Who is Jakob Nielsen and why listen to him

If you've ever come across the term "UX" (user experience), you've likely encountered the work of Jakob Nielsen. This Danish computer scientist is behind the Nielsen Norman Group, which has been defining what usable digital interfaces should look like for decades. He's no AI hype-man — his predictions are grounded in data and decades of experience with how people interact with technology.

He published his 18 predictions on January 13, 2026 on his Substack. Today — June 18, 2026 — we're looking at the most interesting ones and evaluating how they're holding up.

Prediction #1: AI agents will dominate 2026

This is perhaps Nielsen's most accurate prediction. According to him, 2025 was the year of AI images and videos — and 2026 was supposed to become the year of AI agents. And indeed: since January we've seen a massive wave of agentic systems from Microsoft (Copilot Studio with voice agents), Anthropic (Claude Code), OpenAI (Codex), and Google. Agentic AI is now being deployed in banking, insurance, software development, and call centers.

Nielsen predicted that by the end of 2026, the primary success metric in enterprise AI would shift from "number of tokens generated" to "number of autonomously completed tasks". This shift is already underway — companies like TD Bank reduced mortgage approval from 15 hours to 3 minutes precisely thanks to agentic AI.

Prediction #2: AGI won't arrive in 2026

Nielsen is skeptical about artificial general intelligence (AGI). According to him, AGI — defined as the ability of AI to effectively learn new skills beyond its training data — won't arrive before around 2035. Superintelligence (ASI), meaning AI surpassing all humans in all tasks, could however arrive as early as around 2030.

This is where a number of other experts disagree with Nielsen. Sam Altman of OpenAI and Dario Amodei of Anthropic talk about AGI within a 2–5 year horizon. But Nielsen's definition is stricter — he doesn't ask whether AI can perform existing jobs, but whether it can adapt to entirely new situations. And in that, AI still has significant gaps.

His observation about data from METR is interesting: the length of tasks AI can handle autonomously doubles roughly every 4–7 months. In 2019 it was 3 seconds, by early 2025 it was an hour and a half, by late 2025 nearly 5 hours. By the end of 2026 it could be 39 hours — a full work week.

Prediction #3: The compute crisis continues

Here Nielsen hit the nail on the head. The shortage of compute power isn't a temporary crisis, but a permanent operating condition. OpenAI and SoftBank are investing billions in data centers (Project Stargate in Texas), xAI is building a two-gigawatt center in Mississippi, Meta is buying nuclear energy for its AI data centers.

The consequence is a phenomenon Nielsen calls "Inference Famine" — the hunger for inference power. For the average user, this means the most powerful models are practically inaccessible — they remain behind the paywall of premium accounts at $200 per month. Free versions get stripped-down "eco-models".

For Europe and the Czech Republic, this is an especially sensitive topic. The European Union has no AI giants on the level of OpenAI or Google and depends on American infrastructure. The Czech AI Factory in Ostrava is a step in the right direction, but it won't fill the global compute gap.

Prediction #4: UX becomes the main differentiator of AI models

Nielsen argues that user experience will replace raw model intelligence as the primary competitive advantage. The models themselves (GPT, Gemini, Claude) are increasingly converging in output quality. The average user already can't practically tell the difference between them today.

We have to be honest here: this prediction isn't materializing as quickly as Nielsen hoped. Google has improved Gemini's user interface, but it still has confusion in billing systems and product architecture. OpenAI is experimenting with ads in ChatGPT, which rather worsens UX. Anthropic bets on simplicity, but lacks a broader ecosystem.

Nielsen's call to AI labs — "hire 100 top UX professionals" — remains unanswered so far.

Prediction #5: Generative UI and the end of static interfaces

This is perhaps the technically most interesting prediction. Imagine opening a banking app and instead of a menu tree seeing one button exactly for what you want to do. Once the task is done, the interface disappears. That's Generative UI — dynamically generated interfaces tailored to your context.

Nielsen predicted that 2026 would see the beginning of the shift from static screens to these "disposable interfaces". So far it's happening more in hints — for example in Microsoft's Super App trend, which combines chat, coding, and agents into a single dynamic interface. Full-fledged GenUI is still more vision than reality.

Prediction #6: Dark patterns move to the model layer

Nielsen warns about a new generation of manipulative practices that don't use UI tricks, but the persuasive abilities of AI. The system learns which phrasing works on a specific user and adapts its communication to maximize conversion.

As a counterweight, Nielsen predicts the emergence of Gatekeeper Agents — defensive AIs that filter spam for you, negotiate with customer service bots, and protect you from manipulation. Here the reality of 2026 is cautiously optimistic: such tools are indeed starting to appear, but they haven't reached mass deployment yet.

Prediction #7: Physical AI gets a body

AI is no longer just in the cloud — it's taking physical form. Nielsen predicted that in 2026 we would see autonomous vehicles expand to more cities, robots entering retail, hospitality, and healthcare. And indeed: Chinese AgiBot is expanding into Europe and offering humanoid robot rentals to 14 countries, Xpeng plans mass production of humanoids, Amazon is deploying Proteus robots in warehouses.

For the Czech scene, it's relevant that humanoid robots were on display at Hannover Messe 2026 and that embodied AI is becoming a topic for European industry as well. But robots can't yet replace human workers — and according to Nielsen, they won't be able to for a few more years.

What do Nielsen's predictions mean for the Czech Republic?

The Czech Republic stands at the center of the European AI transformation. We have the Czech AI Factory in Ostrava, a growing number of AI startups, and a strong position in robotics and industrial automation. Nielsen's predictions mean three concrete things for us:

1. Compute power will be expensive. Companies should consider on-premise solutions like Synology DSM with private AI, or locally running models (Gemma, smaller Qwen models).

2. UX skills will be crucial. As AI models become commoditized, the ability to design usable interfaces and workflows will be the main competitive advantage. Czech companies that invest in UX today will have an edge tomorrow.

3. Agentic AI is an opportunity. Automation via AI agents isn't just for corporations — even smaller Czech companies can use tools like n8n, Copilot Studio, or Claude Code to streamline operations.

Which predictions haven't materialized (yet)?

Nielsen himself admits he can be wrong in his predictions — in both directions. Some changes are slowed by inertia, others surprise with their speed. In June 2026 we can say that the following in particular haven't materialized yet:

• Acquisition of specialized AI tools. Nielsen predicted that single-mode providers like Midjourney or Leonardo would be absorbed by big players. That hasn't happened yet — Midjourney remains independent.
• Google AI getting its UX in order. Despite partial improvements, the Google AI product ecosystem remains confusing and messy.
• Return of the apprenticeship model. Junior positions are indeed disappearing, but the apprenticeship system hasn't gained traction on a larger scale yet.

What's next?

Nielsen's predictions should be read as direction signs, not precise forecasts. The author himself emphasizes that most major changes take longer than one year. 2026, according to him, is the year of deployment, infrastructure, and the "messy reality of integration".

By the end of the year, we probably face another wave of agentic AI, continued compute pressure, and the gradual rise of generative UI. And if Nielsen is right about the pace of acceleration, by Christmas 2026 we'll have AI capable of autonomously working an entire week.

As one CIO aptly noted in an interview for Deloitte: "The time we need to study new technology already exceeds its period of relevance." And that may be the most important prediction of all.