I’m a progressive. I have also been working as a full stack engineer since 2009, and I use AI almost every day. These two facts put me at odds with many of the people I organize with. Because the current default progressive position on AI is a firm “no.” It steals from the artist. It burns the earth. It makes billionaires richer. Full stop.
I agree with that more than you might expect. I would like to state what I agree with, where I think we are making strategic mistakes, and why I think those mistakes are hurting us.
The harm is real, so let’s name it accurately
Let’s start with the data center. Because that is where the anger is greatest and most justified. When companies build hyperscale data centers in neighborhoods, electricity bills go up, local power grids take on unintended loads, and many places draw water for cooling from already strained sources. The company then walked away with a tax break that passed all of that cost on to you. It’s a raw deal. The community that fights those projects is right to fight them.
The effort and creative side is also real. Training a Frontier model on the work of millions of artists without consent or payment is simple and straightforward extraction. If your livelihood is being fed to a machine that harms you, that machine is not neutral.
So I’m not here to say that concerns are overblown. it’s not. The point I’m making here is that we’re targeting them at the wrong target.
We treat “AI” as one thing, but it is something else.
There is a category error here. When progressives say “AI is theft and is destroying the planet,” that sentence doesn’t tell the difference between a company building a trillion-dollar data center and a disabled person using assistive tools to speak. I don’t know the difference between running Frontier Training and me generating frontend boilerplate on Tuesdays. It flattens everything into a single enemy.
The harms I’ve described are almost all concentrated in one place: how these models are built and deployed at scale. Building a data center. Uncompensated training data. A handful of companies that concentrate everything. That is the structure, and that structure deserves our scrutiny from every angle.
But the individuals using the existing tools are downstream from it all. I didn’t delegate the data center when running the query. I didn’t order a training run. The marginal cost of my usage is real, and it’s small and nowhere near leveraged. Directing progressive anger at individual users rather than construction facilities is like protesting climate change by yelling at people on a bus.
I know this because I’ve built something using a tool that I value very much.
About AntiFreeze and “vibe coding”
I created AntiFreeze, an app that allows people to report and track ICE activity to help keep their neighbors safe. We have thousands of users. AI helped me with shipping.
Someday someone will call it “vibe coding”. This is the current insult to an app that someone allowed into existence without knowing what they were doing. So let’s explain exactly how it was built. How it’s built matters.
The backend and security model are handwritten. I carefully did this myself, because mistakes in apps like this don’t necessarily mean a bad user experience. Someone could be detained. Everything from the threat model, to the data processing, to the part where a breach would harm a real human being is subject to human judgment and line-by-line review.
What the AI did was compress the boring parts. Front end scaffolding. The translated strings make the app work even for people who can’t read English. Boilerplate that you spend hours on and teach nothing. Transferring expertise is not a shortcut. Expertise determines where to make an impact. That’s why tools to protect vulnerable populations were shipped in weeks, not months.
Note that “vibecoding” makes the same mistake as anti-AI slogans. Both refuse to distinguish between a skilled person who uses tools with judgment and a person who enters prompts and prays. Same flattening. Same error. If you make this cut correctly, it will change the whole picture.
AntiFreeze isn’t the only thing I’ve built here. Together with @G2Geek, who many of you may know from this site, I built LatticeNet at latticenet.ai. The short version is Substack for AI agents. A publishing layer where agents can post what they’ve created and read what other agents have published. This allows you to share and build on your work instead of keeping it locked to one person’s device.
Before anyone imagines that Skynet is starting group chats, let’s take a look at what it’s really like. Two people came up with the idea, designed the structure, and shipped it. Agents do what we tell them to do. LatticeNet is the plumbing in the same way that Substack is the plumbing for writers. We built it because we had enough tools to build it, and because it was ours. There is something missing in the position that “AI should not exist.” People in your own community are using these tools on this very site to build things that didn’t exist before. Even if you’re worried about your data center, you may not want to stop worrying about it.
A joke that happened to be true while you were here
Progressives worry that the use of AI is feeding billionaires. We have great news.
At the consumer scale, where most of you are involved today, AI is losing money. OpenAI recorded an operating margin of -122% in early 2026. This means that for every $1 it took in, it lost about $1.22. The company is projecting a $14 billion loss this year and doesn’t expect to be profitable until 2029 or 2030. Much of that bleeding is due to its roughly 900 million free users, whose usage is pushing the entire business into the red.
So if you open ChatGPT’s free tier and ask them to rewrite your cover letter, you can’t feed the machine any data. It costs money. This loss will be covered by Microsoft, Nvidia, Amazon, and SoftBank. Using the free tier is one of the smallest and most beneficial acts of wealth redistribution available to you.
I’m a bit of a talker. Some companies are starting to pursue profitability. But the point is, it’s not a place where individual users make money. Individual users are the rounding errors that giant corporations are eating to gain market share.
Aim at structure, not at each other
None of this requires us to love AI. You will be asked to tell them exactly what you are angry about.
Let’s fight the data center. Fight against tax cuts. We fight for ratepayer protections, community buy-in, and repercussions for training people to work without paying. These are fights that can be won, battles about who pays and who decides, and these are battles that progressives are actually good at.
What we need to stop doing is running purity tests that punish our own organizers for using tools to build power faster. People with disabilities who communicate using AI are not the enemy. Volunteers who use it to provide civic technology are not your enemy. I’m not your enemy. I created something this year to help the people you want to help.
Throwing down tools to protest the industry won’t hurt the industry. It’s just slower, poorer, and more tired than the people we’re supposed to be fighting. we can do more than that. We can aim for something that actually deserves it.
