**Here’s an Overview of How AI Evolved in 2025, and How My Forecasts Fared**
The years 2023 and 2024 were pivotal for AI, but as we neared 2025, both consumers and industry experts raised concerns. Some doubted the trajectory of AI development, while others worried that companies like Google and Apple might overpromise and underdeliver.
At the end of last year, I made multiple forecasts, and the majority have been realized. AI has swiftly embedded itself into operating systems, becoming increasingly agentic and actionable. However, I was taken aback by Samsung’s clarity in differentiating between on-device and off-device Galaxy AI capabilities, permitting users to disable online features with a simple switch.
As we wrap up 2025, let’s assess the performance of the AI sector and the success of its leading entities.
**Evaluating My Predictions for AI in 2025**
After dedicated AI devices like the Humane AI Pin failed, I accurately predicted a transition towards integrating AI within mobile operating systems. Google spearheaded this movement, substituting Google Assistant with Gemini throughout Android and Wear OS. Microsoft’s Copilot evolved into a crucial element of Windows, while Apple faced challenges with Apple Intelligence in iOS and macOS.
Gemini Live’s multimodal video and screen sharing, first a prototype, emerged as a fundamental feature for Android in 2025. Google broadened the availability of Gemini Live to almost all Android devices, marking a noteworthy milestone.
While Google progressed in multimodal functionalities, the development of agentic control was slower than anticipated. The launch of the Gemini Agent and Google Antigravity app happened later than expected, as did OpenAI’s AgentKit. AI agents are now a primary concern for leading corporations.
I had expected a greater number of AI features incorporated into operating systems, ideally featuring on-device processing. However, growth in this domain was not as rapid as predicted, with Gemini Nano and Galaxy AI managing some on-device functions, yet not to the extent desired.
In late 2024, both Google and OpenAI’s CEOs dismissed worries about an AI “wall.” I believed that competition would propel advancement, and indeed, firms competed for dominance in AI benchmarks throughout 2025.
I expected indistinct boundaries between on-device and off-device processing. Samsung’s toggle for AI data processing countered my expectations, and I am anticipating that other companies will implement similar options.
**Google in 2025: A**
Google thrived in AI during 2025, with Gemini 3 reaching remarkable achievements and gathering nearly 350 million monthly active users by November. Gemini’s brief period as the leading free app in the iPhone App Store indicated its potential beyond Google’s ecosystems.
Google AI Pro amassed over 150 million subscribers, generating revenue through advertising trials and AI development tools via Google Cloud. Highlights included Gemini Live’s video capabilities, the Google Antigravity AI-based IDE, and the Nano Banana image generator. Google’s free AI functionalities, even catering to discontinued smart home products, further cemented its success.
Despite a robust AI strategy, Google’s absence of brand collaborations, such as Meta’s with Garmin or OpenAI’s with Disney, may pose a challenge for 2026.
**OpenAI in 2025: B-**
OpenAI seems to be refocusing for the future rather than aggressively creating new AI products. While ChatGPT retains a substantial user base, its current priorities appear uncertain, with recent enhancements like group chats and possible ad placements.
OpenAI’s $500 billion valuation gives it leverage, and its transformation into a for-profit entity was noteworthy. Collaborations with Jony Ive and Disney, along with a functioning prototype of an AI device, may bode well for 2026.
If OpenAI capitalizes on its for-profit structure, develops a successful AI device, and utilizes Disney’s investment, 2025 could be regarded as a pivotal year. However, the current performance justifies a grade of B-.
**The Remainder in 2025: C**
The majority of AI companies, aside from Google and OpenAI, performed moderately in 2025. Meta received a B grade for its sophisticated face-worn multimodal AI technology within smart glasses. The integration of Garmin with Meta AI was commendable, yet Meta’s excessive involvement of AI in unwanted areas overshadowed its accomplishments.
Apple and Samsung retained a standard performance, lacking any groundbreaking AI innovations. Apple’s ability to refrain from overpromising was admirable, whereas Samsung’s top AI features were largely attributed to Google.
In the AI model sphere, unexpected entrants like DeepSeek showed potential but quickly diminished.
**Final Assessment**
Despite the advancements in AI, the overall grade for 2025 stands at a B. While over a billion individuals utilize platforms such as ChatGPT, Gemini, and Copilot, the anticipated transformative effects have yet to be realized. The year 2025 concludes without the attainment of artificial general intelligence or solutions to significant global challenges. AI’s resource consumption raises concerns, and even OpenAI’s CEO recognizes the swift pace of change.
AI models and interfaces have become more potent than ever, yet the enthusiasm surrounding AI’s potential has not fully materialized. Given this context, a B grade is the highest I can assign AI.
