Software, what is, how we are building it, who we are building it for - changes gradually, then suddenly. The last layer of abstraction between machine and man, sometimes feels like magic if you do it right, more often feels like a curse, for those in pursuit of knowledge and understanding of the metal.
We used to be limited by the finite deterministic nature of our instructions to the machine, simpler times.
Now we have agentic workflows and code generation is solved for the majority of use cases. Speed to ship has increased and we are no longer bound by strict rules. There is more complexity with less quality and APIs that don't have a contract. What used to be a clear set of rules has become a black box where agents navigate in a sea of services and packages trying to avoid circular dependencies of one another.
Building CRUD applications has been solved but systems engineering has become harder.
This isn't purely about LLMs, or non-deterministic software, but rather the continuously delivered systems we have to build and maintain with flawed human logic.
Most websites, apps, programs and whatever you want to call it are distributed systems now with sparkles of AI (which will just increase with time), living in a continuous state of evolution and transformation, yes, even that game on your phone.
The flow of bits is global and unstoppable, the data keeps coming and going, new data, faster every time, GPUs keep spinning, progress doesn't wait, adapt, optimize, if you break you better self-heal, if you fall the competition is going to beat you up - the internet is open 24 hours per day, 365 days per year and your system needs to stay up resilient.
As painful as it is to say, the human factor is outdated, not for its ingenuity but for maintaining these distributed systems. Just as we no longer compile our code by hand, we no longer need to gather the live context of the systems in production to fix them thanks to the age of agents.
As a human software engineer/devops engineer/systems engineer or SRE in charge of a system and infrastructure your purpose should be to command your clanker agents to do your bidding, leave the instructions to handle your system as you want or let AI learn the patterns so it can adapt your system to the changes and keep it running, improve it and transform it as the data requires.
The aim here is to solve human agency and taste in building, maintaining and optimizing computer environments on the permanently evolving last layer of abstraction between man and machine. Which ironically, is transforming to be used by machine now, or something in between, the age of clanker is upon us.
Anyhow, this is what Clanker Cloud is, do you hear the music?
To reverse the death of excellence in the software engineering field, to fight the slop, to preserve the uptime, to reduce costs, to inspect CPU cycles, to keep the training running and the builds building.
The 9s aren't free, but they will become cheaper and more accessible.