Kind of awkward to launch this and then go totally silent for three months. Honestly, it’s because I haven’t been building in my own time. Which is a bit weird for me. I should probably say that I haven’t been building things on the computer. I have recently joined Rimbal and pretty much all my appetite for tech is being taken up at work.

My time at home has been taken up largely by gardening and that time gives me a headspace to reflect on the last chapter of my #career, the current one but also what’s going on in what can be described as the next ‘age’ of the tech world?

I’ve realised something recently.. I’m a hypocrite? I’m a trend denier?

I am often a trend denier, not this time

Often is probably a bit strong, I think the most notable thing that I (still) disagree with is the usage of #kubernetes (and GitOps). I say this as BOTH a user of K8s at home and work. I know, a bit weird right?

As I was shovelling the 6th or 7th 500L bag of soil into my massively oversized raised beds, I was thinking about the advice I have given over the last few years vs the advice that I follow myself. It brings you to the classic:

Do what I say, not what I do

I should be clear, I’m not ‘preaching without practice’. I’ve always prided myself on taking the time to bring people on my thought journey and WHY I do things in a certain way. I just can’t help but reflect on the advice that I’ve given to various organisations (sometimes controversial) and how I am executing back as an ‘owner’ engineer.

So contrary to previous advice.. I am

• Running multiple K8s clusters without EKS Auto Mode as a small team
• Running databases on those clusters
• Skipping Control Tower and Config in AWS
• Co-hosting workloads inside the same AWS account
• Using #aws as nothing more than an infrastructure provider, no managed services

Just to name a few, all of this is well justified and well explored through LLDs and prototypes. I actually think these decisions are some of the best work that I have done in recent years. I’m sure there are many people who will enjoy debating with me about why I am wrong. Cool, I’d like a Whisky please.

Would I have done it this way in the past? Probably not.. but our unit economics are totally different. I now have this incredible #ai tool that can let me synthesise knowledge and do research at pace. I can hand low-value tasks off via what I call ‘Vibing on Rails’ (a topic for another post) and focus on the big problems in parallel.

It’s really never been easier to make high judgement decisions at pace, critical thinking and experience have never been more important. I think a healthy scepticism and distrust is probably a core skill these days.

Going from a problem statement to research and then leveraging prototypes to learn (and not end up in prod). Well, I was able to do this for a system that I am designing and try out half a dozen different ways of doing it all in the space of two weeks. In the past, I’d have had to abandon practical demos (and the scars that gives) and only do it for my top idea.

That’s an incredible de-risking tool but perhaps contrary to the productivity gains that people are so focused on. Sometimes it feels like we’re all part of this reinforcement loop that has gotten too biased towards a singular way of completing the tasks.

Mass adoption, the problem that can’t be solved

Part of being a young and fast moving industry is that we haven’t really figured out how to change effectively. We’re also a deeply interesting industry of both philosophy and deep technical topics.. sometimes the core industry and sometimes the tool. So no wonder it’s messy.

My previous employer has been in the news recently about how AI usage tracking and internal recognition systems are interacting in ways nobody quite planned. Honestly though.. that’s the model Bezos wrote about in the 1997 letter. Bold rather than timid. Some of these investments will pay off, others will not, and we will have learned another valuable lesson in either case. Interested to see what the iteration looks like.

It’s hard to find a catch-all way of working, Agile is a nice recent parallel. It started as a way to gain from customer feedback and cut the cost of late changes. It’s now a Scaled Agile monstrosity, the thing built to kill heavyweight process turned into heavier and arguably more confusing process. The Scrum Master went the same way: a hat someone wore became a full time, non-coding headcount with a course and a badge. Later on DevOps too, developers and operations sharing the pager turned into a job title and a tools team you raise tickets with. The industry does this every time. It takes a good idea that got people closer to the work and bolts on enough roles and ceremony that it becomes the bureaucracy it was meant to replace.

There seems to be an allergic reaction to spending time being wrong and a compulsive need to formalise everything into best practices and training. Even if it’s damaging.

With GenAI, it didn’t take long. Anthropic launched the Claude Certified Architect: Foundations exam in March. By June they’d added a Services Track with partner tiers, the lowest of which needs ten certified staff. Ten thousand consultants have passed already. Forty thousand firms applied to the network. Industry-specific specialisations on the roadmap.

OpenAI’s run the same play. Foundations certifications on Coursera. Frontier Alliances with the consultancies. The stated goal is certifying ten million Americans by 2030. Same shape, slightly different brochure.

Build the tool, name the discipline, ship the course, badge it, tier it, specialise it. Same playbook as SAFe, just running ten times faster. Same channel too. BCG, McKinsey, Accenture, Capgemini, the firms that sold the enterprise #cloud and agile. Last year Anthropic’s CEO said AI would wipe out half of entry-level white-collar work in five years. This year, ahead of the IPO, he’s walking it back. Either way, ten thousand certified consultants are signed up to deploy it.

Anyway

So am I a hypocrite? Apparently. The advice was for their teams; the practice is for mine. I can’t help but look forward to figuring out my own ways of working and avoiding the next tech adoption car crash.

The other hypocrisies are less defensible. Haven’t written here in three months. Sick of reading about #ai. And here I am, writing about AI.

Worse yet, I still have more to share..