My 2026 Predictions (Just Kidding)

It’s predictions season. Everyone’s publishing their hot takes on what 2026 will bring.

AI trends to watch! Market shifts ahead! Technologies that will dominate!

So here are mine:

I have absolutely no idea what’s happening in 2026.

Neither do you.

The Week That Predictions Died

Last week alone:

• December 1: DeepSeek released V3.2, matching GPT-5 on major benchmarks
• December 1: NVIDIA launched Alpamayo-R1, the first open reasoning model for autonomous driving
• December 2: Mistral unveiled Mistral 3, claiming “world’s best open-weight multimodal model”
• December 3: OpenAI reportedly leaked performance claims about “Garlic,” their unreleased model

That’s five major AI model releases and one competitive leak in seven days.

If you published annual predictions on November 30th, they were obsolete by December 7th.

This is the environment we’re trying to plan in. Where the competitive landscape shifts weekly. Where capabilities that seemed impossible in October are standard features in December. Where the model you built your product strategy around gets leapfrogged twice before quarter-end.

The Planning Paralysis

Most businesses still operate on annual planning cycles. Build the strategy in Q4. Approve budgets in January. Execute throughout the year.

It worked when things changed slowly. When competitor moves took quarters to roll out. When technology roadmaps were measured in years.

It doesn’t work when Chinese startups can match Silicon Valley frontier models overnight. When Google, OpenAI, and Anthropic are in a launch arms race measured in days. When your carefully researched technology selection becomes outdated while you’re still in procurement.

Your annual strategy assumes a stable world. The world is changing weekly. This is the actual problem.

Businesses aren’t failing because they made bad predictions. They’re failing because they’re treating predictions as commitments.

The Strategy Trap

Here’s what happens in most organisations:

Spend October-December building next year’s plan. Analyse market trends. Consult experts. Run scenarios. Build financial models. Get board approval.

January arrives. Strategy locked. Budgets allocated. Teams mobilised.

February brings three new competitive moves nobody forecasted. March sees capability breakthroughs that change what’s possible. April reveals customer priorities have shifted.

But you’re locked into the plan. Budget’s allocated. Teams are executing. Nobody wants to admit the strategy built three months ago is already wrong.

So you push through. Execute harder. Hope reality catches up to the plan.

It doesn’t.

The Companies That Are Winning

Look at the organisations thriving right now. They’re not the ones with the best predictions. They’re the ones with the best adaptation mechanisms.

They treat strategy as direction, not destination. They build options, not commitments. They run experiments, not rollouts. They measure learning velocity, not plan adherence.

When DeepSeek drops a frontier model that costs 90% less to run, they don’t spend six weeks in steering committee meetings. They test it on Monday, validate performance on Wednesday, and make a decision by Friday.

When customer priorities shift, they don’t debate whether it fits this year’s strategy. They ask whether it’s real and whether they can capture it.

Speed of adaptation beats accuracy of prediction.

What This Means For You

If you’re building 2026 plans right now, you’re not planning for 2026. You’re planning for the version of 2026 that exists in December 2025.

By February, that version won’t exist anymore.

The question isn’t “what should we predict?” The question is “how do we build a business that doesn’t need perfect predictions?”

The Hack

Stop writing strategies. Start building adaptive systems.

Replace annual plans with quarterly hypotheses. Don’t lock in twelve months of execution based on October’s assumptions. Commit to three months of testing based on today’s reality.

Build decision velocity, not planning rigour. The competitive advantage isn’t having better forecasts. It’s making better decisions faster when forecasts prove wrong.

Measure learning, not adherence. Stop tracking “percentage of plan delivered” and start tracking “time from insight to action.” The organisations winning aren’t the ones executing their original plan flawlessly. They’re the ones abandoning bad plans quickly. PACE OVER PERFECTION KLAXON.

This isn’t chaos. It’s strategy for a world that changes faster than planning cycles.

When five frontier AI models launch in a single week, annual predictions aren’t ambitious. They’re absurd.

Build for adaptation instead. That’s the only prediction worth making.

The Problem Hacker* Smart takes on strategy, growth, and insight. From Mark Jefford, ex-Guardian/GWI/YouGov @ // jefford*