“What you will be will depend on the perspective you have”
— Ray Dalio.
I believe 2026 will be remembered as a turning point in how AI transforms work. For those close to the technology, the changes are substantial, and their impact is already visible. AI has already changed my domain (Computer Science) and how people work in fields I’ve been following closely1. And yes, it is true - not everything is solved; significant technical challenges remain2. As these challenges are addressed, changes will continue.
A critical distinction worth calling out lies between automating tasks and automating jobs. While current AI models are capable of performing various specific tasks (e.g., writing coherent text, generating code, or answering questions), this is fundamentally different from automating an entire job, which typically requires managing a complex, multifaceted workflow. In other words, an AI’s ability to handle individual tasks3 reliably does not mean it can perform an entire job. This helps in understanding the real-world economic impact of AI4.
The question is how people can understand, prepare, and reconsider their assumptions about changes they have not yet experienced. Although the technology is widely disseminated, its use is still constrained. Barriers include cost and gaps in education. Access to the frontier intelligence is expensive. However, this situation will continue to change and I expect costs to continue falling (also worth reading scaling laws). As capability density rises - meaning newer models5 achieve comparable performance on specific tasks using fewer effective parameters and the tech community continues to work on many other optimization techniques - these improvements can reduce operational inference costs and make lower API prices possible for end-users.
Another thing worth noting is that today’s most popular computer platform is our phones (DataReportal / ITU Data, Cisco Annual Internet Report). So, for a significant portion of the global population (60-70% according to the most recent numbers), “AI” is what they experience on these devices6. However, this can leave users with only a limited view of what current-generation AI models can do. Someone who uses AI only for brief chats or image generation may not have encountered workflows that combine models with files, tools, systems, reasoning levels, and long-horizon tasks. And this fact holds even when considering the latest product surfaces from the top US labs7, meaning that many people haven’t yet seen how much and how fast things have evolved from the end of 2022 to now (2026).
That gap in experience matters when communities discuss how to respond. The changes to the global economy started back in 20238 and have continued to increase since then. Society and leaders must start having conversations about these changes and address them upfront at a local level and expand from there (and yes, in democracies, elections provide a space to start these conversations). There is no one-size-fits-all solution. However, the more these hard conversations get pushed back the harder and longer it would take to address, align, and move forward to agree on a response (read Bill Gates’ note from Aug 26, 2026, for an idea of important topics that require discussions9). Regardless of creed, culture, nationality, language, or origin, debates and good-old-fashioned human-to-human conversations are needed to address the changing environment, to educate people, set new societal values, and update laws when needed.
This is a once-in-a-generation transition. These moments cause discomfort. Throughout this phase, global societal values will change and adapt. In light of these transformations, I predict that human values - ethical commitments, judgment, and relationships - will become even more important and valued. Finally, education is a must for anyone or any nation to adapt to this new world revolution, as it helps people understand capabilities, recognize limitations, and assess effects on their own work, leading to informed debates.
Footnotes
Today, a significant list of industries have been consolidating changes in how they work (links are only to provide a few examples): Software Engineering, Scientific Research (all areas), Education, Mathematics, Biology, Legal & Compliance, Creative Industry, Gaming Industry, Music Industry, Design Industry, Movie Industry, Writing/Authoring, Health & Medicine, Finance (fintech, investments, advisory, etc) are among the fields whose changes I’ve been following most closely. In addition to task-level execution changes, AI is enabling employees to perform tasks traditionally assigned to other roles, leading to a phenomenon known as ‘toe-treading’ or workplace overlap friction.
See the posts by OpenAI's Greg Brockman (President & Co-Founder) and Jakub Pachocki (Chief Scientist) from Sep 06, 2026 associated with the state of AI detailed on Jakub's recent writing, An Alien Mind. Also worth checking Jacob Coxon's (pretraining research at both OpenAI and Anthropic) thread and Evan Hubinger (Alignment Science lead at Anthropic) take. If you want to read more on technical challenges, feel free to check a list I put together. Aside from the technical challenges, there are still enormous gaps in education and a complete new set of standards of how work will be done moving forward, even within Software Engineering; just look at Boris Cherny (Anthropic) LinkedIn post about emails and messages he receives today (September 2026).
Check METR's study on Task-Completion Time Horizons of Frontier AI Models, last updated on May 08, 2026, and take a look at the plot of the length of software tasks that different LLMs can complete; this will give you a sense of how quickly things have progressed.
The distinction also matters when interpreting employment data. In a March 2026 study, Anthropic found no systematic increase in unemployment among workers in highly AI-exposed occupations since late 2022, although it found suggestive evidence of slower hiring among younger workers in exposed occupations. These findings describe early effects, leaving the longer-term impact uncertain. Source: “Labor market impacts of AI: A new measure and early evidence”, https://www.anthropic.com/research/labor-market-impacts - Mar 5, 2026. My sentiment is that if the same study were updated today, based on my reading of the hiring market (not a rigorous measure), it would show a new dynamic: reduced team sizes (i.e., a reduction in people employed) as more tasks are being automated and people are taking broader roles with neighboring responsibilities that they couldn’t do before.
Major AI model releases have accelerated every year since ChatGPT launched. AI Release Tracker tracks 249 frontier models across 11 labs — the number of major releases rose from 22 in 2023 to 92 in 2025, a 4.2× jump in the monthly release rate, with 76 more already shipped in 2026 (through 10 Sep 2026).
An interesting point to make here is that neither OpenAI nor Anthropic provide analysis or data on smartphone or mobile device usage in their official public reports, as of Sep 10, 2026. You can check the links to their public reports and data websites in the References section of my note here.
I.e., OpenAI, Anthropic, Google DeepMind, Microsoft AI, Meta AI, SpaceXAI.
After the ChatGPT moment on November 30, 2022, when OpenAI launched the initial public release powered by GPT-3.5.
Bill Gates proposes that world leaders, experts, and communities focus on building a comprehensive transition plan across three core areas:
(a) Build a new system for managing the transition: Establish domestic policy frameworks and international governance agreements to navigate rapid economic and social upheaval before widespread displacement occurs.
(b) Set aside some jobs for humans (“Human Reserved”): Intentionally designate specific roles to be preserved for people - such as eldercare or delivering sensitive medical diagnoses - where human empathy and dignity matter more than technical efficiency.
(c) Rebalance how we tax labor and capital: Reform the tax system by taxing AI tokens and robots to offset lost income tax revenue, remove tax incentives that favor replacing human workers with machines, and generate funding for worker retraining programs and a stronger safety net.