[TUHS] A thing I did (AI Slop Alert)

[Will Senn via TUHS]

Matt et. al.

I'm a fan, wannabe expert, skeptic, and something of a Luddite all 
rolled into one.

I study machine knowing, so the current generative systems are 
fascinating to me as objects in their own right. Language is deeply 
implicated in understanding, and suddenly we have inexpensive machines 
whose principal interface is conversation. I find that hard to look away 
from.

At the same time, Matt's concern resonates with me. I dislike 
technological dependence. I want the books, the source, the local files, 
the machine I can understand, and a reasonable chance of surviving the 
zombie apocalypse. Nobody becomes lesser by declining a tool, and nobody 
owes a private company dependence.

Still, I understand the narrower truth behind "embrace it or get run 
over." In some kinds of work, a person who learns to use these systems 
well can operate at a radically different scale. That is leverage, not 
superiority.

For me the leverage is unusually strong because I have aphantasia. I 
don't have an internal visual surface on which I can hold a complex 
structure and inspect it. I tend to know things configurationally: 
relation, fit, pressure, direction, what belongs and what doesn't. So I 
have always depended heavily on externalized cognition... writing, code, 
notes, diagrams, conversation, to give structure a surface I can return to.

Generative AI makes that surface active.

That's why I sometimes call it my Hawking's chair. The chair did not 
think for Hawking. It provided an external mechanism through which 
thought could become expressible and workable. With AI I can externalize 
a half-formed structure, get it back transformed almost immediately, 
inspect the transformation, reject it, correct it, or discover something 
in it, and continue.

This has changed how quickly I can work. It has accelerated my 
exploration of V6 enormously, and I have built an entire small 
programming language with AI writing the implementation from my 
architecture, constraints, inspection, testing, and correction. That 
works not because AI is a great programmer - it really, truly isn't, but 
because it is an extraordinarily fast and ego-free mediator.

That distinction is central to my research. I locate the informing act 
in the knowing subject: an encounter matters when apprehension leaves 
the subject differently configured, K != K'. The machine can transform, 
compare, expose, generate, and mediate. The judgment and apprehension 
remain elsewhere.

And oddly enough, that makes my use of AI feel rather UNIX-like... to 
keep us on topic :).

The UNIX lesson I took was never "use every new thing." It was to keep 
boundaries clean, make tools composable, keep durable things in 
inspectable forms, and avoid confusing one component with the system as 
a whole. A tool can be enormously powerful without becoming the place 
where your work lives.

My source remains mine. My notes remain mine. The resulting code has to 
build without the model. The argument has to survive outside the 
conversation. If one mediator disappears tomorrow, another can take its 
place, or I can work without it. The durable substrate is outside the tool.

So I don't think anyone needs to use AI. But neither do I think the 
scaling effect is imaginary. For certain kinds of intellectual work, 
making the external surface of cognition conversational, responsive, and 
nearly instantaneous is a substantial change.

The trick, very much in the UNIX spirit, is to take the leverage without 
surrendering the architecture.

For more on the mediation side of things, here's the most recent work 
I've done:

https://www.emerald.com/jd/article/doi/10.1108/JD-03-2026-0125/1382093/The-informing-act-apprehension-in-AI-mediated

The non-paywalled AAM is here:
https://www.researchgate.net/publication/407280891_The_informing_act_apprehension_in_AI-mediated_systems

Later,

Will


On 8/10/26 11:22 PM, segaloco via TUHS wrote:
> On Monday, August 10th, 2026 at 21:02, Alan Hightower via TUHS<tuhs at tuhs.org> wrote:
>
>> On 2026-08-09 22:45, Adam Thornton via TUHS wrote:
>>
>>> I will say: I probably would never have finished this project without
>>> Claude's assistance
>>>
>>> Adam
>> ^^ This.  AI gets a lot of bad blowback especially from older
>> programmers.  But results speak for themselves.  Today is the worst AI
>> will ever be going forward.  Embrace it or get run over.
>>
>> -A
>>
> Doesn't feel like a helpful attitude to have though.  Nobody is obligated to reorient their entire approach to their life's work because someone invented some new tool in that field.  Eats me up just the same how most of society expects me to own and use an automobile and cellular device.  Why is anyone beholden to adopting some new private product someone else invented?
>
> Not trying to bicker too much but this is a very, very near and dear subject to my heart.  Any looking to debate lets take it offline, but I do feel a need to advocate that people are NEVER obligated to suddenly be beholden to a product or service just because someone else likes it.  Nobody is *owed* the use of AI, and this mentality of "adopt arbitrary invention <xyz> or get left behind" is so dehumanizing that it verges on sociopathy.
>
> UNIX didn't take hold by force, it didn't grow and improve because people were deemed persona non grata for avoiding it.  Sure, there were evangelists and plenty of flame wars over the years, but that sort of negative dismissal is not what made the system strong.  We build each other up, rather than tearing each other down.  We find ways to accommodate rather than banishing disagreement.
>
> Again I don't want to make a habit of turning chit chat here into debate, but this "Someone invented product <x> and you are less of a person for not paying for their invention" really and truly feels antithetical to what the fellowship that has grown around UNIX is all about.  I want to advocate for being careful in all ways that adherence to some invention or technology hurts our real, human connections and support for one another.  Dismissing someone because won't use or dislike AI is just as bad as dismissing them because they do like it.  In both cases, a human connection has been made worse by an inanimate object.  Ultimately, like it or hate it, our relationships shouldn't suffer over it.
>
> End rant, thank you if you read what I have to say.
>
> - Matt G.