What. A. Year. Biden dropped out, and Kamala dropped in. Also in, the first felon US president. There was an assassination attempt… but it’s the Hawk Tuah girl who made a killing. In other gun news : Olympics shooters, CEO down. Also, breakdancing. Evil, evil, evil wars. Wicked. AI slop was everywhere… and will continue to be. Bluesky exploded, and so did a lot of Lebanese pagers. This is going to ruin the tour. Crowdstrike, Navalny and the Assad regime went down. Chappell Roan stood up. In other standing up news: South Korea, the UK, New Zealand, Brazil. A bridge collapsed, a cathedral reborn. Am I brat? Demure? Mindful? Either way, those lookalikes Not Like Us. Does a popcorn bucket match my freak? Sure looks like it. I never had Meth Lab Oompa Loompa on my bingo card and I want off this ride.

History has always rhymed; in 2024 it did so faster than ever.

But we’ve made it. And though many (but not all) of my predictions for 2024, 2023, 2022, 2021, 2020, 2019 and 2018 have fallen flat, there are two reasons why you should keep reading this article. Firstly… because predictions are fun! Secondly, because the knowledge gained through planning for the future is crucial to the selection of appropriate actions as events unfold. Predictions can act as catalysts to steer the conversation in the right direction. We don’t know the answers, but we can at least ask useful questions.

There is an antidote. The real world is hot again — particularly for younger generations. Meanwhile, Disney is flailing, its crown up for the taking. This is a great opportunity for Netflix to diversify and create experiences, à la ABNB. Why not theme parks? Wouldn’t you like a Squid Game arcade game? A Stranger Things haunted house? A BoJack Horseman bar? It would also be a great way to strengthen the company’s IP. Netflix needs to become Disney before TikTok becomes… everything. This starts in 2025.

There is however a major issue : it is hard for “normies” to use crypto / web3 today. There is no apps, no physical business, no clear way to use meme coins beyond useless “white papers”. We still “see the wires”. 

An app store is a somewhat obvious fix. More specifically, an intuitive platform to let users use the tokens in which they’ve invested, without having to learn the technical lingo in which the industry revels. Apple’s app store changed the world; could something like that happen again? Coinbase is probably the best placed to manage this opportunity… but I have little faith in their operational abilities. Only time will tell.

You’ve felt it for a while, haven’t you? Fundamentals don’t matter anymore. In fact, it feels like reality itself isn’t as important as it once was. Everything is ambiguous, emotional, easily viral, iterative, baked by an Internet reality. Old finance and politics rules don’t work anymore, and may never work again. Trump was a precursor of this. The GameStop “short squeeze saga” was also an early sign of things to come. Neither were anomalies; both are moon-shots in a world where moon-shots are required. Today, the meme of a thing matters more than the thing itself.

“Fartcoin”, a path to generational wealth? Sure! 2024 is the year narrative fully ate reality. In 2025, we will see “rational” actors adapt to this reality. We will see more story-driven strategies, more speed, more adaptability, more reliance on community, more efforts to capture and shapes “vibes”. Professors out, “regards” in : the rule-book is out the window… and we better get used to it.

There is a more likely alternative: president Xi gets on the phone with then-president Trump and coaxes him into letting the company continue to operate (maybe something about golf courses in China?). There’d be some trade-offs of course — my bet is on Larry Ellison owning 20% of the platforms by the end of 2025. I make this predictions safely: TikTok valuation is c.300B$. A third of that value is likely in the US. 100B$ is too much money to leave on the table. Someone, somewhere, will find a way to save US TikTok in 2025.

It doesn’t take a child psychologist to see that phones in school are harming learning. I will let better informed commentators weight in on their overall benefits (?) and drawbacks for young people; but we can all generally agree that their place is not in the classroom. Cheating, distractions, FOMO-induced anxiety… none of it is good.

Predicting phones getting banned from schools is not the bravest call in this list : government-level recommendations on the topic have already been made, and calls for outright bans are becoming louder. In 2025, recommendations will become laws. Mobile-free zones are coming. Tell your kids about it.

When was the last time you saw a real-life friend’s post on Facebook? A Not recently, I bet. Our feeds have turned into ghost towns — relics of a time when “social” meant pokes, relationship updates and family photos. Now, it’s all memes, ads, and sponsored randomness. Good riddance, perhaps? But here’s the catch: with 3.1 billion monthly active users, Meta can’t just let Facebook gather dust. Even if the platform feels deserted, they need users to keep coming back. After all, Meta’s got to keep funding its next big pivot (the metaverse? VR? Dating?)

Facebook will likely soon start auto-populating our feeds with personalized, AI-generated content. It’s a logical next step. TikTok already proved that you don’t need to rely on your friends for engagement. Once the “social” part started sliding, there was never any going back. In 2025, your feed will be an endless stream of tailor-made posts conjured up by Meta’s machines — shiny, catchy, and begging for a “Like.” If your real friends and family no longer post… Meta will find a way to fill the silence.

Anyone who’s worked with a young person over the past 2 years may have noticed that they are quite keen on ChatGPT and other such AI tools. That’s normal! They’re great assets, much like a calculator may have been back in the 70s and 80s. I happily use it daily too — there is no putting the genie back in the bottle, and neither should we want to.

There are however a few troubling stories of under-25s reporting having more and more issues thinking / making decisions on their own. They go straight to the digital helper because, let’s be honest, it’s clearly smarter than we are for everyday topics. That laziness (though natural), is pervasive, and will have second-hand (reduced resilience and adaptability…) or third-hand effect (groupthink and institutional vulnerability…), starting in 2025.

Right now, most AI tools live in somewhat distant data centers, which is fine until you need lightning-fast responses (data has to travel back and forth, which that takes time). That’s where specialized silicon chips come in. These next-generation chips will handle AI tasks — like identifying images or answering prompts — right on the device, cutting down latency / nearing real-time and granting real independence from critical infrastructure.

Scary? Definitely. But the rise of autonomous weapons — “killer robots” — feels inevitable. The future of war is moving toward on-device AI, whether we like it or not.

18 months ago, I wrote about The Weirdest Project happening in the Tech World. In essence, Sam Altman — of OpenAI fame — launched a company aiming to facilitate “proving” one’s humanity online by creating a unique cryptographic code based on one’s irises. Customers would get a pretty useless crypto token in exchange for this biometric data. My premise at the time was that asking people to scan their eyes in exchange of fake internet money is — and I quote — “super weird”.

All that remains true. However, countless companies are currently working on creating AI “agents”, who will take actions / make decisions on behalf of users. In the meantime, we are also seeing AI-generated images, sounds and videos becoming increasingly realistic. It doesn’t take a genius to see the need to quickly develop “proof of humanity” concepts. And yes, they may be similar to what has been proposed by the World Org. Scanning eyes is the wrong way to go about it, but the underlying idea is not as bad as I once thought (only idiots never change their mind, right?). And because I haven’t seen better ideas proposed, and because Sam Altman is a worshipped billionaire, his will likely be the “winning” solution.

This is would be a great tool, but also a stressful one. Who in their right mind would give an untested digital intelligence access to their bank account? Additionally, having those tools whizzing through the internet could do some actual damage to the economy if left unchecked (via bugs or malice). What if they start interacting with each other and get stuck in a never-ending price war?

The easier solution is giving AI agents “pocket money” in the form of crypto token. It’s already digital, traceable, without too many rules, and somewhat remote from the real economy. It’s perfect. In 2025, you will give your AI Bitcoin pocket money, and no one will find that weird.

As AI agents multiply across every corner of our world, it’ll dawn on us that we can’t outdo them by playing fair. They process data at record speed, remember everything, and zip through tasks in a blink. But they’re bound by rules. We’re not.

It’s simpler for a human to spy on competition than decode complex analytics. Seducing a CFO for insider information sure beats trying to match AI’s microsecond decision-making. Feeding competitors’ algorithms trash data is easier than build one’s own algorithms.

And so, every company or person without the ability to fund and use AI agents will cheat. Until lawmakers step in and lawyers will have a field day, at least. It will be a fun couple years. Because in a world dominated by AI, cheating just got cool again.

Over the past couple of years, AI commentators have split into two camps. The first argues that all LLMs do is predict the next word in a sentence based on a prompt. They sneer at simple mistakes, like when ChatGPT can’t count the number of Rs in “strawberry”. It’s a coping mechanism. I would know, I was in that camp until mid-2024. 

The other camp focus on the millions of things such tools can do, and how terrifyingly fast their progression is. They also notice how few guardrails have been put in place to protect every-day citizen from the havoc some tools, left uncheck, could wreck. That camp has long been occupied by the people working on AI tool. I now believe they are right. AI is real and dangerous. In 2025, more people will have a similar shift of conscience, and we’ll likely all be better for it.

Despite being the oldest and most common form of human communication, voice has never really worked as an interface for engaging with technology. Alexa and Siri have always been disappointments, used for playing music, checking the weather… and little else.

Today, however, the evolution of large language models promises dramatically improved voice assistants. Verbal models will soon offer more human-like conversational abilities, making voice interactions more natural and efficient. The digital world hasn’t caught up yet; existing apps aren’t naturally equipped to build / accommodate voice experiences. Neither is our hardware, in fact. In 2025, this will change. We will get new gadgets to speak to (pendants, bracelets, drones…), and better tech to understand and act on what we’re saying.

Have you ever wanted to play a game where NPCs adapted perfectly to your in-game behaviour? To your mood? In 2025, this will become possible as studios leverage AI to reinvent gaming’s nature.

Ready for a world salad? Here’s what the new gaming paradigm will look like. Procedural content generation will ensure limitless environments and challenges, while dynamic storytelling will adapt to player choices to create unique narrative arcs. AI-powered “agents” will create lifelike companions, turbo-charging games with NPCs capable of passing the Touring test. Adaptive difficulty and player behaviour learning will tailor difficulty and content to individual preferences, enhancing engagement. Incorporation of real-world data will continually refresh the game’s relevance. Community-driven content will allow for an ever-expanding universe, and cross-platform experiences will interconnect different games. Ultimately, autonomous AI could independently evolve games, introducing new features and fixes, heralding an era where games continuously grow and change, providing infinite exploration and entertainment possibilities.

Now say that in one breath. Happy gaming.

By 2025, expect a wave of embryo tech start-ups pushing organoids, synthetic embryos, and novel gene-editing methods. With obstacles like manufacturing challenges and ethical debates being addressed, these ventures are poised to unlock next-level regenerative and longevity research, turning once-futuristic ideas into practical realities.

One of the least expected externality of Ozempic will likely be that, pushed by hubris and envy (honestly… same), more people will go to the doctor in 2025 to get a prescription. This will lead to long-overdue check-ups, itself leading to a visible increase in disease diagnoses. Everyone will be confused. RFK will probably blame vaccines. But the reason will be simple : people just want to look good.

I haven’t slept well in about 4 years. I guess the same goes for many people. COVID, wars, politics, climate change… It’s hard to fully give oneself to Morpheus’ embrace when the horsemen of the apocalypse are running circles in one’s head.

The cost of bad sleep to the economy is astronomical ($411B a year in the US alone). If a company manages to offer large companies and countries a way to reduce that number by 20%, while taking a mere 1% cut, they’d be making a bit less than 900 million a year, which would easily warrant a unicorn status.

Many smart entrepreneurs will have done this math. Which is why, in 2025, we’ll see start-ups reach the status of unicorn by promising us a rest that has evaded us for years.

Energy grids the world over are under stress because the like of Meta, Google, AWS and Microsoft are building at such rapid rates. A new paradigm is needed. Hydrogen? Not mature enough. Batteries? Too expensive. Coal? Not green enough. Gaz turbine? Maybe, for the short term. No, the only real and valid solution to power thousands of data centers throughout the world is nuclear energy. Small Modular Reactors to be specific. SMR refers to a new generation of nuclear reactors that are smaller in size and designed for modular construction, allowing for increased scalability and reduced capital costs. These reactors are highly efficient and offer enhanced safety features, making them a promising solution for sustainable and flexible energy generation.

In 2025, we will see the first ones come online after years of discussions and R&D. It will reshape the energy paradigm throughout the world, and likely lead to new innovations that we are yet to imagine. Forget Three Mile Island, Fukushima, Chernobyl, and get excited.

Hydrogen is not matured for full-scale deployment yet, but is nevertheless the topic of many a discussions. It comes in all colours and shapes, but some are more SciFi (and thus interesting) than others. Carbon-negative hydrogen is produced with renewable power and carbon capture, effectively removing more CO₂ than it emits. White hydrogen occurs naturally underground, requiring little intervention. Metallic hydrogen, still largely theoretical, forms under extreme pressure and carries massive energy density. All three variants promise a green alternative for powering AI’s voracious energy demands, aiming to replace coal, gas, and other high-emission sources in data centers around the globe.

In 2025, we will see these hydrogen breakthroughs emerge alongside developments in Small Modular Reactors, upending the conventional energy mix and fuelling AI at scale. Think of it as the next wave of the green revolution.

As of today, two California-based companies command a staggering share of the world’s narrative forces. Meta (Facebook, Instagram, WhatsApp) and Alphabet (Google, YouTube), with their combined 250,000 employees, have the capacity to influence billions of people worldwide. Of course, they’re not alone: Apple, Amazon, Disney, Tencent, ByteDance, Netflix, and other digital heavyweights also wield disproportionate power — often with few effective checks.

This arrangement “worked” while the world was still acclimating to the digital economy. But we are beginning to see policies designed to ensure digital sovereignty, regardless of market-driven logic.

For many industry giants, this shift will be jarring. In the past, enough investment could sway even the most unwilling partners. American behemoths strode the globe, taking what they wanted. They may be in for a rude awakening as countries close off, accept negative short-term consequences, and pursue more favourable long-term political positioning.

If you’ve made it this far in the article, you’ll understand how important AI is likely to be in future economies. Not only in terms of sovereignty… but also when it comes to energy reliability, the latter being directly linked to critical infrastructure build-out. 

European countries are particularly impacted: Frankfurt, London, Amsterdam, Paris and Dublin all have many, many data centers… and significant energy issues because of it (Dublin had to impose a moratorium on new build in 2024). And so, political posturing will prevail, and foreign data center players will get taxed. This will also act as retaliation against the planned Trump tariff, have a positive impact on power supply, and allow national players to potentially emerge. It’s not the choice the markets would make… yet here we are. 

The two predictions above will have a fascinating second-hand effect. Regional powers not bound to the US or China will try to boost their digital economy by hosting AI tools / infrastructure that wuld have otherwise gone to Europe. But, as mentioned, these assets are of national importance, and can’t be shared without some form of protection.

And so, in 2025, countries in South-East Asia and the Middle-East will allow countries from all over the world to build in locations where power and space is abundant. The way they will lure national clients is by promising that data centers built will be regulated under the laws of the country operating it, rather than those of the countries where the building is physically located. This will be hugely beneficial for “middle powers” trying to get a toe-hold in the industry… but may very well end up being a form of neo-colonialism. Only time will tell. 

The AI boom we find ourselves in is really about selling one thing: automate work we don’t value enough to hire a human being to do. This will have a drastic effect, as parts of society automates and gains massive rewards, while the other part sheds jobs at massive speed. 

In 2025, the middle class will continue to disappear, being replaced by “with AI” and “without AI”. We are already seeing it: the top 5% of public companies are running away with the market. The 400 richest Americans holding wealth equivalent to about 17% of GDP, a 15% (!) increase from the 2% level in 1982. And these trends are only accelerating.

Every day, god’s light, we stray, etc.

Yet by 2025, government interventions may disrupt this vision. In pursuit of national interests — commercial, military, or strategic — countries could impose export controls or regulations that stifle open-source development. Instead of letting free market competition foster innovation, these policies may create a fragmented landscape of proprietary “national champions,” undermining the benefits of open collaboration that LeCun and others promote.

For the past 10 years, the tech world has not really delivered on its many promises. Just more monitoring, more nudging, more draining of our data, our time, our joy.

In 2025, let us hope that the industry goes back to doing hard, but necessary work of improving lives. That we invent new tools to help reduce infant mortality, deaths from infectious disease and wealth inequalities between the poorest and wealthiest nations. That great medical advances are made, that education is further democratized, and that green technologies emerges as a long-term force for good. Fingers crossed.