🤓 This Week in AI
PLUS: Smarter AI Is Here—And It’s About to Change Everything
I wanted to kick off with a cheerful 'Happy New Year,' but here in LA, 2025 has started on a devastating note.
As I write this from Brentwood, just blocks from the evacuation zone, wildfires continue to tear through our city.
My heart goes out to everyone affected—I know many of you are facing this firsthand. Please stay safe 💙.
If, like me, you could use a moment to focus on something lighter, let me rewind to my calmer start to 2025.
I spent most of the holiday living my best pajama life, binge-watching the Dublin Murders on Starz (so good), and supporting my local bakery's chocolate muffin business. (You're welcome, economy.)
I also had time to think about what 2025 has in store. From where I'm sitting, this is the year AI becomes unavoidable – not just as a topic of discussion, but as a fundamental part of how we work and live.
Generative AI already has the fastest adoption rate of any technology in history.
Faster than the internet. Way faster than PCs. And faster than anyone could have predicted. I mean, just look at the slope of this line 👇.
Just in the past few weeks, we’ve seen surprising leaps forward like OpenAI’s o3 model (more on this below), and the pace is only picking up.
I believe we'll look back at 2025 as one of the most pivotal years in history—and I don't say that lightly.
As OpenAI CEO, Sam Altman, wrote in his blog this week: “We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies.” The dawn of “AI employees” and “AI coworkers” is upon us.
While I'll continue to be deeply immersed in this fascinating space to bring you what matters most, one reality seems clear: the gap between those who understand AI's capabilities and how to work with these systems and those who don't is widening—fast.
And with AI skills quickly becoming a must-have for hiring managers, those who dive in now will have a big advantage over those who wait.
If you've been watching from the sidelines with curiosity, consider this your invitation to dive in and start experimenting.
Don't worry though—you've got this.
If you’re reading this newsletter regularly, you’re already ahead of the curve. And for the many of you who’ve been through my half-day workshop, continued experimenting, and keep up with my updates, you’re likely in the top 5-10%, even compared to untrained power users.
More importantly, you've got me in your corner. Together, we'll turn these shifts into opportunities...
As I always say, if you don’t know where to start, just hit reply and ask away. I read every email.
Ok, let’s dive in.
Here’s what you need to know about AI this week (clickable links appear in orange in emails and underlined in the Substack app):
OpenAI introduced its new o3 model—a major milestone in how AI scales and solves problems.
This development is a REALLY big deal. Here’s why:
For the past several years, scaling AI has meant building bigger models with more data—which has delivered remarkable progress.
But as the models grow larger, the cost of training and operating them skyrockets, reaching tens of billions of dollars, while gains in performance become incrementally smaller. This has pushed researchers to focus on smarter, more efficient approaches to improve model performance.
That’s where o3 comes in.
Rather than relying on size alone, o3 takes a new path by focusing on reasoning (or “thinking”) and adaptability—traits that make AI smarter, more efficient, and more like a true problem-solving partner. It:
Solves Problems Like We Do: o3 takes on challenges step by step—breaking them into smaller parts, brainstorming multiple solutions, and selecting the best answer. Most AI systems are designed to produce rapid responses based on pattern recognition, but o3 takes the time to think through problems, ensuring its solutions are well-reasoned and accurate. Whether it’s solving complex puzzles or generating detailed plans, it reasons through tasks in ways that feel human.
Excels at Solving New Problems: Unlike earlier AI, which works best with familiar tasks it has been trained on, o3 shines in handling new and dynamic problems. It can tackle challenges it hasn’t seen before, adapting to new scenarios to come up with solutions to novel problems in real time.
Evaluates and Improves Its Own Work: o3 doesn’t just produce answers—it reviews, tests, and refines them. For example, it can generate code, run it, and fix errors on its own, ensuring accuracy and reliability. Think of it as an AI that doesn’t just hand you a draft but actively makes sure it works before presenting it. This feedback loop makes o3 more reliable for more high-stakes real-world applications, where mistakes can be costly.
o3’s surprising advancement will further speed up the development of AI agents—autonomous systems that can take actions and complete tasks on their own with minimal human input—and progress towards superior intelligence that surpasses human abilities. 🤯
It’s going to be a wild wild year.
But there’s a catch: o3’s cutting-edge capabilities are expensive to run, so OpenAI is already working on scaled-down versions to make this tech more accessible for businesses.
Meta rolled out AI-generated users] on Facebook and Instagram to drive engagement. These AI personas, labeled as “AI managed by Meta”, had bios, profile pictures, and shared content like human accounts. Here is what one of them looked like:
Unsurprisingly, the move didn’t go over well. Users were confused about the purpose of these bots and criticized the inability to block these profiles, low-quality imagery, and misleading responses in chats.
Following public and media scrutiny, Meta seems to have either removed or limited access to these profiles—they’re no longer visible on the platforms.
Given all the data Meta has on user preferences and behavior, I continue to be baffled by their use of AI. Last year, they gave us AI celebrity avatars—only to abandon them after they failed to spark any user interest.
This year, another offering that feels completely disconnected from anything users actually want or value…
ChatGPT maker OpenAI officially published its plan for the company’s controversial shift from a nonprofit structure to a for-profit company to secure the substantial funding needed for the next chapter of AI development.
Chinese AI startup DeepSeek released a new powerhouse language model (DeepSeek-V3) that sets new benchmarks in the open-source AI world with performance rivaling industry giants including Meta, OpenAI and Anthropic at a fraction of the cost.
Apple’s AI-generated news summaries are under fire after false alerts claimed a BBC story reported that Luigi Mangione, an alleged murderer of a US healthcare CEO, had killed himself, and that Rafael Nadal had come out as gay. While Apple has promised fixes, critics argue the errors harm media credibility and are calling for the feature’s removal.
AI engineer and AI consultant are the fastest-growing jobs in the U.S., according to LinkedIn's latest report, but significant skills gaps are leaving companies scrambling to fill these complex, critical roles.
AI expertise is becoming essential across functions—marketing, publicity, operations, you name it.
And I can’t stress this enough: If you work for yourself or your company isn’t training you on top tools like ChatGPT or Claude, proactively find credible and trusted sources for in-depth training. Internal tools can’t keep up with the pace of advancements, and learning AI takes time and hands-on experience, regardless of your role or seniority level. Getting ahead now will pay off later.
Religious leaders are experimenting with AI to write sermons, translate them in real time, streamline research, and answer questions. While the tech offers new ways to boost reach and engagement, critics warn of serious risks, including chatbots inventing scripture or giving misleading advice on sensitive topics. They also argue that AI lacks the empathy and authenticity central to faith.
As I always say, understanding what AI does well—and where it currently falls short—for any specific use case is key. AI is great for research, translation, handling simple queries (like 'When’s the next service?'), and even partnering with experts to write sermons
But answering sensitive questions about faith? That’s a big no-no.
Fable, a popular book app, apologized after its AI-generated summaries offended users with racist and insensitive comments, like suggesting a Black reader try white authors.
After the incidents, the company rolled out safeguards like AI disclaimers and feedback tools, but users are pushing for more accountability, saying mistakes like these destroy brand trust and undermine the app’s commitment to inclusivity.
As you begin to experiment with AI for your brand, focus first on non-consumer facing use cases, where mistakes carry lower risk. For anything public-facing, stress-test before launch like your reputation depends on it (because it does) and set up feedback loops to address any issues fast.
Scientists are using generative AI to build “virtual cells” that can predict how drugs, mutations, or viruses affect the human body, speeding up breakthroughs in drug discovery and disease research.
New research from Adobe found that AI assistants drove a 1,300% surge in retail web traffic during the holidays, with consumers increasingly relying on ChatGPT and other chatbots for recommendations and price comparisons.
Travis Scott X Nike ad used 5000+ images generated by Midjourney to shape the visual identity of the campaign for director Gibson Hazard.
Other Interesting Finds 📌
While AI consumes most of my waking hours these days (and makes guest appearances in my dreams), my curiosity has always run deep and wide.
Those who remember my Clubhouse days know I can get equally excited about psychedelics, all things psychology, sexuality, the science of dreams, and of course brilliant marketing campaigns.
So, when something interesting crosses my radar, I’ll drop it here into my collection of non-AI finds—where the only common thread is that it made me think, laugh, or raise an eyebrow...
Think of this little section as your occasional reward for making it through the AI updates 😄
This week, I’ve got:
1️⃣ Solid brand strategy + flawless execution = brilliant & powerfully connective messaging from Eli Lilly
I’m OBSESSED.
2️⃣ A New York Times piece on dating predictions for 2025. And since I plan to be dating this year (if AI lets me see other people – we were pretty exclusive in 2024), I thought I'd see what we're in for.
Here’s an overview:
Rise of the Offline Lover: Grounded, steady, “logged-off” personalities are replacing ambition as the new ideal. People value staying informed without overwhelm and prioritizing real-world connections over endless scrolling. (Maybe I should mention this to my AI... 👀)
More Nonmonogamy and Perimenopausal Pride: Nonmonogamy continues to grow in popularity, and more women are openly discussing the realities of perimenopause with humor and confidence.
Fewer Dates, More Hookups: Inflation and economic pressures are expected to lead to fewer traditional dates and a rise in casual hookups. “Men are not going to be taking women out as they used to.” (Yikes!)
Expecting More from Apps: Daters want apps to foster deeper, more playful, real-world connections instead of transactional matches. (The bar is so low it's underground at this point.)
Less Dating Across Party Lines: Political differences are a growing dealbreaker, with singles seeking partners whose values align with their own.
Parent-Funded Dating App Subscriptions: Parents and grandparents are stepping in, funding dating app memberships and matchmaking services for their children. (This one is hilarious, like PLEASE find one good one – I’ll even pay 😁)
Prioritization of Platonic Romances: Many women are opting out of traditional dating, choosing instead to prioritize deep, fulfilling friendships over romantic relationships.
A.I. Will Help Solve Dating Woes: AI is becoming a digital wingman, handling profiles, coaching conversations, and planning personalized dates. (I’m pretty skeptical of this one…)
Long Distance Will Be More Popular: Long-distance relationships are growing, driven by the internet and the resurgence of phone calls for intimacy.
Setting Up Friends on Dates: Blind dates and matchmaking through mutual friends are gaining popularity as people seek deeper, community-driven connections. (Friends, if you're reading this...👋)
In case you missed the last edition, you can find it 👇:
That's all for this week.
I’ll see you next Friday. Thoughts, feedback and questions are always welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi
💙💙💙 P.S. A huge thank you to my paid subscribers and those of you who share this newsletter with curious friends and coworkers. It takes me about 12+ hours each week to research, curate, simplify the complex, and write this newsletter. So, your support means the world to me, as it helps me make this process sustainable.