Jarkko Sakkinen
jarkko
These older Joel's writeups are not really followed these days:
- https://www.joelonsoftware.com/2001/04/21/dont-let-architecture-astronauts-scare-you/
- https://www.joelonsoftware.com/2000/04/06/things-you-should-never-do-part-i/
It's like stuff was engineered to be so complex that nobody will want to touch it without agents, which would make agents "fake productive" development tool.
I believe that more be research what has happened, the more we will understand the net gain of agents, and this research should include metrics for system complexity.
All claims what have been done by ANYONE can be considered believes not knowledge, when it comes to productivity.
I believe that it is convenience when being serious at all about software. And I've never enable auto-complete in my text editor either because I need space to think - not shit to happen.
- https://www.joelonsoftware.com/2001/04/21/dont-let-architecture-astronauts-scare-you/
- https://www.joelonsoftware.com/2000/04/06/things-you-should-never-do-part-i/
It's like stuff was engineered to be so complex that nobody will want to touch it without agents, which would make agents "fake productive" development tool.
I believe that more be research what has happened, the more we will understand the net gain of agents, and this research should include metrics for system complexity.
All claims what have been done by ANYONE can be considered believes not knowledge, when it comes to productivity.
I believe that it is convenience when being serious at all about software. And I've never enable auto-complete in my text editor either because I need space to think - not shit to happen.
0
0
0
Money involved is most likely at the center. E.g., NVIDIA's return of investment in its plain and raw form with brutal honesty is in fact a figure that I have zero idea of and I don't even know how I would nail that number with public infromation.
1
0
0
Let's imagine we would model human brain just for the sake of an example. It would be still a finite model. And a finite model can be modeled as a finite state machine. This is at the heart how difficult problem we are talking about here.
1
0
0
@liiwi This is still in the area of "unknown unknowns" but we do know already something about spatial models. It comes down to limited aspects of physical reality. If you have spatial understanding in a system that comes at the price of "LLM skills". It would be hard to envision a reality where we would not be having such a "size-fit" problem given not have unlimited resources in this physical reality :-) I don't believe that even in University of Berkeley they would have a magic potion to outsmart these constraints.
1
0
1
There's no AGI coming, "dangerous" LLM that might "escape from the lab and enslave the humandkind". There's just a chatbot with a limited applicability i.e., being a fancy search engine. Personally, it drills my brain equally if I have to hate this type of thing as much it would if I had to love it, or enforcibly increase my use of AI. I don't have any feelings e.g., towards my vacuum cleaner. Either of standingpoints would be an artificial value increase.
1
1
0
Jarkko Sakkinen
jarkko
Edited 26 days ago
The thing with AI and why this IPO feels just "virtual reality" is that AI is migrating to home, as far as consumers are concerned. There's a limited time window to suck the value from proprietary technology.
OpenAI and Anthropic are in pretty much in same situation as Silicon Graphics was in the late 90s.
And it's uncomfortably short distance away, mostly at this point constrained by semiconductor market price, not "forces of the nature". I.e., we are in situation where SaaS based model of providing inference for customers is artificially maitained to be the best option for most of the end users.
This is why trillion evaluations absolutely are out of this world.
OpenAI and Anthropic are in pretty much in same situation as Silicon Graphics was in the late 90s.
And it's uncomfortably short distance away, mostly at this point constrained by semiconductor market price, not "forces of the nature". I.e., we are in situation where SaaS based model of providing inference for customers is artificially maitained to be the best option for most of the end users.
This is why trillion evaluations absolutely are out of this world.
1
0
1
Jarkko Sakkinen
jarkko
Edited 27 days ago
I'm doing all sorts of non-vibecoding sec contributions to keep vibecoders safe ;-) Or at least trying to.
I answered for the first time in my life to a bot:
https://github.com/anomalyco/opencode/issues/2242#issuecomment-4631187152
I do respect the team behind OpenCode. This tells more about code quality in AI industry than Anomaly per se but it has a well-crafted and flicker free TUI engine - the only instance of such in that industry. It is written in a non-scripting language (does not matter but it is Zig), and the boilerplate is in TypeScript.
#opencode #anomaly
I answered for the first time in my life to a bot:
https://github.com/anomalyco/opencode/issues/2242#issuecomment-4631187152
I do respect the team behind OpenCode. This tells more about code quality in AI industry than Anomaly per se but it has a well-crafted and flicker free TUI engine - the only instance of such in that industry. It is written in a non-scripting language (does not matter but it is Zig), and the boilerplate is in TypeScript.
#opencode #anomaly
0
0
0
0
0
0
Jarkko Sakkinen
jarkko
@jani absolutely not! everyone uses contrasts it's a more complex... let's not go to this. but no :D absolutely not and that post did no trigger me. had to just open my crazy thoughts why i thought it was a joke lol.
i'm allergic to claude code, literally. and it enforces me to stay away from sites from which you'd better be away in the frist place :D i guess it's overall on plus side illness...
i'm allergic to claude code, literally. and it enforces me to stay away from sites from which you'd better be away in the frist place :D i guess it's overall on plus side illness...
1
0
1
Jarkko Sakkinen
jarkko
@jani@floss.sociaI I don't have Claude Code myself but somehow I still recognize its handwriting style at instant and it is super disturbing. There's commonalities in text structuring, storytelling and the use contrasts. I've never spent time on leaning any of this but unfortunately my brain has learned its writing style.
And I'm not talking about your text it does not drill my brain.
Sounds insane but I guess there is enough statistical repetition that you might learn recognize something like this. Had to fully logout LinkedIn because of this. That site is not literally driving me insane :-) And I heard from someone that they use a lot synthesized data to train the models, so perhaps that could create "plastic wrapped" feel that the text has.
Lol, I'm suffering side-effects of others use of Claude Code... It's easy not to use it, so I guess this has benefits too. I had it just a short while provided by my employer but within two three weeks I asked to remove my account on basis that I didn't think this does any good on me.
Yeah, I read a lot of books from time to time so maybe that "helps" (it's a curse IMHO so much of that text pollution).
And I'm not talking about your text it does not drill my brain.
Sounds insane but I guess there is enough statistical repetition that you might learn recognize something like this. Had to fully logout LinkedIn because of this. That site is not literally driving me insane :-) And I heard from someone that they use a lot synthesized data to train the models, so perhaps that could create "plastic wrapped" feel that the text has.
Lol, I'm suffering side-effects of others use of Claude Code... It's easy not to use it, so I guess this has benefits too. I had it just a short while provided by my employer but within two three weeks I asked to remove my account on basis that I didn't think this does any good on me.
Yeah, I read a lot of books from time to time so maybe that "helps" (it's a curse IMHO so much of that text pollution).
1
0
0