Sorry folks, you can't avoid AI

I'm on Mastodon, aka the Fediverse. A big reason people are there is because they do not like the big tech, corporate, algorithmic social media platforms, and they want more control of their social media identities and online lives.

My Tribe #

So it's hardly a surprise to see lots of anti-AI sentiment on the platform. And to be honest, I consider many of these people to be part of my tribe as we generally have common values when it comes to the internet.

The hype from the AI companies is crazy intense though. Of course, the AI companies want everyone to believe AI is going to transform the world because they want us all to buy the Pro subscription to their AI models. They want companies and government to buy the hype and replace workers with their robots. They want to capture some portion of employee salaries as subscription revenue.

Corporate executives love the idea because they get to spend less on salaries, which gives shareholders a boner about profits. Their profits go up, the shares go up, and all the C-suite douchebags get to buy their boats and vacation homes because of their supersized bonuses.

Testing the Hype #

I think my tribe has it wrong about AI though. This isn't just another fad, looking for a problem to solve like blockchain crypto bro hype. Agentic, generative AI really is transformative. I too saw all the hype from OpenAI, Microsoft and the like and immediately felt dismissive because it felt like blockchain all over again. I tried out the different AI models to form my own opinion, but I was using it like a search engine. Or I'd copy in some code I wrote and ask the robot to help me improve it or to offer recommendations. I used it to help me write work emails to leadership or customers and to write user stories using our team templates. This use case felt productive and helpful, and I found value in the tools, but it didn't feel like it was going to change the world. It felt like the natural next step of Intellisense or an advanced search engine.

But then my brother sent me Something Big is Happening (sorry about the twitter/x link) from a 'CEO' of an AI company I'd never heard of. We often talk about this topic, so I was prepared to hate-read the article and promptly dismiss it. Who can believe AI company CEOs anyway?

He described a workflow different than the one I've been using, so I installed Claude Code in VS Code so I could get access to the Claude Opus 4.6 model. I wanted to steel man this person's argument so when I inevitably laugh and say this guy is full of shit, I can do so from a position of experience.

I used this setup and workflow to redevelop my website recently. I wanted to go to the latest version of 11ty plus I wanted to change my layout some. I was pretty surprised at how well it performed the job. I didn't just dump my old site in there and tell it to upgrade it with a so-and-so layout. I carefully and slowly built my templates and scripts using the tool until I got what I wanted. I'm really happy with the result, and I believe I made a better website quicker than I could have if I just did it the old-fashioned way.

With that completed, I wanted to experiment more with what agents can do. This led me to Model Context Protocol (MCP), which is a protocol that allows these AI systems to interact directly with a specific system. In my case, I wanted to see if I could use Claude Code to manage my Home Assistant instance and sure enough, there was an MCP server available and I happily tasked Claude to review my instance, automations and scripts. It flagged an issue with my RokuTV. I recently updated the firmware, you see, and that reset a certain setting that needed to be turned on. At this point during troubleshooting, Claude asked me to verify if the IP address of the TV was a specific IP address. I use Unifi networking gear in my house, so I set up the Unifi MCP server and asked Claude to perform the check itself. It did so successfully.

To be clear, an AI agent, running in VS Code on my laptop, can access, query and manipulate my local Home Assistant instance directly. It can create automations, scenes, scripts and dashboards. It can do the same thing with my Unifi equipment and my Synology NAS. I can tell Claude to configure new network storage in Home Assistant, and it will configure my Synology NAS with a network share, then set it up in Home Assistant. It can recommend the Synology NAS device be set to a static IP address and it can perform that step on my networking gear too. The thing that makes it feel transformative is it can run the commands, capture the output and fix errors, as they happen. There's no copy and paste back and forth while we argue about an old command syntax that's been changed. That feels like magic. It's what assistants are supposed to be.

The Haters #

Many of my tribe who reject AI do so for good reasons.

• AI makes us less proficient. The more I use Claude Code to manage my Home Assistant instance, the more I lose my ability to do it myself. I'll forget the syntax of the dumb YAML configuration for mushroom cards. This isn't hypothetical. Researchers call it "deskilling," and it's already showing up across industries.

• AI products make us more dependent on them. If we're losing proficiency because we've offloaded the practice, that's a real problem when OpenAI or Anthropic go down. And they do go down. ChatGPT's June 2025 outage disrupted over 500 million users, many of whom had built it into their daily workflows.

• AI agents could be vulnerable to hacks and data exploits. In order for these agents to access all the different systems, we must configure them with the secret keys and credentials to do the work we're asking them to do. That's a lot of trust to place in tools and protocols that are still very new, and a compromised agent with access to your network gear, NAS, and home automation is a nightmare scenario. Imagine the same scenario at work.

• AI data centers use a lot of water and have a negative impact on the environment. Large data centers can consume up to 5 million gallons of water per day, equivalent to the needs of a town of 50,000 people, and that demand is projected to keep climbing.

• AI data centers use a lot of power and are making energy bills higher for everyone. NPR's Planet Money traced one Ohio couple's electricity bills and found they'd risen 60% in five years, driven in part by grid expansion to serve nearby data centers. The AP reports this has become a bipartisan flashpoint, with politicians from Trump to local lawmakers insisting tech companies, not regular people, should foot the bill.

• AI corporations exploit vulnerable data labelers desperate for work to make the technology "safe" (primarily for Western audiences). Kenyan workers earning under $2 an hour have described their conditions as "modern-day slavery", labeling graphic content including violence and child abuse for eight hours a day, with minimal mental health support, to train systems like ChatGPT. Big tech in general outsources this work to countries with high unemployment and weak labor protections, then distances itself through layers of subcontractors.

It's Coming #

I think points 1 and 2 are the disruptive parts for tech workers like myself. Points 3, 4, and 5 are engineering problems that can be solved. Point 6 is just companies doing capitalism and is not unique to AI.

I'm not an AI champion. I think big tech, capitalist, companies will try to capture the market, and I'm extremely skeptical the US government will do anything about it. In my heart of hearts, I believe AI as it's being envisioned and the way companies will use it is very problematic. I just don't think the right response is to ignore and avoid it like many of my tribe are doing. I will be trying to learn all I can about these tools because I believe it's coming.