Issue #51: OpenAI's o3, Meta's AI Personas, AI dominating jobs already

Good morning.

Welcome to Q1, 2025.

As I said last year, you’ll see a ton of AI agent capabilities and usage this quarter.

Everyone’s working on the all-purpose AI assistant that you can do anything with (and the key feature being that it’s agentic and semi-autonomous).

Everyone’s also working on specialized agents for various industries.

A business that has many parts either fully or semi-autonomous agents running the show is a very real possibility.

You can, already, outsource most of your marketing to agents and automations. Some marketing, like SEO, can be fully automated. Same with ads. Same with content and social media.

But the irony of getting there is to actually not build it to be autonomous out of the gate. You start with humans in the loop. I also think you’ll get the most value from AI agents if you always include a human somewhere.

Anyway, a quick round-up email for today. This newsletter will likely pivot in the next month to go a little bit beyond just marketing.

Let’s get into it.

—Sam

IN TODAY’S ISSUE 👨‍🚀 

• OpenAI's o3 model and the future of human intelligence.

• The Rise of AI Personas: Meta's Next Big Bet

• Real-Time AI Social Media Content

• AI Is Dominating the Job Market

Let’s dive in.

OpenAI's o3 Model for Researchers

OpenAI recently unveiled their o3 reasoning model. It’s significant because from what I’ve seen, it’s expanding on o1 capabilities in a dramatic way that puts some level of model autonomy in reach.

Right now, o3 is in the researcher-exclusive iteration that’s not yet ready for public consumption. We’ll see when, and how, it’ll be released.

If o1 was about building on what GPT-4 could do, o3 is a leap towards what comes next.

From the initial reports, o3 isn’t just faster or more accurate (though it is). It’s designed to solve some of the most persistent challenges in AI, like context depth, reasoning over extended conversations, and adaptive learning.

Researchers are hailing it as a step toward truly dynamic, human-like interaction.

I don’t think too hard about benchmarks for these models. My benchmark for any AI tool or model is:

Am I still using it after 4 weeks? Is it reliably producing quality outcomes (not just content, I’m talking actions and outcomes)?

If so, I keep using it.

The key right now is:

Combine whatever you’re doing with AI.

Maybe one day, AI will do everything. Who knows? In the immediate future (next 12-18 months), it doesn’t matter.

The capabilities of o3 being discussed by researchers are just enforcing how powerful AI is becoming, and how necessary it is that you integrate what you’re doing with it.

As for the main discoveries within o3 from researchers, there’s a lot to unpack.

Here are a few notes worth your time and attention:

1. Memory That Feels Alive

The 03 model introduces dynamic memory expansion, meaning it doesn’t just “recall” past prompts—it adapts and builds a nuanced understanding over time.

For example:

In long-term research projects, o3 could keep track of multiple threads, connect them across weeks or months, and dynamically surface connections without prompting.

For researchers, this means less time re-contextualizing the model and more time pushing into unexplored territory.

(This is where all the “LLMs are just stochastic parrots!” need to sit down and be quiet).

2. Open-Ended Problem Solving

Where prior models excelled at structured tasks, o3 thrives in ambiguity. It’s reportedly better at tackling abstract, multi-step problems with no predefined answer.

On top of this post, a case study from a team at MIT shows they used o3 to generate hypotheses on quantum material properties and refine them iteratively.

The result?

A potential breakthrough pathway they hadn’t considered.

3. Specialist Knowledge on Demand

While o1 partnered with Future for niche content integration, o3 took a different route: collaborative embeddings. This means the model isn’t just smarter—it actively learns from your inputs, offering progressively better insights as you interact with it.

Imagine a medical researcher running live simulations or a policymaker fine-tuning policy frameworks in real-time.

That’s the level we’re talking about here.

The restricted nature of o3’s release isn’t just a play to generate buzz. The reality is, as smart as researchers and doctorates at OpenAI are (and believe me, they’re insanely smart), it’s very hard to predict exactly how o3 usage will play out once millions of users get access.

OpenAI appears to be playing a long game here.

By releasing it into controlled environments, they can refine the model while ensuring it delivers on its promises.

What You Should Be Watching

The research feedback loop created by o3 will shape the next decade of AI development. If you’re in tech, marketing, or business strategy, here’s why you should care:

• Adaptive AI will redefine business tools. Imagine an AI assistant that evolves with your team’s workflows and nuances. You don’t have to prompt it for anything, it just “knows” and “acts” for you.

• Knowledge work is evolving. Specialist AI applications will become more personalized and effective.

• The gap between leading and lagging companies will grow. Those who invest early in understanding these tools will dominate.

• Memory at scale could redefine applications in education, research, and enterprise knowledge management.

• Open-ended thinking will allow new solutions in spaces where intuition and adaptability are key.

• Collaborative intelligence will make the line between human and AI contribution blur even further.

If o1 was OpenAI's big commercial move, 03 is its moonshot.

It’s also a model that makes a semi-autonomous business possible, in theory. We’ll see if anyone (myself included) can figure out how to implement and build one.

The Rise of AI Personas: Meta's Next Big Bet

Meta just unveiled its next step in AI:

AI-generated profiles and characters designed to interact with you (and other AI characters) across its platforms.

From Facebook to Instagram, these AI personas are somehow central to Meta’s bold strategy to drive engagement and reclaim dominance in the social media landscape.

It’s shifting Meta away from social media to just media.

Personally, I don’t really care much for “talking” to an “AI character”. It doesn’t appeal to me. But it does appear to tens of millions of users who are doing so across hunderds of apps. I believe the biggest app in this space, for now, is Character AI.

At the same time, I’m glad to see this because I think we’ve been tricked into believing social media was about social. Out of the hundreds or thousands of “friends” you have following you, how many are actual friends?

Anyway, for marketers this could be an interesting experiment.

Can your product be a character? Can your company be a character? Should your marketing make use of these characters?

Let’s break it down.

Meta’s vision is to create customizable, AI-powered avatars and assistants that users can interact with directly.

These personas will come equipped with distinct personalities, tailored responses, and deep integration into the Meta ecosystem.

Here’s the pitch:

• Need a travel guide? Your AI profile can recommend destinations, build itineraries, and even book tickets.

• Want a workout buddy? An AI persona could create and track personalized fitness routines.

• Just bored? Meta’s AI characters are designed to chat, entertain, and keep users engaged.

On the surface, it’s a clear play to boost time spent on Meta platforms. But there’s more happening here than just adding bells and whistles.

Meta’s Engagement Reinvented

Social media engagement has been plateauing. Younger audiences are flocking to TikTok, Discord, and niche platforms, leaving Meta searching for a way to recapture their attention.

AI personas could be the answer.

By offering users unique, interactive experiences—ones that feel personal and dynamic—Meta might create a stickiness that traditional social media no longer provides.

But there’s an even bigger opportunity lurking beneath this move:

Data.

AI personas interact and learn.

Every conversation, every query, and every customization feeds Meta’s algorithms, providing insights into user preferences, behaviors, and trends. This allows Meta to:

• Refine ad targeting by knowing what you discuss with your AI assistant could unlock a new level of precision for advertisers.

• Redefine content delivery from AI personas. They could serve as curators, delivering hyper-tailored content to keep users hooked.

• Expand Meta’s AI training data, again, from interactions with these personas.

More data leads to smarter AI, which leads to better engagement, which leads to even more data.

This brings me to the conclusion that Meta isn’t just adding AI personas to stay trendy.

This is part of a larger shift toward social platforms as immersive ecosystems, where AI powers deeper relationships between users, brands, and content.

For marketers, creators, and business owners, this move could bring profound changes.

For example, hyper-personalized ads could come from this.

AI personas could become intermediaries between brands and consumers, delivering pitches that feel conversational rather than intrusive.

Conversations with AI profiles may become a key performance indicator for brands. How many convos do you have? How do they “convert”? I think we’ll see the application of traditional funnel metrics to this at first, but most of those metrics are for models that no longer exist or work well.

Expect AI personas to help creators produce content, manage communities, and even collaborate on projects.

The Big Picture

Meta’s push into AI personas is a bet on the future of interaction.

If successful, it could redefine how we use social platforms, blending the transactional and the conversational.

Conversational commerce is already a thing.

On a more philosophical note, if the outcome is that people stop thinking of “social media” as “social”, and spend less energy and time on it, that’s a good thing.

Social media was never about your social life. It was always a play for capturing data, training models and building algorithms. You were always the product, and you still are.

The AI-Dominated Job Market: Growth We’ve Never Seen

ZoomInfo published eye-opening data on the state of AI in the workforce, and the implications are pretty nuts.

Their findings reveal a rise in AI-related roles across industries, skill levels, and geographies over the last few years.

If you’ve been wondering when AI would truly reshape the labor market, this data makes it clear: the shift is already happening.

Where AI is Already Changing the Game

ZoomInfo’s report identifies four key areas where AI-focused jobs are booming:

1. Data Science and Machine Learning:

These roles are leading the pack, with demand for data scientists up 74% since 2020. Companies are doubling down on AI-powered insights to stay competitive, especially in finance, healthcare, and marketing.

2. AI in Customer Service:

AI-driven chatbots and virtual assistants are fueling demand for roles that blend tech and communication skills. Think of positions like AI integration specialists or bot training engineers—jobs that didn’t exist a decade ago.

3. Manufacturing and Logistics:

AI is transforming operations behind the scenes. Robotics programming and supply chain optimization are now some of the hottest skillsets, with opportunities for both technical experts and hands-on operators.

4. Leadership in AI Strategy:

Organizations need executives who can connect AI capabilities to business outcomes. Roles like AI Transformation Manager and Chief AI Officer are emerging as strategic must-haves.

What’s especially striking about this report is how AI-focused positions are no longer confined to Silicon Valley or advanced degrees.

ZoomInfo highlights two trends specifically that are actually reshaping the job market.

The first is the rise of user-friendly AI platforms. With ease of access to highly powerful easy-to-use platforms, even non-technical roles can leverage AI tools.

Marketers, HR specialists, and sales teams are increasingly expected to integrate AI into their workflows.

The second driving factor is AI integration, where the expert has to come in.

Companies are seeking professionals who blend AI knowledge with domain expertise.

For instance, a healthcare analyst who understands predictive modeling or a logistics manager skilled in machine learning optimization.

It doesn’t take a genius to run many AI tools, but to make them effective within, someone has to make that integration.

This still shows the growth of AI roles is a boon for innovation, but it’s not without its hurdles.

If it shows anything, it shows my suspicions are mostly correct (though I’m wrong about a lot of things):

The rise of AI jobs is about how every role is evolving and changing. They’re collaborators, co-creators, and so on.

Expect more roles that mix human creativity with AI efficiency and use.

Keep an eye on this shift. Inside this shift, there’s space to open up new business models, new forms of work, and new ways to generate revenue.

Last year, James Schramko invited me to be on his podcast.

I’ve learned a ton from James over the years. He’s a clear, sharp voice on growing a business that you should pay attention to.

In this episode, we talked about how AI automation tools are transforming workflows, replacing jobs, and shaping the future of business.

One of the most significant ways businesses are benefiting from AI is through the automation of routine tasks.

If it’s repeatable and can be described legibly, broken down into steps, it can be automated.

AI automation tools are now allowing companies to automate workflows that were previously dependent on human intervention.

With the o3 model, a lot of that will get better and even replaced.

Listen to the episode here.

Talk soon,
Sam Woods
The Editor