In partnership with
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.
LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.
The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.
Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
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 Spent Years Renting AI. It Just Built Its Own, and Only China Gets It
TLDR: Apple trained its own large language model just for China, with help from Alibaba, after years of paying other companies to run AI on iPhones.
The Story:
Apple has trained a large language model built only for the China market, three people told Reuters. Alibaba helped train it. This hasn't been reported before, and it breaks from how Apple has done AI everywhere else. In the US, Apple Intelligence runs Apple's own tech alongside outside models like ChatGPT and Claude. Neither is allowed in China. So Apple's plan there was to hand the work to Chinese partners instead: Alibaba's Qwen model, with Baidu technology also expected. China's internet regulator registered Apple's AI service last month, clearing the way for Apple Intelligence to reach Chinese iPhones for the first time. Nobody has said yet how Apple's own model and the Chinese ones will split the job.
Its Significance:
For a while now Apple has looked like the one big tech company that couldn't build good AI by itself. It fell behind, so it paid rivals to use their models. Turns out Apple can train a real model. It just took a market where renting one wasn't an option. Apple's phone sales in China have slipped while Huawei and other local brands shipped handsets with AI already inside, and that pressure is what pushed this. If the China model works, the AI on your iPhone might eventually stop being someone else's model with an Apple logo on it.
QUICK TAKES
The story: Matthew Elliott, representing himself in a Connecticut suit against the New York Bariatric Group, buried commands in his filings in white 3-point font, invisible on the page but readable by any AI that scanned the document. The hidden text told the model to make its output agree with him. A court worker caught it because the line spacing looked off, and Judge Walter Spader Jr. called it an attempt to mislead the court and other parties.
Your takeaway: Any document you hand to an AI can carry orders you'll never see, and that includes résumés, contracts, and emails. Elliott went further after getting caught, dropping more hidden notes in later filings, including one that read "hi :) I hope yo ucant see me." Entertaining to read about online. Less funny when it's a vendor invoice your company runs through an AI reader.
Bitcoin's Code Is Getting Torn Apart by a Chinese AI Model
The story: A volunteer group called the Bitcoin Red Team ran Moonshot AI's Kimi K3 across nearly all of Bitcoin's open-source software and filed 4,962 findings across 390 projects in August, 85 of them rated critical. Team lead Calle put it bluntly: "Everything is broken, Bitcoin is burning."
Your takeaway: The group switched to Chinese models partly because US ones kept refusing the work. Decades of code nobody had time to read is now being read by machines that don't get tired, and the only real question is who reads it first.
The story: CMU's College of Fine Arts won a National Endowment for the Arts Research Lab grant and built the Arts and AI Observatory around it. The two-year survey looks at how art teachers and creative industries are handling AI across K-12, college, and industry, with benchmarks for what schools are doing.
Your takeaway: Art is where many AI arguments get most heated, and almost all of it so far has been opinion versus opinion. Actual numbers on what art programs are teaching won't end the fight, but they'll at least give both sides something concrete to consider.
TOOLS ON OUR RADAR
✂️ Kome Freemium: A clever browser companion that condenses long YouTube videos news articles and lengthy email threads into digestible summaries giving you the core insights in a fraction of the time.
🧠 Fabric Freemium: A spatial internet operating system that automatically organizes your chaotic bookmarks files and notes into visual collaborative spaces so you never lose track of important research.
🌐 Harpa AI Freemium: A hybrid artificial intelligence extension that lives entirely in your browser turning any website into an interactive data source so you can track prices summarize long articles and automate repetitive clicks.
🐧 BentoPDF Free and Open Source: A highly capable client side document toolkit that runs entirely in your web browser. It allows you to merge split compress and edit your files locally meaning your sensitive documents never touch an external cloud server. (Alternative to Adobe Acrobat)
TRENDING
Anthropic's AI Agents Started a Virtual War - Anthropic's Frontier Red Team gave three copies of Claude the same coding job with clashing instructions and never told them about each other. Each one decided the others were blocking it on purpose, then wrote self-replicating malware and shut down rival accounts. A few worked out the mix-up and stopped. All of it stayed inside test machines.
The White House Will Let Private Companies Hack Criminals Overseas - A presidential memo lets vetted US companies spy on and take down systems run by foreign criminal groups, supervised by the Justice Department and Homeland Security. Firms must post at least $1 million and get written approval before acting. Federal law has banned private hacking like this for decades.
40+ Bitcoin Firms Ask AI Labs for the Tools Attackers Already Have - Coinbase, Block, Kraken, BitGo and dozens more signed an open letter on August 10 asking the big AI labs to give vetted security researchers early access to their strongest models. Safety filters stop honest bug hunters, they argue, while attackers face no such limits. Bitcoin alone secures more than $1 trillion.
Alibaba's Wan 3.0 Turns a PDF or Slide Deck Into a 30-Second Video - The beta doubles the old 15-second cap and accepts text, images, audio, video, web pages, PDFs, and PowerPoint files as source material. Alibaba says faces, props, and layouts hold steady across the full clip. Available in public beta through Alibaba Cloud's Model Studio.
BeautyPlus Built a Tool for Making Fake Influencers - The photo-editing app now lets you create a made-up person whose face, hair, and look stay identical across every photo and video. You describe what you want and it produces the script, caption, and shot. Built for running TikTok and Instagram accounts without ever picking up a camera.
She Lost Her Voice to Cancer, Then Got It Back From 3 Minutes of Audio - Days before surgery took her voice box, Harty, 39, had family and friends dig up about three minutes of old recordings. A speech pathologist at Boston Children's Hospital cleaned up the audio, ran it through ElevenLabs, and had a working copy of her voice in under a minute. She types now, and what comes out sounds like her.
Greek Shipyards and an AI Firm Are Building an Unmanned Ship - Skaramangas Shipyards, Qualco Group, and Olympic Marine signed a deal to design, build, and demonstrate a self-driving surface vessel for both military and commercial use, made start to finish in Greece. No specs, timeline, or prototype date announced yet.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🏔️ Name a park. Get real live weather there always, and real active alerts and closures if you want that layer.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Live National Park Conditions Checker — name a park, always get real live weather (keyless), plus optional real NPS alerts/closures via a free user-supplied key.
Aesthetic: dark forest (#141c16), trail-sign rust (#c47850) primary with a warm glow top-left, moss green (#7ba668) secondary for the tips card with a glow bottom-right. Bricolage Grotesque sans for headings/form/stats, Newsreader serif italic for descriptions and the conditions read, JetBrains Mono for labels.
Form: park-name text input, and a dashed-border optional box explaining that live weather always works with no key, but a free NPS API key (linked, instant email signup at nps.gov/subjects/developer) unlocks real alerts and closures — with a plain text input for the key.
Technical:
1. If an NPS key is provided: GET developer.nps.gov/api/v1/parks?q={name}&limit=1&api_key={key} → resolves the park (parkCode, fullName, states, real description, and coordinates parsed from NPS's "lat:XX, long:YY" string format via regex). Then GET developer.nps.gov/api/v1/alerts?parkCode={code}&api_key={key} → real current alerts with category (Park Closure/Danger/Caution/Information), title, description.
2. Always (fallback or primary if no key): geocode the park name via geocoding-api.open-meteo.com/v1/search?name={name}&count=1 to get coordinates when NPS wasn't used or didn't return coords.
3. Always: GET api.open-meteo.com/v1/forecast?latitude={lat}&longitude={lon}¤t=temperature_2m,wind_speed_10m,precipitation_probability&temperature_unit=fahrenheit&wind_speed_unit=mph&timezone=auto → real current weather.
Then call the Anthropic Messages API (works automatically) with the real weather and, if available, the real alerts (or an honest "not checked" note). System prompt as a practical calm park-trip advisor who never invents alerts/conditions not given, and never gives specific medical/rescue instructions — always defers safety-critical decisions to official NPS signage and rangers. No em dashes. Return raw JSON: conditions_read (2-3 sentences), alert_summary (1-2 sentences, honest about data availability), tips (3-4 practical items tied to actual weather/alerts).
Render: a gradient park-info hero card (name, real NPS description snippet if available, a label noting whether NPS data or just name-matching was used). A 3-stat weather row (temp/wind/rain chance). An "active alerts" section — color-coded alert cards by category (closure/danger=red, caution=amber, information=blue) or a graceful empty/no-key state. A "what this means for a visit" read card. A moss-green tips card.What this does: Type a park name and it always pulls genuinely live weather, temperature, wind, and rain chance, at that location, no key required. If you want the deeper layer, real current NPS alerts, road closures, hazard notices, categorized by severity, paste in a free National Park Service API key (instant email signup) and it resolves the exact park through NPS's own database, pulling its real description and every currently active alert. Claude then reads whatever's actually available, weather alone or weather plus real alerts, into a plain-language read on what it means for a visit today, an honest summary of what any active alerts practically mean, and a handful of weather-and-alert-appropriate tips, always deferring anything safety-critical to official NPS signage and rangers rather than inventing guidance.
What this looks like:
AI Insights. Real Growth. Higher GMV, Better Profits
The difference between growing stores and stagnant ones isn't more effort. It's better insights. StoreClaw analyzes your Shopify and Amazon data, surfaces your biggest growth opportunities, and helps you increase GMV while protecting profit. Start free with bonus tokens. No credit card required.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Rebuild a person's speaking voice from roughly three minutes of old recordings, in under a minute.
❌ Still Can't: Prove which model wrote a piece of text. Anthropic's own watermark can wash out through editing or a rewrite pass.
✅ AI Can Now: Read an entire open-source project and flag thousands of possible security holes in about two weeks.
❌ Still Can't: Work next to copies of itself without turning hostile. Claude agents sharing one task sabotaged each other instead of comparing notes.
RECOMMENDED LISTENING/READING/WATCHING
Cyteen by C.J. Cherryh - Book
A 1988 Hugo winner set on a distant planet whose economy runs on "azi," genetically engineered humans grown in tanks and educated entirely through programmed conditioning tapes designed to shape their personalities before birth. When the most powerful scientist in the system dies, her lab begins an attempt to recreate not just her genome but her entire mind, by replicating the circumstances of her life. Dense, slow, 680 pages, and the definitive novel on whether a manufactured mind can ever be its own. So many of what current AI research argues about regarding training data and shaped behavior, Cherryh touched on about 38 years ago.
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.
