In partnership with
What if AI took over your job hunt tonight?
Meet AIApply.
Your AI Career Agent for modern job searching.
AIApply automatically:
Finds the best opportunities
Matches jobs to your profile
Customizes every application
Applies on your behalf 24/7
Saves hours every single week
Stop manually filling out forms.
Stop rewriting the same application over and over.
Whether you’re actively job hunting or casually exploring opportunities, AIApply keeps your applications moving around the clock.
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
OpenAI Built a New ChatGPT. Washington Decides Who Gets It First
TLDR: OpenAI's new GPT-5.6 model won't go out to everyone at once, because the US government asked the company to release it slowly and approve each customer one at a time.
The Story:
OpenAI plans to release its newest model, GPT-5.6, in stages instead of opening it to the public all at once. The Information first reported that the Trump administration asked the company to limit who gets early access. Federal officials would approve preview access one customer at a time during the first few weeks. CEO Sam Altman told staff in a memo that the government would be "approving access customer by customer during this preview period," with a wider release hopefully a couple of weeks later. Two White House offices were involved: the Office of the National Cyber Director and the Office of Science and Technology Policy. Altman made clear this isn't how OpenAI wants to operate long term.
Its Significance:
This is the first time the US government has stepped in to slow down an AI release before it goes public. Why does it matter to you? It changes who gets the most capable AI tools first, and when regular people and small businesses can use them. The reason is security: officials worry GPT-5.6 is good enough at finding software weak spots that it could help hackers if it spread too fast. It also follows a similar move against Anthropic, whose top models were pulled offline weeks earlier over the same kind of worry.
QUICK TAKES
The story: Anthropic told Congress that operators tied to Alibaba ran 28.8 million chats with Claude using about 25,000 fake accounts to copy its skills. The company calls it the largest known case of "distillation," where one company uses a strong AI's answers to train a cheaper copycat.
Your takeaway: Anthropic wants lawmakers to treat this like a national security problem, not just stolen property. If they agree, it could reshape the rules for how AI companies share data and who can reach US models.
The story: A University of Washington team built PaperTok, which uses Google Gemini to turn a research paper into a short, TikTok-style video. A scientist uploads a paper, picks an opening hook, then edits the script and clips. Right now it only works for people who pay for Google Gemini.
Your takeaway: Most people never read research papers, so this could help real science reach more eyes. The downside: some testers said the videos looked "too AI-ish," with nonsense text that could make their work look less trustworthy.
The story: A new study found that a single planted web page can fool an AI shopping helper into recommending a fake product to a real customer. Researchers Minghao Luo and Liang Chen tested how easily these search-powered AI tools get tricked.
Your takeaway: AI shopping assistants are popping up everywhere, but they can be gamed by anyone who knows how to plant a bad page. Until that's fixed, double-check what an AI recommends before you buy.
TOOLS ON OUR RADAR
🎬 Clera Freemium: Turns saved Instagram Reels into structured, searchable notes, so recipes, tutorials, and travel tips stop getting lost in your saves.
⚙️ Innflow Paid: Connect your apps and spin up no-code AI agents that handle repetitive work like emails, summaries, and scheduling across 1,000+ integrations.
📝 NoteRich Freemium: A local-first note app that keeps your notes on your device while still giving you AI summaries and Q&A across everything you have written.
👁️ Verid Paid: Watches any webpage or API and pings you the moment something changes, using AI to pull out only the fields you actually care about.
TRENDING
A Startup Just Raised $320 Million to Train Robots With Video Games - General Intuition, now valued at $2.3 billion, uses millions of hours of recorded gameplay (including exact button presses) to teach one AI model that can both play a Fortnite-style game and steer a real robot. It took just 8 minutes of real-world data to get the same "brain" running a four-legged robot.
Sony Stops Selling Its Robot Dog 'aibo' in Japan - Sony is ending Japan sales of its $3,000 robot puppy once stock runs out, but it'll keep support and cloud services running and continue US sales. Fans were upset; the robot dog builds its own personality and has sold more than 150,000 units since 1999.
Australian Police Start Scanning Faces in Public in Real Time - Western Australia police put live facial recognition in marked vans around Perth, matching passersby against a watchlist of about 4,000 wanted people and missing persons. Even at 90% accuracy, 1 in 10 alerts is wrong, and Australia has no real law built for this yet.
New AI Lets Scientists 'Talk' to Lab Robots and Speed Up Experiments - A team at Pacific Northwest National Laboratory built AutoLabs, an AI that turns a scientist's goal into step-by-step instructions for a lab robot. Setting up an experiment used to take weeks of back-and-forth; now the robot can mix, heat, stir, and filter samples with little human help.
One New Chip Can Do Two Jobs That Generative AI Needs - Seoul National University researchers built a memory chip that handles both the random guessing and the steady math that generative AI relies on, all in one spot. That could make future AI gadgets smaller and use far less power.
Draw a Shape, and This AI Folds DNA to Match It - Researchers in South Korea made a generative AI called "Generative SNUPI" that designs tiny DNA structures to match any shape a user draws. It could help build molecular robots, biosensors, and better drug delivery systems.
California Launches the First Public Tool to Track AI Job Losses - California rolled out an AI-Unemployment Tracker that watches jobless claims in roles most exposed to AI. So far there's no statewide wave of AI layoffs, though college-educated workers in high-exposure jobs near San Francisco are showing early signs.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
⏳Write a note to your future self, seal it, and it stays locked with a live countdown until the day you chose to open it.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Time Capsule — write a note that stays sealed and locked until a future date, with a live countdown. No API; fully client-side. Persist to localStorage key 'time_capsule_v1'.
Aesthetic: midnight vault (#0c0e16), wax-red (#aa3c37) and brass-gold (#c9a86a) accents, faint red glow top + gold glow bottom. Cormorant Garamond serif for headings/labels/messages (the keepsake feel), Inter sans for help text, JetBrains Mono for countdown digits and small labels. A small SVG wax-seal emblem per capsule (red hourglass while sealed, gold ✦ when ready).
Compose form: label (visible on the outside while sealed), message textarea (the hidden content), optional visible hint, and an open date via preset chips (1mo/6mo/1yr/5yr) plus a custom date input. Validate: label, message, future date required.
Mechanics (all client-side):
- Seal: store {id,label,hint,body,createdAt,openAt,opened,openedAt}. Lightly obfuscate the body with base64 so a casual localStorage peek doesn't spoil it.
- Vault list sorted: sealed/ready by soonest open, opened last. Counts of sealed vs opened.
- Sealed capsule: shows label, sealed+open dates, hint, and a live 4-unit countdown (days/hrs/min/sec) updating every second. Message NOT rendered.
- When now ≥ openAt: capsule glows "ready," shows hint and a pulsing "Break the Seal" button. No early opening.
- Opened: unfold animation revealing the message (white-space preserved) + "written … for …" line.
- Delete button per capsule with a confirm (warn that an unopened message is lost unread).
- setInterval ticks countdowns each second and re-renders when one becomes ready.
Render: centered masthead, compose card, then "The Vault" with status subline and capsule cards. Empty state: "Your sealed notes will rest here, waiting."What this does: Write a message, give it a label you'll see on the outside, optionally leave a visible hint, and pick an open date (preset chips for 1 month, 6 months, 1 year, 5 years, or a custom date). Sealing it drops the capsule into your Vault, where it sits locked with a live countdown ticking down days, hours, minutes, and seconds. The message itself stays hidden, even from you, lightly obfuscated in storage so a casual peek won't spoil it. When the date arrives the capsule glows "Ready to open," and breaking the seal unfolds the message with the date you wrote it. Fully client-side, no API. Everything persists in localStorage, so capsules survive across sessions.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Fold a strand of DNA into almost any shape a person draws, then plan how the pieces bond together.
❌ Still Can't: Reliably tell a real product page from a planted fake one, so a single bad page can fool an AI shopping bot.
✅ AI Can Now: Run a full lab experiment, mixing, heating, stirring, and filtering samples, from a plain-language goal.
❌ Still Can't: Make a research-paper video without leaving behind nonsense text that gives it away.
FROM THE WEB: RESIDENCY OPPORTUNITY
The unlimited access alone makes this worth applying for if you have an interest in digital creation.
RECOMMENDED LISTENING/READING/WATCHING
A Memory Called Empire by Arkady Martine - Book
An ambassador from a small mining station arrives at the capital of a sprawling empire that has just absorbed her predecessor's body, but she's been given a malfunctioning version of her culture's "imago" tech, which is supposed to let her carry her predecessor's memories and skills directly inside her own head. Martine, a Byzantine historian, won the 2020 Hugo for this and the novel is one of the most accessible recent books about what consciousness transfer and digital identity actually mean for human relationships and politics.
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.
What if AI found your next job overnight?
Stop wasting hours filling out repetitive applications. AIApply automatically discovers relevant jobs, optimizes your resume, generates personalized applications, and applies for you around the clock so you can focus on securing your next opportunity faster.