2026 Predictions: The Year AI Goes Local and Reality Catches Up
Another year coming in hot and another round of wild (and not so wild) predictions. With the total chaos of 2025 I’ll take pride in getting roughly 60% of my 2025 predictions right. When trying to make bold claims, this is a WR% I would take, balancing between good enough to keep going and honest enough to have whiffed on a few “out of the box” calls.
The ones that landed felt obvious in hindsight. AI layoffs accelerated exactly as expected. Meme coins made their comeback (for a hot sec). AGI discourse went from “maybe in five years” to “wait, did that just happen?” The ones I missed? GTA 6 blockchain integration was wishful thinking (more so on simply shipping on time), and gaming DAOs didn’t fail so much as fade into irrelevance, which might be worse.
But here’s the thing about predictions. The value isn’t in a perfect scorecard. It’s in forcing yourself to take positions, document your reasoning, and then face reality twelve months later. That’s the fun of the game.
This year I’m structuring things differently. After a year of studying and trading on Polymarket and running 150+ AI agents daily through Alexandria, my lens has sharpened. I now see a cleaner split between signal and noise than I did last January.
So I’m breaking this into two buckets:
- Not This Year - Things you will hear about constantly in 2026 that won’t meaningfully land but are inevitably part of the zeitgeist
- Actually Arriving - The shifts that are real, tangible, and hitting this calendar year
Here’s what I’ll cover:
- Not This Year (The Hype That Won’t Land)
- AI and Technology
- Crypto and Finance
- Gaming
- Health and Human
- Culture (yes, including cheese)
Let’s get into it.
Not This Year: The Hype That Won’t Land in 2026
This is the section I wish more prediction pieces included. Half of being right about the future is knowing what isn’t happening just yet. These are topics that will dominate headlines, flood your timeline, and generate enormous amounts of commentary… Only to see none of them meaningfully resolve this year.
Space Infrastructure
You’re going to hear a lot about server farms in space and orbital data centers, especially given the SpaceX IPO. I get the excitement. The convergence of compute demand, cooling challenges, and launch cost reductions makes the narrative compelling.
But the engineering timeline doesn’t care about your narrative. Putting meaningful compute infrastructure in orbit is a few years away. The FAA challenges alone make it a question mark that is pseudo-helped by SpaceX already deploying its Starlink systems. But people have forgotten how long the FCC approval process was to get that rolling, and I can’t imagine the US government operating much faster given it current state.
Add a layer of lawfaring competition and I just don’t see it happening in 2026. If this heats up this year, Jeff Bezos will fall back to his usual tactics of sue to slow down.
Will we see prototypes? Sure. Will we see announcements? Absolutely. Will any of it matter for your portfolio or your product roadmap in 2026? No.
ASI Discourse Reaches Fever Pitch
Notice I said ASI, not AGI. That’s the shift that really matters.
I’ve been writing about AGI indifference for months now, and the core thesis holds: we’ve entered a phase where the AGI conversation has been absorbed and normalized. The discourse has already moved past it, with many claiming we need new ways to measure AI intelligence.
The new goalpost is Artificial Superintelligence, and it’s going to consume an absurd amount of oxygen in late 2026.
Every major lab will have something to say about it and likely cause an escalation to the doomer thesis. Meanwhile, the accelerationists will push harder, causing deeper widening of the new “K-shaped” economy and workforce. The Senate will hold hearings.
None of it will produce ASI.
The models will get exponentially better this year. No doubt we’ll see new lineage models pushing terrifying updates. But the gap between “really impressive AI” and “intelligence that exceeds all human capability across all domains” is still enormous.
With that said, the discourse will outpace the reality, but let’s understand that there is a crack in the door… I think we’re a handful of model upgrades away from recursive learning and building. I have my agentic setup doing this, but add in scale and its a different story.
The exponentiality and tightening iteration cycles will go from months, to weeks, to days. Just not to the point of ASI in 2026.
If you’re building products, ignore the ASI timeline and focus on what the models can actually do today.
The Energy Deficit Hits Beattle Mania Levels
This one is real but won’t resolve in 2026. The energy demands of AI training and inference are growing faster than grid capacity in most markets.
Data center construction is outpacing power infrastructure, and nuclear is getting renewed attention, but moves on decade timescales. Even Small Modular Reactors (SMRs) aren’t proven enough for production and fusion has been “almost there” for a decade.
Every country will be talking about these impacts. For longer-term followers, you know my thoughts on reducing everything to energy costs… well, it’s about to get very real.
In the US, wind and solar have been growing steadily over the last two decades and now represent ~20% of energy generated. A great trajectory, but still faces persisting hurdles like grid/infrastructure, economic/supply chain, and regulatory/social issues.
You’ll see this pop up everywhere: earnings calls, policy debates, ESG reports, conference panels. Everyone will talk about the problem. No one will solve it this year. The actual buildout of energy infrastructure sufficient for AI’s appetite is a late 20’s story.
What you’ll see is creative workarounds: edge computing, more efficient model architectures, and inference optimization. The constraint will drive innovation at the margins, but the macro energy gap is not closing in 2026.
Robotics Goes Mainstream
Boston Dynamics videos will continue to go viral, Tesla will show more Optimus demos, and China will run more Olympic-like events to show off its capabilities lead. Figure and other humanoid startups will raise enormous rounds and the vibes will suggest robots are about to be everywhere…
They’re not, or at least, not this year.
The hardware is further along than most people realize, but the software stack for general-purpose robotics is still a generation behind. We’ll see impressive demos and limited commercial deployments in warehouses and manufacturing.
Consumer and general-purpose humanoid robots in your daily life are for 2028 at the absolute earliest. Also, why do you need a $30k robot to slowly fold your laundry?!
Actually Arriving: The Real Shifts of 2026
Now for the stuff I’m actually putting positions on. These are predictions where I’m willing to be scored and graded.
AI and Technology
1. AI Goes Local: Desktop Agents Become THE Game Changer
This is my highest conviction prediction for 2026 and the one I’d stake my reputation on.
The browser-based AI paradigm is about to get obliterated by local desktop agents that run directly on your operating system. No browser sandbox, no security permission battles, and no “I can’t access that file” limitations. Straight direct system access that empowers your agent to help you run your tasks.
I’ve been building agentic systems for over a year now. Alexandria runs 150+ agents organized into functional teams. I know exactly where the bottleneck sits, and it’s in the interface layer between the AI and your actual workflow.
Right now, even the best AI agents are fighting the browser. They can’t see your file system natively. They can’t interact with desktop applications without hacky workarounds. They can’t manage system processes or orchestrate between local tools without middleware.
Desktop agents solve all of this in one move.
Think about what that means. An AI agent that can open your IDE, write code, run tests, check results, and iterate, all without you copying and pasting between browser tabs. An agent that manages your email client, your calendar, your project management tool, and your Slack as a unified workflow instead of twelve disconnected plugins.
An actual assistant, just better.
Google, Apple, and Microsoft are both positioning for this. Apple Intelligence is early (and objectively crappy) but directionally correct. Microsoft’s Copilot integration across the OS is more aggressive. And Google likely has the most real-world data to feed its models.
But the real winners will be the third-party agent frameworks that plug directly into the OS layer.
This is the year the “AI assistant” stops being a chat window you visit and becomes an ambient system that runs alongside everything you do. For builders already at Level 4 or 5 on the agentic AI maturity model, this unlocks an entirely new tier of capability.
2. Agentic Teams Become Ubiquitous
Related to the above but distinct enough to call out separately.
In 2025, building agentic teams was a niche activity for people like me who were willing to spend months wiring together custom systems. I documented the entire painful journey from single prompts to orchestrated teams because almost nobody else was doing it publicly.
In 2026, this goes mainstream.
The tooling has caught up. Cursor, Claude Code, Devin, and a dozen other platforms have made multi-agent workflows accessible without needing to build your own orchestration layer from scratch. Companies that were skeptical in 2025 are seeing the 3x-5x productivity gains from early adopters and scrambling to catch up.
I’ll make a specific sub-prediction: by Q4 2026, at least 50% of software companies with 50+ employees will have some form of agentic AI team deployed in production workflows. Not experimental, but working in production.
The transition from “AI as tool” to “AI as team member” is the defining technology shift of this year. If you’re still at Level 1 or 2 on the maturity model by December, you’re in BIG trouble.
3. Cognitive Costs of AI Usage Surface
Here’s the prediction nobody wants to hear.
We’re going to start seeing real research and real discourse around the cognitive downsides of heavy AI usage. Context switching, reduced deep thinking, attention fragmentation, and a subtle erosion of the skills we’re offloading.
I notice it in myself. When I have agents handling research, drafting, coding, and analysis, my own ability to sit with a hard problem for two uninterrupted hours has gotten worse. Not dramatically, but noticeably, and I’m someone who’s hyper-aware of this dynamic.
I build agentic systems for a living, and even I catch myself reaching for Claude before I’ve finished forming my own thought. This should concern people… A LOT.
The analogy that works best is GPS. Navigation apps made us functionally worse at spatial reasoning. We traded a capability for convenience and most of us decided the trade was worth it.
The same thing is about to happen with AI and cognitive work, except the capabilities being traded are higher-order: synthesis, original analysis, sustained attention.
Here’s the part that doesn’t get talked about enough: the average human processes roughly 100k-200k tokens of information per day. That’s reading, conversations, media, daydreaming, everything.
Working alongside agents, you’re ingesting multiples of that in a single work session. The bandwidth mismatch is real. Our agents are operating at a speed our biology wasn’t built for, and the result is this low-grade cognitive fatigue that’s hard to name but easy to feel.
I notice it most when I’m around other AI power users. We easily identify that our speed of input/output in our conversations feels too slow compared to what we’re used to with agents. We’re recalibrating what “normal” processing speed feels like, and the human version is starting to feel broken by comparison.
This won’t slow AI adoption. But it will reshape how the smartest companies think about AI integration. The ones paying attention will design workflows that keep humans engaged in the hard parts, the judgment calls, the creative leaps, the ethical reasoning, rather than automating everything and wondering six months later why their team can’t think independently anymore.
The ones not paying attention will optimize for speed, burn out their people, and mistake output volume for actual productivity.
4. Anti-AI = Proof of Human (Authenticity Counter-Culture)
A counter-movement is building, and it’s going to crystallize in 2026.
“Proof of Human” will emerge as a real category. Think verified human-created content, authentication protocols for original work, and premium pricing for demonstrably non-AI output. This isn’t anti-technology, it’s a market response to saturation of low-grade AI slop.
When AI-generated content becomes the default, human-created content becomes scarce. Scarcity creates value and the market will respond.
You’ll see this in creative industries first. Music, art, writing, photography will see a hard push for “100% human-made.” I imagine we’ll see a premium label, similar to organic food or fair trade coffee. Verification systems will emerge, probably blockchain-based, to authenticate the provenance of creative work.
This prediction connects to a broader theme I keep coming back to: the question of what constitutes valuable human action when AI can replicate most outputs?
The answer, at least partially, is that the process of human creation holds value independent of the output. People will pay more for a painting by a human, not because it looks better, but because a human made it… and especially a specific human made it.
Ultimately, this becomes a philosophical shift with real economic consequences.
Crypto and Finance
5. Prediction Markets Fully Integrate with Daily Life
Damn, this feels like a softball, but I’ve been waiting for this one.
Prediction markets have been an interest of mine for the past few years. I trade on Polymarket regularly, and I’ve watched the platform go from crypto-native niche to genuine information infrastructure during the 2024 election cycle.
In 2026, this integration goes much further. Prediction markets will be embedded directly into news platforms, financial terminals, and social media feeds. Not as a separate app you visit, but as a contextual layer on top of existing content.
Reading an article about a central bank decision? There’s a prediction market right there showing real-time probabilities. Following a geopolitical crisis turns these real-money forecasts into incredible insightful data aggregators that assist alongside the reporting thats closer to the “truth”.
The infrastructure for this is already being built. Polymarket’s API is being integrated by third-party developers. Kalshi is pushing hard on regulatory approval for broader event contracts. The concept of “putting your money where your mouth is” as a default part of information consumption is arriving.
I’ll add a specific call that should help the industry: at least one major news outlet (top 10 by traffic) will natively integrate prediction market data into their editorial content by December 2026.
This matters because prediction markets are the most honest information aggregation mechanism we have. When people have real money at stake, the noise drops out fast and information asymmetry with it.
I will take a sec to point out that there are many negative effects of these markets, too. Gambling in a new, almost unlimited form will undoubtedly hurt society (as we’ve already seen with apps like DraftKings).
I’ve watched people in my own trading circles go from “informed speculation” to degenerate behavior disturbingly fast once the markets got liquid enough.
But to turn this into a parlay, I’ll add that we’ll have a shockingly large arrest from someone wetting their beak from inside information on a market. The transparency cuts both ways and although counter-inuative, I think it helps the industry.
It’s a year of greater transparency, and people will learn the hard way that it can be a good and a bad thing.
6. X402 Takes Hold of Transaction Volume
This is a slightly nerdier call but one I feel good about.
The X402 protocol, HTTP-native payments built directly into internet requests, is positioned to capture a meaningful slice of internet transaction volume in 2026.
The concept is simple: payments happen at the protocol level, and not the application level. Essentially, your browser handles micro-transactions the same way it handles cookies.
Why does this matter? Because it solves the fundamental monetization problem of the internet. Every paywall, every subscription, every ad-supported model is a workaround for the fact that HTTP doesn’t natively support payments. X402 fixes this at the infrastructure layer.
The early use cases will be API access (pay-per-call instead of monthly subscriptions with specific rate limits), content gating (pay $0.02 to read this article instead of $15/month for the whole publication), and AI agent transactions (agents paying for services on your behalf).
That last one is the dark horse and the one I’m most excited about. When your desktop AI agent can autonomously pay for API access, data sources, and services using X402, the entire economics of agentic workflows change.
Agents stop being software tools and become economic actors that don’t need handholding and to constantly approve activities. I’m already running into this bottleneck with Alexandria where agents need access to paid APIs, and the authentication/payment layer is always the friction point.
I don’t think X402 dominates internet payments in 2026. But I think it goes from “interesting experiment” to “meaningful transaction volume” and sets the stage for a much bigger story in 2027-2028.
7. Omni Banks Rise, Traditional Finance Bridges to Blockchain
The line between traditional banking and crypto is dissolving faster than most realize. I’ve been saying this for years, and I still think people underestimate the speed.
Omni banks are financial institutions that natively operate across both fiat and blockchain rails, and will see rapid growth in 2026.
These aren’t crypto-native companies bolting on banking services or banks nervously dipping into crypto. They’re purpose-built for both worlds simultaneously, crossing the threshold into borderless banking.
At the same time, the incumbents are moving with constant news of Visa, Mastercard, and major banks building blockchain-connected payment infrastructure. The USDC integrations, the stablecoin settlement, and the tokenized deposit programs have moved from pilots to production systems processing real volume.
The specific call: total stablecoin real transaction volume will exceed $20 trillion in 2026. For context, 2024 cleared roughly $10 trillion. That’s not a moonshot projection, it’s the trajectory we’re already on as traditional financial infrastructure plugs into blockchain rails.
This is the boring version of the crypto revolution. Moon missions with 100x returns are going to be few and far between. Just the plumbing of global finance quietly switching over to better infrastructure. It’s less exciting than meme coins but orders of magnitude more important.
Gaming
8. Arc Raiders’ Real Test: Can Embark Execute LiveOps at Scale?
This was my game of the year for 2025 and I honestly hadn’t had this type of stay up until 4am fun in a long time.
Also, I’m finishing a deeper analysis of Arc Raiders and the prisoner’s dilemma, so I won’t rehash the full framework here. But I want to make a specific prediction about what happens next, because the honeymoon phase is going to end.
Arc Raiders proved something the gaming industry has rarely demonstrated: that deliberately designing around game theory principles produces measurably player behavior. But is it a great gimmick, or has the goods to be sticky for years?
I’ve logged 200+ hours and watched the community evolve in real-time. The cooperation dynamics, the betrayal patterns at extraction, the emotional rollercoaster of every run. We can all agree Embark knocked it out of the park.
But here’s where it gets interesting. Every major studio is watching the updated PvPvE mechanics and it’s about to get saturated the same way battle royale did after PUBG and Fortnite.
The clones are coming, and when they arrive, the only thing that will separate Arc Raiders from the pack is LiveOps execution.
This is where I think Embark is in for some short-term pain. They’re building the plane mid-air. The game’s foundation is brilliant, but running a live service at this scale is a completely different muscle than shipping a great gameplay loop.
Day 1 retention across the industry sits at just 26%, dropping below 4% by Day 30. LiveOps events across top F2P games have increased 35% since 2023, and players are burning out.
The studios that figure out quality over quantity in their event cadence will win. The ones that dump content for the sake of dumping content will churn their player base (Warzone, anyone).
My prediction: Arc Raiders remains a top-5 extraction shooter through 2026 but faces a meaningful retention dip in the early parts of this year. From my understanding they just weren’t prepared for the demands of the breadth and depth of high level LiveOps.
The real question is whether Embark has the operational depth to navigate that, and whether they can evolve the game fast enough to stay ahead of the copycats.
9. The Great Studio Compression: AI Lets Small Teams Eat AAA’s Lunch
As someone with a lot of experience working at top publishers, this one has been building for years, and 2026 is when it breaks open.
One-third of US game industry companies have had laid off in the past two years. Ubisoft, Microsoft, EA, they’ve all been cutting. And the layoffs aren’t slowing down, but accelerating. Because the bloated companies and studios are finally realizing what smaller outfits have known for a while: you can do more with less, especially now.
Here’s the stat that tells the whole story: 75% of the top 20 highest-rated games on Metacritic in 2025 were made by indie studios. Nearly double their share from 2016.
Meanwhile, AAA budgets often cross $100’s million, and the mid-tier ($5M-$20M) is nearly extinct.
AI is the accelerant once game engines, like Unity, release their proprietary models. Once AI can prompt full casual games into existence with natural language and generative art, the gig is up.
By Unity’s own account, 62% of the engines developers are already using AI for coding assistance. The gap between what a 10-person team can produce and what a 500-person team can produce is shrinking every quarter.
I predict layoff rounds in 2026 outpace totals we saw in 2024, the peak year in the industry. Sadly, this would mean over 15k jobs.
Simultaneously, indie and small-studio titles will continue to dominate critical reception and increasingly compete on commercial metrics. The $300 million AAA model isn’t sustainable, and everyone inside the industry knows it.
The question is whether the executives will adapt or keep throwing money at diminishing returns until the market forces their hand.
10. GTA 6 Ships with Blockchain Integration
Yes, I’m pulling this one out of the bin and dusting it off. I missed it last year because the game didn’t ship. But I think the delays themselves are part of the story.
My theory: Rockstar has been quietly building blockchain infrastructure into GTA 6’s economy. The in-game economy of GTA Online already functions like a proto-crypto ecosystem, complete with inflation, arbitrage opportunities, and a thriving gray market.
I think at least part of it is integration complexity. You don’t bolt blockchain rails onto a game this size in the last six months of development. You build it in from the foundation.
If any studio has the install base and the audacity to normalize blockchain rails in a AAA title, it’s Rockstar. A GTA 6 with native crypto integration would be industry-altering, not because the tech is new, but because 100+ million players would suddenly be interacting with blockchain without knowing or caring that it’s blockchain.
I’ll be upfront: this is a low-conviction call, but the upside if it hits is enormous.
Health and Human
11. Health and Longevity Breakthroughs Accelerate
I’ve been tracking biotech from the sidelines for years, but 2026 is when several convergent threads hit critical mass. I’ll be honest, this is the section where I’m least deep in the weeds, but the signal is too strong to ignore.
AI-driven drug discovery is about to produce its first wave of candidates that were genuinely impossible without machine learning. Not “faster” drug discovery, but genuinely novel discovery.
Compounds and pathways that human researchers wouldn’t have identified without AI modeling. The timeline from target identification to clinical trials has compressed from years to months in some cases, and that acceleration is only getting faster.
On the consumer side, the biohacking movement has graduated from fringe to mainstream. Blood panel analysis, continuous glucose monitors, personalized supplement stacks driven by genetic data. This was expensive and weird three years ago. Now it’s a subscription service with a slick app (although many could use some help on the app side).
I started tracking my own biomarkers last year, and the amount of actionable data you get from a $200 panel is shocking and deeply thought provoking. Everyone I’ve talked to who starts doing this becomes slightly obsessed.
My specific calls:
- At least one AI-discovered drug enters Phase 3 clinical trials in 2026
- The “longevity economy” (supplements, testing, optimization services) exceeds $50 billion in market size globally
- Wearable health tech adoption crosses 40% of adults in all Tier 1 economies
The bigger story here is cultural. Health optimization is becoming a status marker the way fitness was in the 80s. Knowing your biomarkers is the new knowing your macros and that cultural shift drives commercial adoption faster than any technological improvement.
Culture
12. The Great Cheese Prediction
OK, stick with me here because this one is actually a metaphor for how value perception works in consumer markets.
In the early 20th century, the cheese industry had a problem with rinds. You know the hard crust that forms during the aging process and renders a portion of the cheese inedible. Not to mention you need workers cleaning the rind as it dries.
Lots of resources were poured into this problem and finally, in 1916, J.L. Kraft figured out rindless cheddar. He had a new premium product with consistent texture, no waste, and easier to portion. Supermarkets stocked it front and center and rinded cheddar was the budget option, with the rind seen as a defect, not a feature.
Fast forward a century later and “artisanal” cheddar with the rind is the premium product. You’re paying a 40% markup for what your grandparents considered the default. The rind isn’t a defect anymore, but “character,” more “traditional,” or an ode to the “craft.”
Nothing about the cheese changed… But everything about the story changed.
This is the same dynamic playing out in every market I watch. In AI, the “rind” is human involvement. In content, the “rind” is imperfection and authenticity. In gaming, the “rind” is friction and challenge. In finance, the “rind” is opacity and exclusivity.
The premium always migrates to whatever feels scarce. When efficiency is abundant, inefficiency becomes a premium. When AI output is everywhere, human output becomes artisanal.
Yes, using cheese as an example is a little goofy, but the principle underneath it is one of the most important market dynamics you won’t unsee.
Closing Synthesis: The Year Reality Catches Up
If I had to compress 2026 into a single theme, it’s this: reality catching up to the narrative.
For the past three years, the narratives around AI, crypto, and technology have outpaced what’s actually deployable. We’ve been living in a hype gap where the stories were bigger than the products. That gap is closing in 2026.
AI agents aren’t a demo anymore, they’re THE infrastructure.
Prediction markets aren’t a novelty, they’re THE information systems.
Blockchain payments aren’t an experiment, they’re THE plumbing.
The hype cycle items (space, ASI, robotics, energy) are real but premature. They’ll have their year, I don’t think it’s this one.
The arriving items (local AI, agentic teams, prediction markets, proof of human) are past the hype phase and entering the deployment phase. That’s where the real value creation happens. Not when everyone’s talking about something, but when everyone’s using it.
My positioning for 2026 is straightforward: go deep on the things that are actually arriving and ignore the noise of things that aren’t.
Same as last year’s closer: the only way out is through. But this year, the “through” is clearer, and maybe a little more frightening, than it’s ever been.
See you in January 2027 with the scorecard.
Disagreements? Predictions I missed? Better cheese takes? Find me on Twitter/X or LinkedIn. I’m genuinely interested in what you’re seeing that I’m not.