My big bets for AI in 2025

Hi Futurist,

I hope you all had a wonderful holiday season and have started 2025 full of energy. In the spirit of ‘New Year, New Me,’ I’ve given the newsletter a fresh new look. Hopefully, you like it! Personally, I believe 2025 will truly be a year of acceleration. The curve of exponential growth is becoming ever steeper. That’s what I’ll be diving into with you today. Therefore, I am sending you insights, inspiration, and innovation straight to your inbox. Let’s dive into the depths of the digital future together and discover the waves of change shaping our industry.

💡 In this post, we're dipping in:

• 📣 Byte-Sized Breakthroughs: AI in 2025 will leap ahead with reasoning models, multimodal intelligence, and agentic apps. My predictions show a future where AI doesn’t just assist—it transforms every industry.

• 🎙️ MarTech Maestros: This whitepaper gives you everything you need to create AI agents tailored to your needs—tools, insights, and strategies to turn them into your ultimate productivity drivers for 2025.

• 🧐 In Case You Missed It: From Krea’s custom trainings and Bolt’s Supabase integration to NVIDIA’s Project Digits and OpenAI’s Infinite Memory rumours, this roundup keeps you ahead with the latest tools and innovations shaping 2025.

Do you have tips, feedback or ideas? Or just want to give your opinion? Feel free to share it at the bottom of this dip. That's what I'm looking for.

No time to read? Listen to this episode of Digital Dips on Spotify and stay updated while you’re on the move. The link to the podcast is only available to subscribers. If you haven’t subscribe already, I recommend to do so.

Headstory: My big bets for AI in 2025

AI is accelerating. Not by a step. By a leap. The past year felt like watching fireworks—each launch more dazzling than the last. 2023 was the year of text and image. 2024 gave us video and speech. But 2025? It promises to take us further, faster, and deeper into a world dominated by agents, multimodality, and intelligence that feels almost magical. Are you ready for what’s next?

Here are my predictions for AI in 2025:

1. AI surpasses human intelligence: By the end of 2025, we’ll see AI systems outperforming human intelligence across various domains.

2. Reasoning models become the norm: Forget simple answers—AI will reason, debate, and make decisions.

3. Large token windows will be standard: Models will handle massive amounts of information seamlessly, enabling nuanced and context-rich responses.

4. Agentic apps, Text-to-Action, and AI agents: In 12 months, everyone will have a team of AI agents working on their behalf.

5. 2025 is the year of 3D: AI will create immersive 3D experiences, revolutionizing industries like gaming, design, and retail.

6. New AI models are multimodal by default: AI models that seamlessly integrate text, images, video, speech, and 3D will dominate.

7. The pace of development accelerates: By year’s end, ‘innovator AI’ may surpass human understanding in key areas.

8. The rise of Spatial Intelligence: AI will interact with the physical world, manipulating environments and hardware autonomously.

9. AI apps controlling your devices: AI apps are controlling your devices on your command, for example through speech.

10. At least one lab will claim ‘AGI’: The definition may remain debated, one lab will announce they’ve achieved Artificial General Intelligence (AGI).

These predictions aren’t just milestones—they’re signposts. Signposts urging businesses to rethink strategies, realign priorities, and ready themselves for the next leap. With AI agents poised to redefine workflows, spatial intelligence reshaping interactions, and reasoning models enhancing decision-making, companies must innovate or risk irrelevance. The mantra for 2025? More AI delivery, less AI talk. More action, less hesitation. More experimentation, fewer excuses.

Whether you work in legal, operations, HR, finance, or supply chain, the impact of AI is undeniable across every field. Marketing is no different. Marketers will need to rethink how they reach audiences, run campaigns, and manage budgets. AI won’t just help—it will drive smarter and faster decisions. It’s time to get practical and make marketing work for the future.

Here’s how AI will impact marketing in 2025:

1. AI becomes Martech’s new normal: SaaS tools will merge with agent platforms, creating systems that don’t just assist but actively perform. Products that fail to integrate these capabilities won’t survive.

2. SEO becomes Search Experience Optimization (SXO): With AI-driven search results on Google Gemini, OpenAI’s Search GPT, Bing Copilot, and Perplexity, most searches will no longer lead to clicks.

3. AI will reshape ad creation and placement: AI won’t just place ads; it will generate concepts, draft copy, manage budgets, and monitor performance in real-time.

4. Social media flooded with AI characters: Expect social feeds flooded with AI-generated personas. These “characters” will engage audiences, creating new opportunities—and challenges—for marketers.

5. Hybrid content creation: Marketers will combine AI’s efficiency with human originality. AI crafts data-driven frameworks while human creators infuse emotion, nuance, and brand authenticity.

6. Proactive decision-making through AI: AI evolves from reacting to results, to predicting them. Marketers will rely on data-driven insights to forecast consumer trends, tailor campaigns, and pivot before problems arise.

7. Smarter budget management: Real-time optimization is the norm. AI monitors where each dollar works best, shifting budgets instantaneously to high-performing channels for maximum ROI.

8. Brands must own their values: In an age of deepfakes and AI-synthesized messages, authenticity matters more than ever. Consumers will demand clear, honest, and value-aligned communication.

9. Data quality as a competitive advantage: Organizations with vast amounts of high-quality, domain-specific datasets will dominate their market by developing AI solutions that competitors cannot replicate.

10. AI truly gives loyalty its value: By integrating loyalty programs with AI-driven marketing, businesses will unlock deeper customer insights, boost retention, and create personalized experiences that resonate.

The stakes? Your survival. The prize? Dominance in a market driven by intelligence. AI is no longer just a strategy—it’s the foundation of everything. Businesses need a new engineering playbook, one that combines agility, innovation, and AI-driven precision. What you do in the next couple of years could define not just your trajectory—but the legacy of your entire organization.

The takeaway is clear: most jobs as we know them will become obsolete. In 12 months, everyone will have a team of AI agents working tirelessly behind the scenes, automating tasks, making decisions, and driving growth. 2025 is the time to act. To embrace AI, master its potential, and place yourself miles ahead of the competition. In past editions of Digital Dips, I shared tools to build your first AI workflows. Now is the time to use them. Go for it.

OpenAI’s o3 breaks barriers in artificial intelligence

TL;DR

OpenAI’s o3, announced just three months after o1, has shattered benchmarks across coding, math, and science. With an 87.5% score on ARC-AGI (above the human threshold of 85%) and incredible improvements in PhD-level science and software engineering, o3 represents a step into uncharted AI territory. Alongside o3, OpenAI introduced o3-mini, a faster and streamlined version optimized for coding, set to be publicly available in early 2025.

Read it yourself?

Sentiment

Online, the reaction is explosive—many are saying, “AGI is here.” o3’s performance has sparked conversations about whether computer science degrees still make sense, given that the way we write code will fundamentally change before current students even graduate. Soon, everyone will carry a "grandmaster programmer" in their pocket. The estimated IQ of o3, calculated at 157, places it in the top 0.0075% of humanity. People are going nuts over this groundbreaking leap, which some are calling the dawn of a new paradigm.

My thoughts

The launch of o3 confirms that the pace of AI development is accelerating faster than anyone anticipated. Your timeline needs to be adapted. What should a computer science student—or any knowledge worker—do in this rapidly changing world? The reality is stark: even if these models cost $2000/month, they’re still cheaper than hiring a graduate employee. The solution isn’t to resist change but to embrace it. Mastering AI collaboration will become the most valued skill in this new era. Practice working with AI. The future will belong to those who can effectively combine human and AI capabilities, as this combination will be far superior to either working alone. The global economy, as we know it, must adapt. We need new systems, new frameworks—because the machines are now smarter than most of us, and they’re venturing into conceptual and scientific realms that, for us, will remain like shadows of far-off lands. AIs will build and manage this new world, while we humans grapple with understanding it. Strap in—things are about to get very weird.

DeepSeek-V3 is a giant leap in AI at a fraction of the cost

TL;DR

DeepSeek-V3, a new AI model from China, is shaking things up. It has 671 billion parameters and was trained on a budget of just $5.6M. That’s over 10 times cheaper than its competitors, yet it performs as well as leading models like GPT-4o. And the best part? It’s completely open-source, meaning anyone can use it.

Read it yourself?

Sentiment

People online are blown away by how little this model cost compared to others, while still delivering top-notch results. And the fact that it’s open-source is just the cherry on top.

My thoughts

China’s relentless progress in AI development has been a recurring theme in previous Dips. Yet again, they’ve achieved something remarkable: developing a model nearly on par with leading alternatives at a fraction of the cost. If this trend continues, I wouldn’t be surprised if building your own AI model becomes financially viable for SMEs in just a couple of years.

More byte-sized breakthroughs:

• Google introduces Gemini 2.0 Flash Thinking
  Google's experimental Gemini 2.0 Flash Thinking model is built for reasoning, promising stronger logic by "thinking through" its answers. Available in AI Studio, it targets complex problems in programming, maths, and physics. This experimental model is part of Google's push to compete in the AI reasoning space, following the surge of similar models by other tech giants.

• ChatGPT expands app integration on macOS
  ChatGPT now connects with apps like Notion, Apple Notes, and coding tools such as Warp and IntelliJ IDEA on macOS. With o1 reasoning models and Advanced Voice mode, you can manage workflows, and even navigate apps hands-free using your voice. This update empowers users to work smarter and faster across a growing range of applications, making multitasking more intuitive than ever before.

• Hume introduces Octave for emotional AI
  Hume AI’s Octave takes emotional intelligence in AI to the next level. This tool enables machines to understand nuanced human emotions through voice, text, and facial cues. By integrating Octave, developers can build AI that connects on a deeper, more human level. It’s designed to create meaningful interactions, bridging the gap between technology and emotion.

A must-see webinar, podcast, or article that’s too good to miss.

AI agents are the new workforce behind every knowledge worker

As highlighted in the headstory of this newsletter, 2025 will mark the rise of AI agents as indispensable tools for knowledge workers. These intelligent assistants combine reasoning, real-time data access, and task automation to execute complex actions independently. From personalized recommendations to automating transactions, they will redefine productivity. This whitepaper explores how AI agents work, their transformative potential, and why businesses should embrace them to stay ahead.

A roundup of key updates that are too cheesy to ignore.

• Krea Editor introduces custom trainings, enabling users to seamlessly add real products to images in seconds.

• Bolt.new announces official Supabase integration, making seamless data handling easier than ever.

• LTX Studio debuts Face Expression, letting you fine-tune characters’ emotions with presets or custom controls in seconds.

• Gemini launches Spotify integration, letting Android users stream songs and discover playlists directly in the app.

• OpenAI’s GPT-5 fails to meet expectations as a successor to GPT-4, with users citing minimal upgrades.

• OpenAI’s Infinite Memory for ChatGPT is rumoured to roll out soon, enabling seamless access to all past chats.

• OpenAI and Microsoft reportedly define AGI as AI systems generating $100 billion in profits—a goal still years away.

• OpenAI misses its 2025 deadline for the promised opt-out tool, leaving users without a way to exclude data from AI training.

• OpenAI is now focusing on developing Artificial Super Intelligence (ASI), aiming to surpass human intelligence in various tasks.

• Groq launches Appgen, a platform for creating and sharing micro-apps with lightning-fast code generation in milliseconds.

• Arizona adopts an AI-driven curriculum in some charter schools, replacing teachers with algorithms for two hours daily.

• Grok launches its standalone mobile app, offering a free AI assistant and image generator from xAI.

• Grok and xAI rumored to outpace OpenAI with upcoming Grok 3 model, fueled by last year's massive investments in GPU clusters.

• South Korea enacts their AI law, becoming the second country to pass a comprehensive AI law inspired by the EU AI Act.

• Tldraw unveils “Computer,” a Gemini powered tool linking components for seamless workflows with lightning-fast execution.

• Amazon unveils SageMaker Lakehouse with zero-ETL integrations, simplifying data workflows across applications like Salesforce and SAP.

• Engineers achieve quantum teleportation over active internet cables, paving the way for secure quantum networks alongside regular internet traffic.

• A Chinese research team publishes a step-by-step guide to replicating OpenAI’s o1 model, emphasizing reinforcement learning techniques.

• Meta envisions a future where AI-generated users fill social media platforms, reshaping digital interactions at scale.

• Meta scales back fact-checking and relaxes content moderation policies, sparking debates over platform responsibility.

• Stanford debuts STORM, an AI tool that curates Wikipedia-like reports by analysing major findings across hundreds of websites.

• Sky-T1-32B-Preview, an open-source reasoning model trained for just $450, outperforming o1-preview in MATH and AIME2024 benchmarks.

• Perplexity starts integrating ads directly into queries, moving closer to a zero-click future.

• Microsoft is set to invest $80 billion in AI-enabled data centers during fiscal year 2025, with over half of this investment allocated to the US.

• NVIDIA announces Project Digits, a personal computer priced at $3,000 that’s 1,000 times more powerful than a typical laptop.

• Alibaba Qwen launches its AI web app with support for artifacts rendering, file uploads, and advanced interactions.

This was it. Our twenty-seventh digital dip together. It might seem like a lot, but remember; this wasn't even everything that happened in the past few weeks. This was just a fraction.

I’m here to help you. In the coming weeks, I’ll be sharing all sorts of things with you. From a custom GPT that always improves your prompts, to a complete list of over 125 AI tools, and a 5-day email course to help you put your AI strategy on paper. I want to do everything I can to get you started. Do you have questions, want to brainstorm, or have great coffee? Hit reply and send me a message!

Looking forward to what tomorrow brings! ▽

-Wesley