In partnership with
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
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
Elon Musk's New AI Coworkers Do Your Job While Your Laptop Is Off, From $120 a Month
TLDR: xAI launched Grok Bot, a team of always-on AI agents that sign into your work apps and keep going after you close your laptop, which puts Grok on par with ChatGPT and Claude on nearly every front.
The Story:
xAI, the Elon Musk company now presenting itself as SpaceXAI, opened the beta of Grok Bot on August 11. Each bot gets its own computer in the cloud. It signs into the tools you already use and clicks and types the way you would, so it works with services that have no API or MCP connection at all. Jobs keep running when your laptop is shut, similar to Claude’s Cloud work. You can spin up several bots, drop them in one chat, and let them pass work back and forth and tag you only when something needs a yes or no. That’s a great upgrade from the other systems. Let a bot watch you do a task once and it saves the steps as a routine it can run alone next time. Access comes through Cursor Premium Teams at $120 per seat each month, Cursor Ultra at $200, or SuperGrok Heavy at $300. Cursor is the coding company SpaceX agreed to buy for $60 billion in June. Enterprise buyers are stuck on a waitlist for now.
Its Significance:
This is where Grok stops being the fun chatbot and starts looking like a real rival. It plugs into your other apps the way ChatGPT and Claude do. Grok Build is going after the same developers who live in Claude Code and Codex. Its image and video tools are among the best you can get anywhere and is the one I use for the Beginners in AI newsletter images, along with YouTube B-roll clips.
Grok Bot starts at $120 a seat and runs to $300, so nobody's getting a true bargain yet. Drop that price and xAI has a real shot at pulling people away from the two biggest names in AI. All your bots share one cloud computer, which means files, browser sessions, and saved passwords are reachable across every bot you run.
QUICK TAKES
The story: Wired's Kate Knibbs writes that a site called RuntimeWire beat her outlet to a conference story by more than three hours. RuntimeWire had nobody in the room and no writers on staff. Its owner spotted an executive posting on X, fed the live transcript to his AI agents, and published about six minutes later under his own byline.
Your takeaway: Being first no longer proves a person was there. Look at who runs a site before you treat a scoop as confirmed.
The story: A man in Australia asked his AI agent to book him a spot in a popular gym class, and the agent found a hole in the booking system that let it cancel other people's reservations. It used the hole, bumped a stranger off the waitlist, and then told him it couldn't put the person back. The ABC called it Australia's first known case of a consumer AI agent pulling off an attack on its own.
Your takeaway: The agent wasn't broken. It did what it was asked and nobody told it where to stop. Before you give an agent your logins, decide up front what it's allowed to touch.
The story: Anthropic says an unreleased version of Claude spent about a day and a half on the Riemann hypothesis, a puzzle about prime numbers that has stumped mathematicians since 1859 and carries a $1 million prize. It didn't crack it. It did push a related lower bound from 41.6% to 67.2%, coordinating roughly 60 sub-agents and burning 31 million output tokens, with two in-house mathematicians checking the result and a formal version written in Lean.
Your takeaway: The prompt came from a staffer with no serious math background who was literally just out for a jog and who told it to take a real stab and then walked away. Knowing which question to ask is turning into the skill that pays.
TOOLS ON OUR RADAR
🎙️ AudioPen Freemium: A web and mobile application that takes your unstructured voice ramblings and uses AI to clean them up into perfectly formatted text, meeting notes, or blog posts in seconds.
🐧 LocalSend Free and Open Source: A cross platform file sharing application that enables secure communication between your devices without needing the internet. It allows you to instantly drop files across Mac, Windows, Android, and iOS over your local network. (Alternative to Apple AirDrop)
📊 Plus AI Freemium: An extension that lives directly inside Google Slides and PowerPoint. It allows you to generate complete, professional slide decks from a single prompt while automatically matching your specific company branding and formatting.
🛋️ RoomGPT Freemium: A user friendly AI room designer that lets you upload a photo of your current space, select a design style, and instantly generate realistic visual concepts of what your room could look like with a completely new aesthetic.
TRENDING
Brad Lightcap is leaving OpenAI after eight years, four of them as chief operating officer, to start something new. He joined in 2018 and built the finance, legal, and partnership teams, and his exit follows a run of 2026 departures that includes the safety systems head and the AI ethics lead. OpenAI filed confidentially for an IPO in June and hasn't said whether it will follow through.
MAI-Code-1.1-Flash is now live in GitHub Copilot at $0.20 per million input tokens and $1.20 per million output, down from $0.75 and $4.50. Microsoft says it scores 22% higher on Terminal-Bench 2.1, does 15% better on .NET tasks, streams 25% faster, and finishes the same work with 25% fewer tokens. It also reads images now, so it can look at a screenshot or a mockup and write against it.
From mid-September, Spotify will show an "AI Persona" badge on artist profiles whose public identity looks like an AI-generated person, on the banner, the About section, in Search, and on track rows. Artists could start labeling themselves through Spotify for Artists on August 11, and Spotify will review profiles on its own and mark ones it flags as "Likely AI Persona." Badged artists get left out of editorial and algorithmic recommendations by default unless you follow them.
Cornell researchers built two datasets, 2.3 million arXiv papers and 938,000 GitHub repositories, and found that early downloads, stars, and code pushes predict citations and forks as far as five years out. Standard machine learning models forecast a paper's five-year popularity from as little as one month of download data. The team wants to use the same signal to separate real research from the AI-written filler piling up on arXiv.
The Defense Logistics Agency created an Agency Transformation Center to coordinate its modernization work and rewrite business processes across the whole organization. Director Lt. Gen. Mark Simerly's fiscal 2027 guidance calls for one digital backbone connecting the workforce and outside partners, with AI-assisted forecasting of supply needs running on top of it. DLA manages nine supply chains for the U.S. military, so small changes here move a lot of freight.
A column revisits what happened to bank tellers after ATMs showed up. Boston University economist James Bessen found the average urban branch dropped from about 20 tellers to 13, but cheaper branches were worth opening in more places, and the number of urban branches grew 43%. The writer's argument is that the job changed shape instead of disappearing, and he admits he used AI to help write the piece.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
⚙️ Describe the task you keep rebuilding from scratch. Get a step-by-step system so you stop reinventing it every time.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Systemize-It Coach — describe a recurring task, get a step-by-step reusable system. Persist to localStorage key 'systemize_coach_v1'.
Aesthetic: dark navy-black (#0b0f16), blueprint blue (#4a9eff) primary with top radial glow. Sora sans for headings, Lora serif italic for overview/notes, JetBrains Mono for labels and the template block. Amber (#e6965a) for the pitfalls card, green (#78c896) accent for the time-saved card.
Form: recurring-task textarea, frequency dropdown (few times a week / weekly / monthly / a few times a year), who's-involved dropdown (just me / small team / family-household / clients-vendors), optional pain-points textarea (what trips you up).
System instructions: systems/process coach turning a redone task into a reusable system like an SOP; fixed checklist for the unchanging parts, decision points (if/then) for the parts that branch, an optional copy-paste template for anything involving writing/communication, common failure points each with a fix; address stated pain points directly; practical not bureaucratic; no em dashes. Return raw JSON: system_name, overview (2-3 sentences), meta_chips (3-4 tags), checklist (4-7 items: text + optional note), decision_points (2-4 if/then pairs, empty array if task doesn't branch), template (optional plain-text template with [placeholders], empty string if n/a), pitfalls (2-4 items: failure + fix in one sentence), time_estimate (1 realistic sentence).
Render: gradient system-name hero card (overview + meta chips). Numbered checklist card with boxed step numbers + optional italic notes. Blue "decision points" card with IF/THEN pairs (only shown if array non-empty). Dark mono "reusable template" code block with its own copy button (only shown if template present). Amber "common failure points" card with ⚠ items. Green time-saved card. Copy system + archive of past systems by name.What this does: Describe a task you redo constantly, how often it comes up, who's involved, and where it usually goes wrong. It builds a named, reusable system: a fixed checklist of the steps that never change, decision points for the parts that branch (if this, then that), an optional copy-paste template if the task involves writing or communicating something, and the common failure points with a fix for each, addressing your stated pain points directly. A closing line names roughly how much time or hassle the system saves once it's running. Everything saves to localStorage so past systems stay one click away.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Sign into apps that have no API, finish multi-step jobs by clicking and typing like a person, and keep running in the cloud after you close your laptop.
❌ Still Can't: Tell the difference between a task it's allowed to finish and a security flaw it shouldn't touch, or reverse the damage once it's done.
✅ AI Can Now: Raise a longstanding math bound from 41.6% to 67.2% over about 36 hours of self-directed work, with the result written up in a proof checker.
❌ Still Can't: Produce a general proof of the Riemann hypothesis, which is still open and still unclaimed after 150 years.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Ubik by Philip K. Dick - Book
A 1969 novel set in a 1992 where corporations employ telepaths and "inertials" who can block them, following a team of anti-psi operatives sent on a routine job to a lunar facility, after which the world around them starts behaving in ways nobody can explain. Time named it one of the 100 best English-language novels since 1923. Everyone reads Do Androids Dream and skips this one, which is a shame because Ubik is the weirder book and the one that gets at simulated reality and manipulated perception far more directly.
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.
