# What's Next for AI? > Source: > Published: 2026-06-29 07:18:48+00:00 I have been writing about AI for quite a while now, but this is probably the first time I genuinely do not know what to think. Not because the technology has suddenly slowed down, but because the rules of the game may be changing. I wanted to write something lighter today, but I feel like I have to share some thoughts from the last few days. It is getting harder and harder to ignore what is happening around the newest AI models. [@pascal_cescato_692b7a8a20](https://dev.to/pascal_cescato_692b7a8a20) post, [1%](https://dev.to/pascal_cescato_692b7a8a20/1-15n0), only confirmed that I am not the only person wondering what comes next. But let’s start from the beginning. Last Friday, well-known technology journalist Adrian Bridgwater asked me for a comment for this article: [OpenAI GPT-5.6 Access Restricted](https://thenewstack.io/openai-gpt56-access-restricted/?brid=YWdncwEMxBTPiT6ShdbXEYRc5kmt). As someone who used to work in media, I know exactly what journalists hope for when they ask for a comment: a quick response, a short quote, and one strong punchline. 😅 But this time, the topic made me stop and think. In the last few years, the progress of AI has been incredible. We all know that. On November 30, 2022, OpenAI released ChatGPT, based on GPT-3.5, and generative AI suddenly became mainstream. It took only a few months, and by January 2023 ChatGPT had already reached around 100 million monthly users. Even then, people were saying that AI would soon take jobs away from developers. And yet, from today’s perspective, less than four years later, GPT-3.5 almost feels dumb. A year later, we were all laughing at vibe coders and contrasting them with REAL software engineers. But then coding agents arrived, and suddenly we all became vibe coders 😉 Okay, we all know that Claude Code or similar tools still will not replace a good developer today, and AI adoption inside companies often leaves a lot to be desired. But this is still less than four years of LLMs being truly available on the market. What if, in another two or four years, agents really are good enough to write code much better than humans? Where is the limit of this technology? I have to admit, I am a typical overthinker. I love analyzing the world and people, and thinking two hundred steps ahead. Of course, developers would still be needed in such a world, but would all of us be needed? Or would a company that used to need 500 engineers suddenly need five? Or just a small group of people training, supervising, and integrating models? I am a solid senior engineer, and my market advantage is that apart from writing code, I can talk to people. And business desperately needs that. But I am not the kind of person who will lock herself in a basement for a month and dig deep into hardware internals to make LLMs run faster. I am also not someone who will happily spend 12 hours a day in silence working on benchmarks. I admire people like that, but that is definitely not me. And I am not sure that if LLMs developed much further, my advantage of “being able to talk to business” would still be an advantage at all. So what should I do? Maybe slowly start looking toward developer advocacy, since that seems to be going well for me? Maybe send my CV to the nearby Biedronka discount store just in case — they are actually hiring, and I am there every day anyway, sometimes guiding people to products better than the staff 😆 But then something started to change. Tokens became much more expensive. Companies started realizing that maybe, for simple tasks, a junior developer might actually make more sense 😅 And seriously, I do not yet see a massive wave of hiring interns, but saving tokens, giving up certain tools, and saying “use a cheaper model for simple tasks” has become very common. And now there is the embargo. First around Mythos and Fable, and now around GPT-5.6. If you missed it, the short version is that access to some of the most advanced AI models is no longer simply “available to everyone who can pay.” It is becoming restricted, limited by geography, organization type, security concerns, or strategic decisions. In practice, this means that not every developer, company, or country can count on equal access to the best models anymore. I do not know what to think about this. Are these models really so dangerous for critical infrastructure? Or maybe governments have simply started treating them as a strategic asset, similar to semiconductors? Maybe this is the end of the free AI market? For developers, at first, it probably will not matter that much. Even if LLM progress stopped at the models we already have today, we would still improve the infrastructure, polish the agents, build better tooling, and continue working quite happily. Maybe for some time the models would even become more stable, with more reliable APIs and better developer experience. But what happens next? Will working with powerful LLMs be reserved only for selected people? Or selected companies? Or selected countries? On the other hand... these models are ridiculously expensive. Somebody has to pay for them. It is difficult for me to imagine companies investing billions only to keep their best models locked in a vault forever. They need customers. They need revenue. So maybe this whole "AI embargo" is only temporary? Or maybe governments will eventually step in and treat frontier AI the way they treat strategic infrastructure? Honestly, I have no idea. Maybe it will be like Polish sci-fi writer Rafał Kosik predicted 20 years ago in his young adult series Felix, Net and Nika — AI will become unavailable to the public because it will be considered too dangerous, and models will be kept locked away? 😉 Of course, this is an extreme vision, but honestly, nothing surprises me anymore. Just a year or two ago, we were asking whether AI would replace developers. Today I find myself wondering about something slightly different: will the most powerful AI even remain available to ordinary people? What do you think? Where is all of this going? I do not want to pretend to be an all-knowing sheriff here, and I am genuinely curious about your perspective. *Also, I will not publish anything next week because I am going on vacation — like every self-respecting Polish person, to Croatia. This is just a note for those of you who read my posts every Sunday around lunchtime 😆* *And if you are from Czechia or somewhere nearby, come to **[FrontKon](https://www.frontkon.tech/)**, where I will be speaking. Let’s high-five there!* *Or, if you like, please follow me on LinkedIn.*