I understand where you’re coming from, but I think there might be some misconceptions about the resource requirements. You can actually host LLMs on a local computer without needing a $10,000 GPU. For example, it’s possible to self-host the full Deepseek model on a $2000 setup and open it to your organization for browser-based use, or smaller models on a 400$ GPU.

I also find it compelling that LLMs like Deepseek are designed to be very efficient in their cloud versions, especially when compared to Western tech that isn’t incentivized to prioritize environmental concerns because there are no mechanisms in place to force them to care about the environment. This (the fact that capitalism won’t save the environment) is a much stronger argument than a blanket “no datacenters,” since a datacenter is powering Lemmygrad as we speak. To put it in perspective, China has about 450 datacenters while the US has over 4000, yet their tech sector is just as advanced. It shows there are different, more efficient ways of doing things that we (the state) can tap into if we only wanted to.

This also seems like it could erode trust in Communist organizations

To be perfectly honest, I think you overestimate the existing level of trust the general masses have in communist organizations.

I’m coming from a place of wanting our movements to succeed globally, it’s just that it worries me when I see us hesitating to adopt tools that could give us a real edge. We already use technology, including the internet and automated stuff in our organizational work. I believe we need to move past a certain hesitation toward new tech (a sort of “return to Pravda” mindset) and embrace whatever makes our praxis more effective. We don’t have the luxury of refusing efficient tools. Looking at how China integrates technology provides a practical, existing blueprint for this.

I’ve often seen proposals to automate tasks or improve efficiency in orgs get shut down with responses like, “Oh, that sounds complicated,” or “I like the way we do things already.” But we have to try new things if we want to close the gap. I’d be happy to help build out a tech stack if given the chance! And yet many still prefer to rely on manual email lists when a simple Telegram channel could coordinate communication.

It’s a bit like how the MIA gained its foothold in the 90s while other communists were still debating whether the internet was a fad. We got shown up by trots!

Just recently, we launched a Telegram broadcast channel with ProleWiki to share news. It’s only the first week, and we already have 80 subscribers. That’s 80 people we can reach directly, without being subject to algorithmic filters. The bot for the channel was coded by our dev with some LLM assistance, it uses RSS feeds and custom filters to select the headlines we want and posts them automatically on a schedule. Eventually, we might use something like Deepseek to scrape sites that no longer offer RSS, and maybe even analyze the articles for relevance before posting. At this moment the channel runs automatically, it requires literally 0 labor to sustain. I’m not aware of any org that have a low-stakes, public-facing point of entry like this. They seem to assume that the more labor they put into something the greater its impact and this results in a lot of wasted effort. This automated approach lets us maintain a presence with 0 effort while freeing up energy for other things we want to work on. It’s essentially self-sustaining, I mean, how cool is that!

perpetuate a surveillance state

I mean by many metrics China would be considered a surveillance state (and not just liberal metrics). They have a different cultural and legal approach to online privacy and device security, in fact researchers that work in China like that accessing data, even medical data, is more straightforward there. Our distrust should be directed at capitalism, not the ‘surveillance’ itself.