Sponsored by
Find Your Best-Fit 3PL.
Skip weeks of sales demos. Tell Fulfill your products, order volume, and shipping zones, and our team screens a network of 2,800+ vetted 3PLs to hand you a ranked shortlist of your best-fit partners. You reach out to the matches you like, get pricing directly from them, and decide on your own terms, whether that means switching to a better fit or negotiating with your current 3PL. There is no fee and no pressure, and most brands get their shortlist within days. See your best-fit matches.
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
The US Banned Anthropic's Fable 5 for 19 Days. It's Back Today
TLDR: The US government's ban on Anthropic's most powerful AI model, Fable 5, is over, and the model is back for everyone today, July 1, right after tests showed cheaper models could find the same security flaws it did.
The Story:
On June 12, the US government put export controls (rules that limit who can use a technology) on Anthropic's two newest models, Fable 5 and Mythos 5. The rule blocked any foreign national from using them, even Anthropic staff who aren't US citizens. Anthropic couldn't check everyone's nationality in real time, so it shut both models off for all users. The trigger was a report from Amazon researchers who got Fable 5 to point out software flaws, and in one case write code showing how one flaw could be used. But when Anthropic ran that same task on weaker, cheaper models like Claude Opus 4.8, GPT-5.5, and Kimi K2.7, they could spot the same flaws. On June 30 the government lifted the controls, and Fable 5 is back today, July 1 across Claude.ai, Claude Code, and Claude Cowork, now with a stronger safety filter that blocks the reported trick in over 99% of cases and sends those requests to Opus 4.8 instead.
Its Significance:
This is the first time the US government has stepped in to pull a company's AI model off the market. For regular users and developers, the model is back, but the new filter is stricter, so some normal coding and debugging jobs may get blocked by mistake and bounce to Opus 4.8. Anthropic and rivals like Amazon, Microsoft, and Google are now writing a shared rulebook for how bad an AI "jailbreak" (a trick that gets past a model's safety rules) really is, which could shape every model release from here on.
QUICK TAKES
The story: Netflix's new reality competition "Wonka's The Golden Ticket" uses an AI recreation of Gene Wilder's voice, built with ElevenLabs and approved by Wilder's estate. Wilder played Willy Wonka in the 1971 film and died in 2016.
Your takeaway: This is one more famous voice getting an AI second life, joining Michael Caine and Judy Garland. Even with the family's blessing, plenty of fans online called it ghoulish, which shows how raw the "should we even do this?" question still is.
The story: Anthropic launched Claude Sonnet 5, its "most agentic Sonnet yet," and it nearly ties the pricier Opus 4.8 on real-world work tests while costing far less, at $2 per million input tokens through August 31 versus $5 for Opus. It's now the default model for Free and Pro users.
Your takeaway: If you use Claude, you're probably already on a smarter model. The catch is a new tokenizer that can eat up to 35% more tokens per task, so cheaper per token doesn't always mean cheaper per job.
The story: Google is bringing its Gemini Spark agent to the Mac, where it can sort files, build a spreadsheet from your invoices, and soon run tasks on your computer while you're away. Spark now also links to Canva, Dropbox, Instacart, OpenTable, Zillow Rentals, plus Google Tasks and Keep.
Your takeaway: This is Google pushing Gemini from a chatbot that answers questions to an agent that does chores on your real desktop. For now the Mac app is beta and only for Google AI Ultra subscribers in the US.
TOOLS ON OUR RADAR
📝 Drafts Freemium: A brilliantly fast text editor that opens instantly to a blank page allowing you to capture quick thoughts before deciding where they should be organized.
🔒 AppFlowy Freemium: A highly secure digital workspace that gives you total control over your data while providing a beautiful interface to manage your projects and notes locally. (Alternative to Notion)
📹 Fathom Freemium: A powerful video conferencing assistant that joins your online meetings to automatically record transcribe and highlight the most important moments of your conversations.
✂️ Munch Freemium: An intelligent video repurposing engine that scans your long recordings and automatically extracts the most engaging short clips perfectly optimized for modern social media algorithms.
TRENDING
X Now Offers an MCP Server to Make Its Platform Easier for AI Tools to Use - X launched a hosted server that lets AI assistants like Claude, Cursor, and Grok tap into its data using your own account, no custom setup needed. It's read-only, so tools can search and read posts but can't auto-post spam.
This Smart Ring Hides a Trackpad and Lets You Whisper-Control Your Computer - The OASIS 1 ring has a tiny mic for quiet dictation and a trackpad to fix mistakes, for $289 with shipping around Christmas. If privacy is your worry, Pebble's open-source Index 01 ring only records when you press its button and does the work on your phone, no cloud needed.
Google Opens Nano Banana 2 Lite and Gemini Omni Flash to Developers - Google's fastest image model can make a picture in about 4 seconds, and its Omni Flash model turns those pictures into short videos you edit just by talking to it. Both carry SynthID watermarks so people can check if content was AI-made.
OpenAI's GeneBench-Pro Tests Whether AI Can Do Real Biology Research - The new test hands AI 129 messy biology problems where it has to pick the right way to analyze data, not just recall facts. OpenAI's best model solved 31.5% of them, and reviewers said each problem would take a human expert 20 to 40 hours.
Japan Wants 10 Million More Robots by 2040, Some for Medical Care - Japan set a national goal of 10 million AI-powered robots across 18 fields by 2040, backed by up to $6.1 billion, partly to fill a shortage of about 570,000 care workers. A new group called Noetra, owned by SoftBank, NEC, Sony, and Honda, will build the AI model behind them.
US Air Force Picks General Atomics and Anduril to Build AI Fighter Drones - The Air Force will build over 150 AI-piloted "loyal wingman" drones to fly beside F-35 and F-22 jets, with software that can pass control between different AI "pilots" mid-flight. The long-term plan calls for 1,000 or more.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🎁 Describe someone like you're talking to a friend. Get genuinely thoughtful gift ideas matched to who they actually are, not generic junk.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Gift Whisperer — describe a person, get thoughtful gift ideas matched to who they really are. Persist to localStorage key 'gift_whisperer_v1'.
Aesthetic: warm plum-black (#160f14), rose-pink (#e6789a) primary with rose glow top-left and gold glow bottom-right. Fraunces serif (italic for prompts, the read, gift names) for headings, Inter sans for body, JetBrains Mono for labels. Gold (#e6b478) for prices and the wildcard card, red (#d26e6e) for the avoid card, green tag for "safe."
Form: big person-description textarea (encourage little details/quirks/inside jokes), occasion dropdown, budget dropdown (under $25 → sky's the limit), relationship dropdown (very close → coworker).
System instructions: thoughtful gift expert who reads people; suggest specific findable gift types (categories/product types, NOT made-up brands, no generic "candle" or "gift card"); use the little details to guide picks; respect budget and relationship boundary (coworker ≠ partner); range from safe to one stretch; every idea tied to a detail they mentioned. Return raw JSON: read (2-3 warm perceptive sentences reflecting who they are), gifts (4 items: name + why with **bold** on the connecting detail + price in budget + tag safe|thoughtful|wow), wildcard (name + why, one unexpected idea), avoid (a category to skip + why), card_note (short warm non-cheesy line for the card matched to relationship/occasion).
Render: gradient "here's who I think they are" card (italic serif). Gift cards each with name + gold price pill + why (bold detail) + a colored tag chip (safe green / thoughtful pink / wow gold). Gold-bordered wildcard card. Two-column split: red "don't get them" and pink "for the card" (quoted). Copy ideas + archive of people shopped for (snippet + occasion).What this does: Describe the person, the little details and quirks and things they've mentioned wanting, then pick the occasion, budget, and your relationship. It reflects back who it thinks they are to show it understood, then gives four gift ideas ranging from safe crowd-pleaser to wow, each with a price in your range and a reason tied to a specific detail you mentioned (tagged safe / thoughtful / wow). You also get a creative wildcard they'd never think to ask for, a category to avoid for this person, and a short non-cheesy line for the card matched to your relationship. It respects the relationship boundary, so a coworker gift stays different from a partner gift. Saves lists to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Turn a single photo into a short, edited video you steer just by talking to it (Gemini Omni Flash).
❌ Still Can't: Reliably tell a harmful cybersecurity request from a normal coding question, which is why Fable 5's stricter filter blocks some safe requests by mistake.
✅ AI Can Now: Recreate a late actor's voice closely enough to narrate a whole TV show (Gene Wilder for Netflix).
❌ Still Can't: Handle most real biology research judgment calls, with the top model solving just 31.5% of GeneBench-Pro problems.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Metropia (2009) - Movie
A Swedish adult animated dystopia set in a near-future Europe where all the underground trains have been connected into one gigantic transit system, and a shy call-centre worker starts hearing an unfamiliar voice in his head. Directed by Tarik Saleh and rendered in a distinctive photomontage style influenced by Terry Gilliam, with Vincent Gallo, Juliette Lewis, and Stellan and Alexander Skarsgård voicing an eerie, dreamlike political thriller about surveillance and thought-shaping tech.
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.
Moda is the AI design agent with taste
Moda is an AI design product where you prompt what you need, get a complete on-brand design, and edit every element on a full canvas.
Our viral launch hit 4.4M views in days, tens of thousands signed up, and executives at major finance and tech companies now use it.