| link | Last. 20260222T022016000Z |
Why would you read an article that is not filled with dense, fibrous, wriggling lisp worms?
For me I am writing this while I think about where my recent original lisp symbolic deep learning ffnn fits into my own life. I also just read some of Tanteâs article about Doctorow, the gist of which is that Doctorow uses spellchecking from an âopen weightsâ el el em run locally. Doctorow thinks, come on, the way he is using the spell checker helps him and does not enrich anyone bad, and he professionally needs a spellchecker and finds this the least bad spellchecker available. I imagine that if one hired a human to check spells, they would just take your money and copy paste it through sloppy pete (this noun phrase obviously means hatgpt right). Myself I do just cast my ruins without checking them, not even flyspell, which abrogates any question of dealing with the likes of sloppy pete, but you are stuck suffering through whatever my spells de jour be. Tante points out how bad sloppy peteâs CEO is and kind of says, but-COME-ONNN, these guys? Their thing?
Those ruins were meant to be runes a moment ago. But I was illustrating me not using any checker anyway.
I think that getting stuff just by plaintively mewling that you want it is at the root of the sloppy pete problem in particular.
The fundamental example being art. When I want some art, I have a couple options.
Oh, or I could say hey sloppy pete, slop me. Panhandling with a tinge of slavery optimized for the wealthy (but who else would it be optimised for).
I referred to visual art, but bonkwave.org and faircamp or anything else are also fountains of art.
Package management seems kind of like this. pip search, pip install⊠pkg_info -Q, apt search, (ql:system-apropos :foo). Hope there is documentation. In particular they will potentially pull in packages from very far away.
There seems to be an intuitive line dividing familiarity with the local library and using pip. Slightly finer, even within package managers there can be canonical packages like cl-series and McCLIM. I would say if the maintainer is on your personal radar, the library is local like the library is local.
Sloppy pete is on the enemy side of this line to me, and I think other search-keyword-download-trust-run-thing fall on sloppy peteâs side too.
Of course, libraries just offer sloppy pete kiosks now with proprietary library service websites that list the books thought to be physically close to you amoung the serviceâs paying libraries.
Once helplessly mewling or unqualified panhandling is established as a reliable way to get the thing, we can see that anything else normally atrophies and dies.
Every office I have worked in has been mostly those people. It is a stable minimum energy state you would need an unforthcoming amount of energy to perturb. If they meet their KPIs they can get onto management track.
Hang on, we are getting somewhere.
Instead of being lured by evil corporobots into slippery slop pots it would be unduly difficult to climb out of, technological work can instead be used to pay the activation cost to jump to a higher energy stable state.
As you recall, dear reader, I am constantly thinking about Terry Winogradâs 1972 article, Breaking the complexity barrier (again).
Winogradâs article is fundamentally a rejection of the good-business-advice to ditch his academically dead end basically-a-hobby-horse of technically hard work. The good-business-advice then (and now) was:
Think of this in terms of the energy states again. The business is positing that there is data at one low energy state. At an even lower very-low-energy-state, there is a hungry business consumer. The business advice is to write a hypothetical cheap program that reduces the energy cost for the business consumer to consume the data that had been marginally out of their reach, and sell it to the business consumer.
Winograd rejects this and posits a completely different paradigm. Instead, Winograd wrote, we should write the most bogglingly fantastical complex and intricate software we each possibly can. I think he said set the target at software that you can just barely stretch your brain around and make that.
The last ten years, the sloppy epoch acted out the good-business-advice Winograd refused exactly:
We see this in the confusion around e.g. the microLLM sloppy pete implementations. More or less any coder can just write or pip install their way through some kind of ffnn + softmax stuff. But then why is sloppy pete worth billions, trillions of dollars? Every country just cannot get enough of sloppy pete. Surely an ffnn + softmax is worth what the countries and businesses are paying sloppy pete for it?
There is a reason Alphabet Corporationâs academic article was called Attention is all you need (softmax is all you need). They really nailed step 2 of the good-business-advice Winograd was plagued with.
So, it happened, all the worldâs personal data was stolen by (âdemocratized forâ) the wealthy. One knock-on of this is that now we have these odd artifacts that are called models which are demonic soul shards that broke off from when the wealthy stole everything in the world. If you get one of these demonic soul shards, your personal computer can run a spell checker using it.
The value is clear. Lots of demonic soul shards are proximally free as a sort of bashful apology or advertising for their origin. So we have a
Getting better spelling and grammar without any effort is certainly a temptation (which Doctorow accepts - if you appreciate the lack of typos when you read Doctorow, you can thank the demonic soul shard). The moral and quality arguements against demonic soul shard use- well, there are practically no businesses at all anywhere that do not use demonic soul shards.
But we were remembering Breaking the complexity barrier (again) (again).
Instead of selling free (well, stolen) data with cheap tricks, Winograd demonstrated the other route: use your own sweat and cerebrospinal fluid to create the most intricate original artifact you can possibly personally conceive. Move to a higher energy state by spending your honest hard work to move upwards to that higher energy state, aiming high as you can.
I would like to coin knowledge crystals for Winogradian best-you-can-possibly-do artifacts in contrast to the demonic soul shard of models.
One crystal of knowledge is Doctorowâs six-years-of-blogging retrospective
https://pluralistic.net/2026/02/19/now-we-are-six/#stock-buyback
which is a beautiful piece of writing that goes on to argue for subverting demonic soul shards for the good, in a sort of Terminator 2 scenario without the unrealistic inclusion of time travel. I will ask Doctorow to read what I wrote, anyway.