The reality of AI is going to hit soon and most will be caught with their heads in the sand.
When a nuclear bomb detonates, there is a sky-spanning flash often described as “brighter than the sun”. Then, nothing – not even sound - until the shockwave hits. The duration of this period of awful grace depends entirely on the distance from the epicentre. For some it’s seconds, for others, handfuls of minutes. Survival depends on what you do in that window.
The world of digital technology is currently living in a similar state of suspension. The widespread availability of large language models is a world-changing event that occurred. Generative AI isn’t coming, it’s not about to, it won’t – it happened. The flash happened. It was powerful and undeniable.
The survival of any industry that uses computers in any capacity now relies on reacting coolly and with foresight. Typically, as expected, there doesn’t seem to be a whole lot of that happening.
Professionals are terrified for their jobs, as apparently AI is coming for them all. Consultants sprouted AI services to future-proof against the oncoming storm. Programmers, writers, artists, musicians, engineers, knowledge workers and personal assistants all have a polarizing hot take on what is coming next and why Gen AI is either good or evil.
These reactions, while heartfelt and entirely human as part of the ever necessary period of grieving, are largely useless. No one is preparing for the true impact of generative AI models.
Scientific papers are pointing at the industrial revolution and at the arrival of the printing press, and drawing conclusions based on historical models. While this instinct is at least facing the correct cardinal direction, it still falls wide off the mark.
Large language models have already led to a competence compression event so powerful and pervasive that the world of technology is still reeling. Translators are struggling to establish their relevance. Grandmothers can operate image generation diffusion models without even knowing how to spell words correctly. Executives can receive complete, largely correct breakdowns of complex situations in minutes.
Competence compression is what happens when tool access collapses the gap between novice output and professional output. As a result the floor rises to meet “good enough” requirements in many domains. And it now is both free and freely available to anyone with a phone.
This is the consequence of generative AI in its infancy, in its experimental growth phase that changes every quarter and keeps improving with every generation.
Looking to history does lend a fairly good idea of the mechanisms about to play out on a social scale. Just like when the printing press, the industrial revolution, or photography happened, technology will bring crafted goods to the masses.
All professionals who occupied the mainstream market will be displaced, and will have to find work based on either their exquisite skills that the technology can’t yet replicate, or working as control layer to advance the new technology. There has never been a third mainstream option.
While the past offers some insight on the generalities of what to expect, the specifics are, this time around, terrifying to any sane mind. Generative AI will cause a widespread knowledge compression across all knowledge-based trades, writing, art, and programming all at once, at digital speeds.
What’s coming isn’t just a shockwave, it’s a Krakatoa-sized explosion aimed right at the heart of the entire tertiary sector in the form of a re-pricing of competence itself.
Survivors of a nuclear blast have to worry merely about seeking protection within minutes, and then stay put a few days while the environment is bathed in skin-melting radiation. There are guides and protocols informed by the dozens of tests, and real life deployments of nuclear ordnance.
There is no comparable source of comfort for digital natives facing commercial irrelevance and a future radically different from the landscape that shaped them. This does not mean we can’t trace some paths and draw some reasonable conclusions.
The first truth to swallow is that Gen AI isn’t going back in the bottle. Even if the current crop of AI developers, from OpenAI to Anthropic were to go bankrupt in the next 24 hours, this technology is going nowhere. Its advantages to industry and automation, the access it provides to enormous troves of knowledge and complex technologies are far too precious to surrender.
Many artistic professionals aimed their hope at the courts. However, in the US, a federal decision held that training AI on books, under the circumstances the case presented, could qualify as fair use of the material (https://www.theguardian.com/technology/2025/jun/25/anthropic-did-not-breach-copyright-when-training-ai-on-books-without-permission-court-rules) . The general view is that Gen AI isn’t infringing copyright by just being trained on copyrighted material, as AI learns just like a human mind: it absorbs a lot of sources, remixes them, and makes new stuff. The law is not coming to the rescue.
Next, short to medium term survival relies on bluntly realizing that skilled workers remain necessary, but not in the way their job descriptions currently imply. Gen AI performs extremely well in scenarios where both the input and output of its process can be defined precisely. It struggles greatly against complex contexts, shifting requirements, hidden constraints, and accountability. That is the gap that remains human, for now.
This is why “control layer” roles will expand, and why mainstream production roles will compress. There will be demand for professionals who can translate intent into constraints, define success criteria, anticipate failure modes, and evaluate outputs rigorously. There will be far less demand for people whose primary value is the mechanical production of first drafts, first implementations, first passes.
It will be a brutal downward slide for employed professionals who do not learn how to incorporate AI into their workflows, and how to remain competitive. The positions at the top, the ones that set direction and own outcomes, will remain a fraction of the current workforces in writing, programming, translation, research, and data manipulation.
In a world where “good enough” is free, being good enough will no longer suffice. Mastery is not safety either. It is simply the minimum price of admission to the next labor market.
True long term survival requires a painful separation from the past, which many are greatly struggling with. While writing and art play by slightly different rules, where gen AI is able to replicate but not create fully, there are many professions that will either cease to exist or be so radically transformed that insisting on keeping things as they are will cause tremendous loss.
Programming, for example, will see the complete annihilation of the human implementation layer. Three decades ago, high performance programming was carried out in assembly – one step up from machine code. Two decades ago, managed high level languages such as C# freed programmers from having to manually manage memory on a device. A decade ago, markup languages and interpreted scripting languages such as Javascript further removed knowledge of machine architecture and operated on completely software-driven substrates.
It is not a crazy prediction to see that 15 years from now, most junior programmers won’t know what their actual code will look like, as they work on an abstraction layer that leaves hardware implementation purely to the AI, and leaves them to make sure the system is well designed and performs as indicated.
These programmers of the future will not need to know what today’s programmers had to suffer through when they couldn’t get pointers right in ANSI C back in university. They will need to be schooled in what matters, which is system architecture, storage strategies, data transport layers, and deployment pipelines.
Truly dedicated alpha nerds may choose the ascetic path of actual programming languages to help advance the substrates of implementation and hardware architecture, but the majority won’t be able to actually write a script without AI assistance.
This is a perfectly predictable path. All knowledge professions, many parts of art production pipelines, translation services, and all data manipulation shops are heading in similar trajectories, in their own time.
The true change generative technologies brought is to remove an entire industry of craftsmen who kept technology running. What kind of future does that leave for those who will struggle to adjust to the new reality?
There will be pain, for sure. However, the new world generative AI technologies unlock isn’t an inherently unfriendly landscape. The parts that generative AI compressed away in its mighty cybernetic folds were, frankly, the horrible, repetitive bits that most professionals have been trying to automate away for years.
The future that’s already happened will mean that we have to train ourselves to refocus what truly matters. Detail work and craft will shift to the organization and structure that encode the true sense and meaning of what we’re trying to accomplish.
A lot more minds will spend a lot less time bogged down in the minutiae of formatting and converting and struggling with colour spaces, and will be able to focus on planning ahead.
This is a future the world is terrifyingly unprepared for. Universities can’t prepare young professionals for a future where each profession is a cybernetic union of AI and human ingenuity. Current professionals are still busy deciding whether and how they like AI, they haven’t even thought of how to preserve what truly matters of their crafts.
This is the time for cool heads and foresight to prevail, not to marvel at the beauty of the flash, or to suddenly decide you always were a pacifist after all.