Kaden Cartwright

The reason LLM’s are so good at generating code is in part because code is roughly English like. Also, the text generation is what powers the reasoning aspect of reasoning models. So not technically feasible. An alternative way to reframe that discussion would be to ask about the removal of chat interfaces. Many of the Christian humanist types have particular issue with the “deceptive” anthropomorphism of the ubiquitous chat based interface

@camoore1979 @suzania Code doesn't have to be readable text. It could spit out assembly language.

In my experience the major predictive factors are: 1) are you a stem person, especially any kind of tech or engineering? 2) are you low church/non-sacramental? If the answer to either of those two is Yes you’re more likely to be bullish on AI

@annielcrawford @suzania Everyone is ideological, no matter where you are on the spectrum. The real question is, who’s correct?

@suzania The low church tech guys don’t realize that they are ideological.

The ontological distinction between divine and human words we see implied in 2 Tim 3:16 is an essential feature of a reformed, i.e. word centered, piety. The way this grounds reformed sacramentology is actually one of my chief reasons for being confident that the AI doomers are wrong. Creative production such as writing or visual art can, in a congruent form to the sacraments, be a way to partake of God. That is to say, only mediated through faith, which is an inherently word centered faculty. The derivative implication of this is that I think you’re right, and those who value the form over the substance of communion with God get things tragically out of order. Don’t get me wrong, I like a pretty church building, and I think the undervaluing of aesthetics is an error of our time, but I think those using this as a critique would have their own arguments more fully undercut by the implications of their own sacramentology

I’m a baptist, and while some get mocked for worshipping in the equivalent of a retrofitted Pizza Hut, the "high church-low church" framing is often just a pretentious, urban intellectual Kellerite smokescreen used to obscure the primacy of one’s actual doctrinal commitments.

@crwiley1962 What constitutes a Christian “cyber-head”?

Yeah, but will Anthropic consider the possibility of blasphemy? I predict the cyber-heads in Christianity will go full transhumanist in a generation. Yet another split is coming.

I’m sure you’ve noticed this at a macro scale, but theological eclecticism is incredibly in vogue right now. John Mark comer’s medieval mysticism and Francis chan’s Eucharistic fascination are some of the bellwethers for this phenomenon. Jonathan pageau seems to be at the orbital center of many of the Christian AI doomer talk I see on twitter. Lots of them reference his symbolic world podcast. As an Eastern Orthodox icon carver by trade, he has what I would consider a very unbiblical sacramentology. It makes sense that he would be a thought leader among those who have a tendency to over spiritualize art and work. Back to eclecticism - lots of evangelicals are swallowing pageau’s sacramentology hook line and sinker. This seems to be a common thread in the twitter AI conversations

@kadenbc Can you explain that to me? I’m not familiar.

The most alarming thing about the AI debate is how many people believe man can breathe life into our creations.

T3 Code just got better remote support. Use it to connect your phone to your desktop app, or connect to an external machine anywhere in the world.

As long as the llm actually reads the source the summary is usually solid. They don’t always search the web, but still confidently spout nonsense as if they did. Not as common today as it used to be, but still something to watch for A flow like you describe with structured inputs typically works really well though

@kadenbc @DrJordanBCooper I started using Claude to read youtube transcripts. A great way to go through a hour long video and distill the important things to a 10 minute read

If I see one more comment that says, "I was doing some research on ChatGPT, and..." I am going to lose my mind.

@odd_joel @broady @theo Very interesting! I’ll have to give these a try soon

@kadenbc @broady @theo the whisperflow background app thing is a pain. been using Moshi — on-device Whisper so voice works without switching apps, and Mosh protocol instead of SSH means no more reconnecting when your phone sleeps

I want to control my agents from my phone. I'm sure you do too. What workflows do you want? Do you have one that works now? Tell me all about it.

@tpauls @DrJordanBCooper I got half way through that flow and was about to abort, then they offered me 50% off to retain me. You might give it a try if that makes it worth it for you. The eleven labs stuff is excellent as well! I’ll probably switch to that when my half price year is up

@kadenbc @DrJordanBCooper I need a free easy to use TTS option. I'm too cheap for Speechify.

Several things about the terminal in t3code are janky on mobile. I usually only use it to launch tmux in the worktree then connect via ssh and do terminal stuff from Termius. Whisperflow is still not a great experience on iOS because it has to open the app in the background to work. There are often times weird states that the chat input + stop/send button gets into as well

@kadenbc @theo what's not ideal about this?

Out of curiosity, when was this? I personally avoided coding agents for a very long time because of the slop factor. I definitely believed they were a net negative on productivity. I could never use cursor or copilot because I got neovim-pilled in 2022 and can’t stand the experience of VS Code. I do believe I was genuinely faster in neovim than most people using cursor because modal editors are just better from an efficiency perspective. It wasn’t until December of this past year that these tech came together enough to break that previous mental model for me. A side note, I think it may be muddying the waters to talk about if something is needed or not, especially in the tech world. If a tool brings more negatives than positives, then by all means we should reject it. But the converse is true also. There are all sorts of tools I wouldn’t want to touch with a 39.5 foot pole, that we managed just fine with as an industry for a long time. For example, I’m never going to choose to write Perl, even though there was a time when that’s just what you did for the web.

I didn’t mean to imply that in my comments above. I was speaking more about the fact that big data analytics doesn’t require the recent AI add-ons. Coding agents aren’t needed for all big data projects. I ran big data projects for decades without them. But in terms of net negative on productivity, that depends on what you are doing. In my professional experience, the agents we used to translate text to code and to automate the code writing caused a lot more work for us and was a productivity killer. We gained some efficiencies, but because the results were often unreliable (because the generative nature made the agents continually make stuff up that broke our code), using coding agents make our QA/QC process much longer and more difficult.

@MattInformed Are you implying that using a coding agent is a net negative on productivity/impact for big data projects?

Something to keep in mind as we consider “AI;” the LLM and most of the underlying capabilities are available without it needing to be presented to us as a personified “intelligence,” or as a means for mass replacement of the work of man. The packaging exposes the ideological (and deceptive) nature of all this. Really. If they want to cure cancer with “AI” and think it’s possible, they don’t need Grok or GPT or Claud or Gemini etc. They can digitize, sequence and analyze large sets of data without “AI.”

@JoshDaws @theo Low effort solution would be Tailscale and bind your dev server to that nic, that’s what u do currently. You could probably wrap a Cloudflare tunnel create script in a custom tool and get even farther

@theo Whatever workflow you come up with, I just want the ability to open whatever it’s building on my phone. My remote work always grinds to a halt when I get: “It’s running at: http://localhost:3000”

@annielcrawford Eh, give it a few years

Total insane nonsense.

@bendellwerry Very interested

Building a group for right wing guys who are positive about AI and want to network / discuss ideas. We’re looking to add men who are using the tools already and building things. Not a meme, politics, or religious discussion group. Reply and I’ll add you as I have time.

@dokimazete I would love to know this too. There definitely seems to be a disproportionate number of anti AI Christians in the broadly sacramentalist traditions

Sincere question: Are there any normal-ish Anglicans on here that are pro-AI?

The biggest issue for me is how you get from “transhumanism is bad” (which I agree with) to “chat based LLM use necessarily involves participating in the moral evils of transhumanism” Maybe this isn’t exactly the point you’re trying to make, but this is what I interpret you to mean at this point

@MattInformed @PractiveSec This was the ai revolution workshop. Looks like you dropped a new video just after I watched that, I hadn’t seen that yet

I’ve often explained that I have used AI professionally for a long, long time. But AI isn’t just technology. It’s not just a tool. The technology underneath is, but AI is a concept - an ideology. That sets it apart from the tech as “simply a tool.” My objection today is rooted largely in the revolution and how the ideology has repurposed the technology toward ends that are harmful. Those ideological roots involve dark enchantment - so it’s very hard for me to set that aside. Today, it’s very hard to separate the tech from the ideology and the Revolution because they are all intertwined now. 3 years ago that was a totally different case. Would I use it again professionally? Maybe. Under certain circumstances that would need to align with my values - and probably only for information processing by ML. I would try to not use the natural language processor as an interface, though I understand why it’s helpful. I would never use it to replace people or undermine others. It would really depend on the context, the outcome, and the impact on me and others. I wrote an ethics paper on that. Do I use it personally? Not by choice, except for research related to my work - to test it and stay current with features so I can write about it. Do I use AI powered services and tools? Of course. You almost can’t avoid it, just as you can’t avoid QR codes to access the menu at local restaurants. But I opt-out when I can. Will I ever choose to use AI personally? Dunno. Maybe. No plans to. In fact, when given the choice, I intentionally avoid it. I am intentionally steering my family away from AI. I’m setting us to the effort of preserving the old ways. I do not want the effects of AI use to transform us - or maybe I should say, deform us. But that’s us. That’s our mission. Do I think businesses should not use AI? That entirely depends on the ethics of it all. But never to replace people. That, I believe, is immoral. I am deeply concerned about the dark and deceptive nature of all this, the harm already being caused, and the direction this is leading us. That’s why I’m giving it such attention. I also believe that as Rod Dreher has written, that this is a spiritual war. I do believe the Transhumanists are under spiritual influences that they have opened themselves up to. They have openly admitted this. I believe that puts us all at risk. Does that mean the underlying technology of AI is 100% always bad? I’m honestly not sure. I could probably make the case that it is, but there are reasons that I don’t. Would I end this Revolution tomorrow if I had the power to? Yes. 1000% yes.