Actually, in professional usage in a technical setting this is my prime objection to heavily LLM driven development. Were the tools in usage deterministic then I'd be a lot less objecting to the mandating of their incorporation into workflows.
I want to be reading, writing, testing and maintaining the software at the same layer. Right now extreme AI usage leads to reading, testing and maintaining happening in a less expressive language than writing and guess which of those four activities developers enjoy the least and find the hardest - it sure isn't writing.
> You're learning to manage idiot savants, which is a very useful skill.
I think the real bifurcation is whether you will settle on that belief.
Some of us are settling on the belief that the idiot savant, lacking the coherence of a functional mind, cannot be managed. It's essentially a chaos agent masquerading as something more cooperative.
Have you looked at the feasibility of making your own CIQ app to push data from the watch to your alternate internet endpoint?
I have the impression there are permissions and APIs to access sensor history and activity records, but I haven't had a need to dig in and learn what restrictions there might be...
Ironically, I use "new" on other sites to bypass their algorithms and see the raw stream. But I've never even had the urge here, nor really remembered it exists.
It's probably because I have always looked at HN as a sort of amusement or curiosity. The curated listing is part of the entertainment. Not because it usually resonates for me, but because it gives me this glimpse into Otherness. The comments can be more resonant, once in a while.
It's a bit like wandering through the bar district and finding your old haunts missing and replaced with weird facades and fashions. But here and there, you notice some patrons are interjecting with stories eerily similar to the one on the tip of your tongue...
I salute your introspection. In my mind, it is better than the alternative cope.
My wife has an ongoing frustration with a colleague who has adopted the mindset, "I reviewed it, so I wrote it". I guess he must sleep well at night, and probably votes in the "AI gives me superpowers bloc", but it is pretty apparent he doesn't really review it much either, because it is full of flaws and absurdity.
Having had experience dealing with people with conventional psychosis, I don't see it as a binary thing. Aside from a full-on psychotic break and full remission, there is a broad gray area. It can be a miasma of reality and non-reality that the sufferer may mask to varying degrees, but which influences their behavior and logic.
So, to me, AI psychosis seems apt to describe the murky areas where people are misapplying AI agents and thinking of them as social entities or suitable to drop into previously human roles, rather than carefully defining appropriate risk management strategies for this new technology.
I adopted Linux in college in 1993 and, like many peers, brought it to my R&D job and observed this wave of expansion through the mid to late 90s. Linux was already "going somewhere" in 2000 for IBM to even notice it. Lots of federal grant money was directly or indirectly improving Linux due to FOSS folks like me.
It was getting so much commercial and academic engagement that we had the idioms (cliches?) of the "LAMP stack" for basic web servers and "Beowulf clusters" for high performance computing. Even SGI was already revealing a Linux plan, before 2000, when they still seemed like a fixture of the HPC industry rather than an also ran.
I apologize for the hyperbole, but you are arguing my point: if something took "lots of federal grant money" to become usable in universities and amount to anything more than a research project, then we are no longer about something "free", are we?
From that point of view nothing that requires human input is free. Which is true in a sense, people are using free to mean free to use, not free to improve.
TANSTAAFL does not need a qualifier to apply. "Nothing is really free, so whatever you got 'for free' from a community member or some non-commercial effort was bound to have limited reach" is more like the point I'm trying to make.
Of course we put labor into it. It's not some seance or wormhole communicating with the software dimension.
This is the way FOSS is meant to work. I got jobs where an employer was happy to run other people's FOSS software "for free", happy for me to contribute bugs/requirements/patches back upstream, and happy to release our own projects under FOSS licenses.
It is a win-win for all involved. That's the whole point of it.
You seem to imply that work on FOSS projects is a second-class activity, meant to be done after companies and employers have secured their revenue sources.
This is like trickle-down economics for FOSS and it doesn't work.
I wouldn't call it second class. Maybe second-order?
To me, it is no different than management, planning, logistics, marketing, etc. which is done for the purpose of supporting some other objective.
It simply means that you perform software development as work-for-hire in support of that other objective, rather than for the purpose of licensing revenue. It provides wages for services rendered, just like the vast majority of other job types.
It just doesn't provide for scalable virtual rent extraction for a "publisher" or other middleman. To me, that is a benefit of it. It removes a bunch of perverse incentives from the table. Incentives that tend to harm the developers and users for the benefit of those middlemen.
In principle, I don't disagree. The problem is that in practice the large corporations managed to neutralize any chance for independent by commoditizing the software service layers that supported their business and invest all their resources they could to package their proprietary solutions on top of it, AWS and "OpenElastic" being the textbook example for it.
The one way to get out of this mess would be to have the market paying a premium for companies that do R&D in FOSS directly. It can not be a secondary goal, and we can not be telling them they shuold find some other way to make a living.
Of course they have to double down on yet another compliance regime. Why not converge on an existng NIST 800-53 baseline, or some HHS "tailored" variant? Or CMMC, if they want to push for more strict certification processes instead?
It's getting absurd with how many different compliance regimes a modern research university will have to follow simultaneously, if they do a broad set of defense, energy, basic sciences, and health research as well as having an attached medical school and teaching hospital.
How insane is it to advocate the usage of these non-deterministic compilers, where each time you compile may produce different semantics?
And then people resort to saving and hand-editing the compiled output.
But when they want to change the source, they recompile and have to start over hand-editing the output again.
reply