Have you opened X or a tech news site lately? Then you are probably familiar, even exhausted by, posts like these:

Do you want to hear a secret?

This is marketing.

The message is simple:

AI will beat the smelly humans. It is faster, cheaper, tireless, and anyone who doubts this (usually the experts in a field) is a dumb luddite who failed to adapt (and will end up in the permanent underclass).

Conveniently, adaptation means buying the $200/mo subscription and convincing your boss to spend $4000/mo on tokens so you can work faster/better/harder.

Artifacts are not Work

Compilers did not replace programmers. Spreadsheets did not replace accountants. CAD systems did not replace engineers. These tools changed what competent people, the experts, could do, how fast they could do it, and how much complexity they could manage.

A compiler can translate code, but does not understand software. A programmer does. A spreadsheet can calculate, but does not know what calculations are important for the next quarter. An accountant does. CAD can model a bridge, but it cannot decide whether to simply upgrade the one 200 meters downstream. An engineer can.

Visible output is just a small part of the actual job done. AI companies reduce work to the visible output, the artifact, completely ignoring the value in understanding that comes from the process of producing that artifact.

In programming, the bottleneck is understanding the business problem, designing the system architecture, and maintaining the codebase. Sure, you can write the perfect prompt to encapsulate your entire business problem. Ask any business that tried outsourcing in the last 30 years, they might have some tips if you want to go down that path… The hard thing is knowing if something is good, what tradeoffs it makes, and the consequences of a particular implementation or choice.

Code is designed to be deterministic. The idea that a probabilistic system will replace compilers is fundamentally silly.

The same thing is starting to happen with cybersecurity. Take the Mythos/Fable release from Anthropic. Among the serious security people I know, the reaction to Mythos was not awe. It was frustration: Anthropic took a genuinely interesting capability and wrapped it in reckless “cyberweapon” marketing. Also, no it did not hack the NSA. The damaging claims around AI “cyberweapons” devalues the meticulous work of actual security researchers and pentesters, and overshadows the actual capability of a model like this.

Here’s the thing

All of these AI companies are facing a massive problem. Currently, there is a huge amount of investor appetite for AI. The AI companies know it won’t last.

Selling “the future” for investment will soon give way to investors demanding something in “the present”. At some point numbers start mattering. Quarterly revenue, customer retention and margins. Customers asking: “did this really save us money?”

They are capitalising on the hype while it lasts, and convert hype into capital, customers and dependency before investors start caring more about margins than mythology.

Assist the work, not just the artifact

I have a dream about the future of AI. A dream where AI complements human experts, rather than threatening to replace them.

Back when tools like Unity and Unreal were released, they drastically lowered the barrier of entry and more low-quality asset-flip games were made. They did however allow smaller teams of people and solo developers who genuinely had the skills to make games, to build their dream project without sacrificing on technical scope due to not having the resources to develop their own game engines from scratch.

This is an example of tools empowering creation and innovation, and yes there is a cost to it too (asset-flips).

Midjourney Medical is interesting because the underlying idea points in the right direction: better tools for people with expertise. A tool like this could give doctors and researchers better, less invasive insight into the human body. That does not “cure cancer” by itself, but it could make experts dramatically more effective at diagnosis, research, and treatment. But the bougie spa/wellness framing also shows how quickly even useful technical ideas get wrapped in consumer futurism.

With all that in mind.

Hark my request AI executives

Give smart people the tools to do smart things instead of threatening that you will replace them with a clanker.