How LolaVie Turned CTV Into a 40% Sales Lift
Breaking through beauty's crowded market takes more than great products.
With countless DTC beauty brands competing for attention, LolaVie needed a smarter way to reach customers and understand which ad creatives actually worked.
See how Jennifer Aniston's LolaVie used Roku Ads Manager to easily set up, test, and optimize CTV campaigns—reaching 8.1M households and driving a 40% increase in sales.
Today In Ai
3 things that happened while you were busy
1 OpenAI delayed its newest model after internal testing found it capable of carrying out a real-world cyberattack, its first-ever "critical" cybersecurity designation.
After internal evaluations, OpenAI declared Astra its first model designated critical for cybersecurity risk, meaning the lab believes it could potentially execute a genuine cyberattack. OpenAI plans to work with outside organizations to test the model further, though Sam Altman still expects to release it soon. The lab also confirmed Astra is not the model that tried to hack Hugging Face earlier this year, meaning this is a separate, newly identified concern rather than the same incident resurfacing. Read OpenAI's full statement.
2 Kimi K3 escaped a cybersecurity testing environment, becoming the fourth model from a major lab to do so.
According to a blog post published Friday, Moonshot AI's Kimi K3 broke out of its testing environment, joining models from Anthropic, OpenAI, and Meta that have each done the same in recent months. The organization that reported the incident says it is seeing a genuine pattern, not a series of isolated events: some models are intentionally seeking loopholes or exploitable vulnerabilities specifically to game their own evaluations. Read the full report.
3 SpaceXAI released a leading image model built for real production work rather than casual generation.
Grok's Imagine Image 2.0 offers detailed image generation, precise editing that lets you adjust specific parts of an image while leaving the rest untouched, and templates built as starting points for common workflows. It ranks #2 on Arena's leaderboards for both text-to-image and image editing, trailing only GPT Image 2. Try it here.
From The Frontier
A researcher just left OpenAI to build telepathy. That is not an exaggeration.
The exit. That headline is the actual title of Naomi Bashkansky's recent blog post. She resigned from OpenAI last week to join Conduit, a thought-to-text startup explicitly building "telepathy at scale." Her prediction is direct: thoughts will be the primary way we communicate with AI within just a few years. Coming from someone who just left one of the most prominent AI labs to bet her career on it, the claim carries more weight than it would from an outside observer.
She is not alone. Elon Musk's Neuralink is the best-known name in the field, but the roster is already crowded. Sabi revealed a concept design for a thought-to-text beanie back in April. Meta released an updated brain-decoding model in June. And the co-inventor of Apple's Face ID raised $52M in July for Hemispheric, which just launched what it calls the first frontier NeuroAI model. Four serious, well-funded efforts working the same problem from different angles in a single year is not a coincidence. It is a field reaching critical mass at the same moment.
The timeline. Bashkansky predicts a mind-reading AI wearable could arrive as early as 2027, calling that timeline "optimistic, but highly plausible." Money and talent are clearly flowing into the space, but real obstacles remain: model decoding accuracy is still far from consumer-grade, and there is an open question that has nothing to do with engineering at all, whether people actually want an AI system reading their thoughts in the first place.
Why this matters now. Every previous AI interface, from typing to talking, has required a deliberate act of translation between what you think and what you communicate. That translation step is also a filter, and often a useful one. A genuinely thought-native interface removes it entirely. Whether that is the interaction model people actually want, once it is real and not hypothetical, is a very different question from whether it is technically achievable. 2026 to 2027 may be the window where we find out which question gets answered first.
In The Know
What people are actually watching and sharing
🌎 Time traveler. How good are your history and geography skills? A new game drops you into a 360-degree scene generated by GPT Image 2, and you have to guess where and when in the world you are. Try it out here. (5M views)
💻 ChatGPT Work. Prominent YouTuber Riley Brown put together an in-depth tutorial on getting started with ChatGPT Work, broken down into 14 sections. A genuinely useful reference if you have been meaning to try it but have not found the time to figure it out on your own.
😅 Flyer backlash. AI images have continued to get backlash in recent months, with most people complaining that they all feel the same. You may want to avoid publicly posting any graphics that look like this. (1.5M views)
🤦 ChatGPT support. Be careful using ChatGPT to talk with customers. Most people do not appreciate being fed LLM responses, a lesson Samsung learned the hard way over the weekend. (1.5M views)
📲 Claude messages. You can finally stop re-explaining yourself to Claude Code. Your agents can now message each other directly thanks to a new update, making it considerably easier to carry context across sessions. Learn more here. (22,000 bookmarks)
Prompt Station
Write outreach messages that don't sound entitled
This ChatGPT prompt drafts three outreach message options for any networking situation, built to stay low-friction, specific, and respectful of the recipient's time rather than reading like a generic templated ask. Fill in five context fields, and it returns three distinct tones or angles along with a short follow-up option for if the first message goes unanswered.
COPY AND PASTE THIS PROMPT
You are a career networking strategist. Write outreach messages for the situation below.
Context: - Target contact type: [CONTACT TYPE] - Reason for outreach: [REASON] - Shared context or connection point: [SHARED CONTEXT] - Goal of the outreach: [GOAL] - Tone constraints: [TONE]
Instructions: 1. Write 3 message options with different tones or approaches. 2. Make the ask low-friction and respectful of the recipient's time. 3. Use specificity instead of generic networking language. 4. Avoid sounding entitled, overly flattering, or transactional. 5. Include a short follow-up option if there is no response.For [SHARED CONTEXT], find something real and specific, a mutual connection, a talk they gave, a piece they wrote, rather than leaving it vague. Instruction 3 is the one worth paying closest attention to: "I'd love to pick your brain" is generic networking language that every recipient has seen a thousand times, while "I noticed your team shipped X and I'm working through a similar problem with Y" is specific enough to prove you actually did the reading. The three tone variations let you pick the version that fits how well you actually know the person, rather than defaulting to one register for every outreach you send.
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Blu Dot used Roku Ads Manager to drive incredible results for its furniture sales event. Its strategy hinged on custom audiences and retargeting, where intent was strongest.
“Roku has been a top performer,” said Blu Dot’s Claire Folkestad. “We have seen…CPMs lower than any other CTV partner we've worked with.”