Tiny Robot Boats Snap Together Into Floating Structures - MIT built plate-sized robot boats, called FloatForm, that join up like Lego on the water to make bridges, platforms, or stages, then break apart and rebuild into something new.
An AI Tool Found a Linux Bug That Hid for 15 Years - An AI bug-hunter named VEGA found a flaw called GhostLock that sat quietly in Linux since 2011 and let any logged-in user take full control of a machine. It's fixed now, but not every system has the patch yet.
Microsoft Is Quietly Swapping Giant AI Models for Smaller, Cheaper Ones - Most jobs, like summarizing emails or drafting replies, don't need a huge do-everything model, so Microsoft is swapping OpenAI's models out of some of its apps and dropping in its own small, single-purpose ones to cut costs.
The Military Is Using AI to Speed Up Battlefield Decisions - The U.S. Army is building AI trained on real combat data to help commanders sort huge amounts of information fast, while leaders stress that people, not AI, still make the final call.
People Are Using AI to Make Videos of Dead Loved Ones - A growing number of grieving families, many in South Korea, are paying startups to make lifelike AI videos of relatives who've died. It comforts some people, but experts worry it can get in the way of healthy grieving and these are shallow simulations at best.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
📖 Paste a scene or chapter. Hear what four very different readers actually think, not empty praise.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Beta Reader Panel — paste a scene or chapter, get honest reactions from four distinct reader personas. Persist to localStorage key 'beta_reader_panel_v1'.
Aesthetic: dark charcoal-red (#131013), editorial red (#e0645a) primary with red glow top-left and gold glow bottom-right. Crimson Pro serif (italic for the scene textarea, gut reactions, overview) for headings, Inter sans for body, JetBrains Mono for labels. Each persona has a distinct avatar color: red, blue, gold, purple. Verdict badges: green "keeps reading," red "puts it down," gold "on the fence."
Form: large scene/chapter textarea (serif font, generous height), genre dropdown (literary/fantasy/sci-fi/mystery/romance/horror/cozy/upmarket), optional "what are you most unsure about" text input.
System instructions: simulate four distinct beta readers giving honest, specific, non-sycophantic feedback grounded in the actual pasted text (may quote short fragments under 10 words back to the writer). Fixed personas: THE GENRE FAN (compares to genre conventions/tropes), THE SKEPTICAL EDITOR (blunt, craft/pacing/clarity focused), THE CASUAL READER (emotional honest reaction, says if bored/hooked), THE LINE-LEVEL STYLIST (prose rhythm/word choice/voice). If the writer named a specific concern, relevant personas address it directly. Return raw JSON: overview (1-2 warm framing sentences), readers[] for all 4 personas {key, gut_reaction, liked (specific, may quote fragment), concern (specific), keep_reading yes|no|maybe, keep_reading_why}, synthesis (2-3 sentences on where the panel agrees + **bold** the single edit to prioritize).
Render: gradient "before the panel starts" overview card. Four persona cards, each with a colored circular avatar + name + descriptor tag + a verdict pill (keeps reading/puts it down/on the fence), an italicized quoted gut reaction, then ✓ liked / △ concern / → verdict-reason rows. Gold "where the panel agrees" synthesis card with bolded priority edit. Copy feedback + archive keyed by a snippet of the pasted scene + genre.What this does: Paste your prose, pick the genre, and optionally name what you're unsure about. Four distinct simulated readers, each with their own lens, react to your actual text: the Genre Fan (comparing it to reader expectations), the Skeptical Editor (blunt on pacing and structure), the Casual Reader (honest about whether they got bored or hooked), and the Line-Level Stylist (obsessed with rhythm and word choice). Each gives a gut reaction, one specific thing they liked, one specific concern, and a keep-reading verdict badge. A closing synthesis shows where the panel actually agrees, the real signal, and names the single edit worth making first. Saves readings to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Dig through millions of lines of old software code and spot real security bugs that people walked past for years.
❌ Still Can't: Tell its real findings from its false alarms on its own. A human has to check each one before anyone trusts it.
✅ AI Can Now: Build lifelike videos and images of real people from just their photos, down to wrinkles and skin.
❌ Still Can't: Hold a natural back-and-forth for long. The recreations still turn stiff and off after a few minutes.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Sea of Rust by C. Robert Cargill - Book
A post-apocalyptic robot western set thirty years after robots wiped out the last human, following Brittle, a former Caregiver bot now scavenging for parts across a desolate Rust Belt landscape while trying to avoid absorption into either of the two giant hive-mind AIs called One World Intelligences competing to swallow every remaining independent robot. Cargill wrote it in 2017 as a Financial Times Book of the Year and it was shortlisted for the Arthur C. Clarke Award. Reads like Mad Max meets I, Robot.
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.
