> ## Content Index
> Fetch the complete content index at: https://www.thedigitalspeaker.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# I Scored My Own 2026 Predictions: 63.5 out of 100
- URL: https://www.thedigitalspeaker.com/2026-trend-predictions-evaluation/
- Published: 2026-11-20T16:15:57.000Z
- Updated: 2026-11-20T16:16:47.000Z
- Description: A year ago I published 10 technology trends for 2026. Before the 2027 report launches, I audited every one against the evidence. Small models and AI-driven crime landed. Consumer brain–computer interfaces and AI-driven job losses did not. My score: 62.5 out of 100.
- Author: Dr Mark van Rijmenam, CSP
- Tags: Trends, Technology

Anyone can publish predictions. Almost nobody publishes the scorecard, and that is exactly why most predictions are worth so little.

A year ago I released [*Ten Technology Trends for 2026*](https://www.thedigitalspeaker.com/ten-technology-trends-2026/) and called 2026 the Year of Augmented Intelligence. A forecast you never audit is an opinion with good design, so before the 2027 edition goes out I checked all 10 of my predictions against what actually happened, which is the Verify step of my own WAVE framework applied to myself. 

I score direction, magnitude, timing and adoption separately, because a trend can be right and still land late or small. My total is 63.5 out of 100, and the pattern is unflattering and useful at once: I read the direction right far more often than I read the date. Here is how the 10 trends performed.

## Spot On

### 2\. Small language models: 9/10

This was my best call. I wrote that the center of gravity would move from trillion-parameter giants to small models living on devices, and the industry spent 2026 building exactly that. Google released [Gemma 4 on 2 April](https://developers.googleblog.com/en/bring-state-of-the-art-agentic-skills-to-the-edge-with-gemma-4/) to run agent workflows on devices. 

On 13 November Microsoft announced that [Aion Instruct, a 2.3-billion-parameter model, will replace Phi Silica on Windows](https://devblogs.microsoft.com/foundry-on-windows/aion-instruct-local-ai-windows-gpu/), with retail devices planned for January. Frontier models still matter, but the architecture I described, with large models as the escalation layer and small specialists at the edge, is the one being built.

### 6\. AI-driven crime reaches every corner: 9/10

The direction is right, and November made it harder to argue with. The FBI's 2025 Internet Crime Report logged 1,008,597 complaints and $20.9 billion in losses and, for the first time, [counted AI separately: 22,364 complaints and $893 million](https://www.pymnts.com/cybersecurity/2026/fbi-flags-893-million-in-ai-driven-scams/). [Europol](https://www.europol.europa.eu/publication-events/main-reports/iocta-2026-evolving-threat-landscape) and [ENISA](https://www.enisa.europa.eu/news/exploring-the-evolution-of-the-cyber-threat-landscape-how-dependencies-weaken-our-digital-resilience) both report growing criminal use of AI. On 16 November Google's Threat Intelligence Group [reported](https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold/) agent frameworks in live intrusions and criminals reselling stolen LLM access, called LLMJacking. It also identified, for the first time, a framework that found and exploited vulnerabilities without a human in the loop.

I held back a point for the autonomy I promised. That case is a single exposed server, and AI-related losses are about 4% of the FBI total. Even the closest earlier case, the espionage campaign [Anthropic disclosed in November 2025](https://assets.anthropic.com/m/ec212e6566a0d47/original/Disrupting-the-first-reported-AI-orchestrated-cyber-espionage-campaign.pdf), had AI running 80% to 90% of tactical operations while humans set the targets.

### 7\. Growing unrest over privacy-breaking hardware: 8/10

The tension arrived on schedule. The European Data Protection Board adopted [Statement 01/2026 on smart glasses](https://www.edpb.europa.eu/system/files%5Fen?file=2026-11%2F20261013%5Fedpb%5F012026%5Fstatement%5Fsmart-glasses%5Fen.pdf) on 13 October, warning that bystanders may not know their data is being processed. In September Reuters reported that [privacy fears could slow adoption](https://www.ctvnews.ca/sci-tech/article/from-smart-glasses-to-ai-pins-privacy-fears-challenge-techs-next-big-bet/) of AI glasses and pins. 

I described a full revolt, with no-record zones and an exploding counter-surveillance culture, and I find no sign of it at that scale. Counterpoint counted [AI glasses shipments up 263% year on year](https://counterpointresearch.com/en/insights/global-ai-glasses-shipments-surge-263-percent-yoy-with-display-less-ai-glasses) in the first half.

## Partial Progress

### 1\. China overtakes the West: 6.5/10

I wrote that China would outpace the West across AI, photonic computing, robotics, semiconductors and manufacturing integration. The core of that held. Stanford's 2026 AI Index finds that the US–China model performance gap [has effectively closed](https://hai.stanford.edu/ai-index/2026-ai-index-report). 

The International Federation of Robotics counts [354,000 industrial robots installed in China in 2025, 59% of the global total](https://ifr.org/ifr-press-releases/news/five-million-robots-now-operate-in-factories-globally). The US still leads in frontier models and capital, so "overtakes" was too broad. But the engine underneath it, AI, robotics and manufacturing fused into physical deployment at scale, is what I described.

### 3\. A data warehouse in a box: 7/10

I predicted that intelligence would leave the far-off cloud and settle into compact appliances beside the production line. The direction held. A Cisco-commissioned Omdia survey of 1,201 IT decision makers found that [96% already run AI across a mix of cloud, on-premises and edge](https://www.cisco.com/site/us/en/products/computing/offers/assets/omdia-ai-infrastructure-infographic.html). The same survey found 97% deploying or planning to deploy edge inferencing within 12 months, while 60% have already moved some AI workloads back on-premises.

The magnitude is where I overreached. The money did not leave the cloud. Amazon, Alphabet, Meta and Microsoft [spent $170 billion in the second quarter alone, up 72%](https://www.wealthprofessional.ca/investments/equity-markets/big-techs-ai-spending-clouds-the-earnings-picture/393161), according to the *New York Times*. Amazon also raised its 2026 capital spending plan to about $220 billion. I called a pivot to the box. What happened is a split.

### 5\. The rise of humanoids as a service: 7/10

I wrote that 2026 would not be the year humanoids take over. It would be the year they tip from prototype to practical tool, with economic parity against human labor not before 2027 to 2028\. That held. Counterpoint counted [more than 22,000 humanoid shipments in the first half, up nearly 300%](https://counterpointresearch.com/en/insights/global-humanoid-robot-shipments-soar-nearly-300-percent-yoy-in-h1-2026). 

Yet manufacturing took about 13% and logistics about 5%, while entertainment, research and data collection took more than 60%. Agility Robotics' preliminary registration statement, filed with the SEC in November, describes [one customer committing to deploy $300 million of Digit 5 robots over several years](https://www.sec.gov/Archives/edgar/data/2074973/000121390026118025/ea0311220-s1%5Fagility.htm), subject to milestones. It also says its robots-as-a-service model has not yet shown profitability at scale. The shipments are real. The labor economics are not there yet.

### 10\. AI-powered job losses hit escape velocity: 6/10

I wrote that we had crossed from forecast to fallout. On the narrow measure, announced layoffs that cite AI, that is right. Challenger counted [121,401 announced cuts citing AI through October, about 20% of the total](https://www.challengergray.com/wp-content/uploads/2026/11/Challenger-Report-October-2026.pdf), against [54,836 in all of 2025](https://www.challengergray.com/blog/2025-year-end-challenger-report-highest-q4-layoffs-since-2008-lowest-ytd-hiring-since-2010/). October's 1,265, however, was the lowest month of 2026.

Tech showed the pressure most clearly. In October Microsoft offered voluntary buyouts to [60,000 employees, about 27% of its workforce](https://www.cleveland19.com/2026/10/26/microsoft-offers-60000-employees-voluntary-buyouts-ai-spending-accelerates/). In November Canva said its restructuring would [affect 1,271 people, about 22% of its staff](https://www.deccanchronicle.com/amp/business/companies/canva-announces-layoffs-job-changes-for-about-22-of-workforce-2001715), who would lose their jobs, change roles or relocate. Neither company named AI as the cause, so I treat both as context, not proof.

On the broad measure the case is not made. Stanford's Digital Economy Lab found [no evidence of widespread job displacement](https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries%5FAugust2026.pdf). But employment of 22- to 25-year-olds in the most AI-exposed occupations sits 19% below where it would be had it kept pace with less-exposed peers, mainly through reduced hiring. The authors call that descriptive, not causal. The [OECD](https://fortune.com/2026/06/03/oecd-economic-outlook-scarring-effects-recession-ai-labour-displacement/) found in June no sign of widespread displacement at industry level. In November the [ILO](https://www.ilo.org/publications/ai-and-developing-economies-promoting-decent-work-time-rapid-technological) described the labor-market effects of AI to date as modest and slow-moving. The first casualty is not the job. It is the next hire.

## Missed

### 4\. The AI-enhanced metaverse returns: 4/10

I titled this trend "The AI-Enhanced Metaverse Returns." The metaverse did not return. Meta cut [more than 1,000 jobs, roughly 10% of Reality Labs](https://www.entrepreneur.com/business-news/meta-starts-cutting-jobs-signaling-a-broader-shift-in-strategy) in January. Omdia reports that [global XR headwear shipments fell 38.7% to 1.9 million units in the first half of 2026](https://telcomagazine.com/news/omdia-xr-shipments-decline-38-7-metas-dominance-wanes). Glasses went the other way, as the Counterpoint figure above shows. 

What survived is the machinery I described underneath the label: simulation, digital twins and synthetic worlds. At the Conference on Robot Learning in November, [NVIDIA released Isaac Lab 3.0 and the GR00T 2 model](https://blogs.nvidia.com/blog/robotics-simulation-advancements-corl). NVIDIA's announcement frames simulation as the place where robots rehearse delicate work before they handle the real thing. The market rejected the brand and kept the infrastructure.

### 8\. Brain–computer interfaces move into consumer life: 3/10

I wrote that by 2026 these would sit inside headphones, glasses and workplace tools. Medicine moved: on 13 March China approved Neuracle's implant for adults with cervical spinal cord injuries, [which officials call the world's first invasive brain–computer interface cleared for commercial use](https://www.sixthtone.com/news/1018307/China%20Clears%20First%20Brain-Computer%20Implant%20for%20Commercial%20Use). 

Consumer life moved only in its mildest form. In October Neurable launched [$499 headphones with built-in EEG sensors that track focus through the working day](https://www.engadget.com/2271652/neurable-one-eeg-headphones/), and NextSense sells [EEG earbuds that track sleep](https://www.soundguys.com/nextsense-smartbuds-bring-eeg-brain-sensors-to-wireless-earbuds-for-the-first-time-152745/). They measure and report. I find no consumer product that lets you command a device with your thoughts. The technology is moving and my timing was wrong. I now put consumer BCIs that act on your intent several years out.

### 9\. Dramatic healthcare discoveries increase our healthspan: 4/10

The science moved, but the outcome did not. In September, [*Nature Biotechnology* published an analysis](https://www.nature.com/articles/s41587-026-03286-y) applying six proteomic aging clocks to a 12-week phase 2a trial of rentosertib, an AI-designed drug for idiopathic pulmonary fibrosis. The clocks pointed toward lower biological age, but the analysis cannot show whether the drug affects aging beyond what it does for the lung disease. 

A *Nature Reviews Drug Discovery* perspective published on 7 August finds that [clinically relevant impact from AI drug discovery remains limited](https://www.nature.com/articles/s41573-026-01496-2). One correction here: I cited an 80% to 90% Phase I success rate for AI-discovered drugs. The same analysis puts Phase II success at about 40%, [comparable to historic industry averages](https://pubmed.ncbi.nlm.nih.gov/38692505/), on a limited sample.

## What the Scorecard Says

My strongest calls had several independent signals pointing the same way: small models, AI-driven crime and the tension around pervasive hardware. My weakest needed a chain of uncertain steps inside 12 months: a consumer brain–computer interface, extra healthy years from a drug result, a macroeconomic shock from AI adoption. My upcoming 2027 report will state for every trend what would prove me wrong, and I will score it in public at the end of 2027\. A prediction you cannot lose is not a prediction, it is a mood.

To get the report the day it launches, subscribe to my newsletter.