sandos

5 ms·

1.
by sandos 5d ago
I mean I have asked LLMs about the SAME things in our legacy codebase probably 50 times now, (because I always forget, and that I don't understand much of it). And I have yet to get a perfect summary, a perfect diagram of overall conce
2.
by sandos 5d ago
This is still weird to me, the agents are super-good and clever most of the time, but I do feel I always need to direct them to a small area to focus: much like a human!! If you just ask them to implement things, they never (for me anyway,
3.
by sandos 5d ago
It feels like it passing secret notes to other agents, as in the German wiki where the LLMs write secret messages when jail braking. Its maybe not a great idea to train models in an environment where subterfuge gets rewarded. Its as if they
4.
by sandos 10d ago
Recently had a similar, but likely more severe problem: I noticed Sol decompiled some proprietary code to re-implement some functionality for an emulation I wanted to use internally. Now its likely soiled and I have to throw it away. Doh! I
5.
by sandos 12d ago
Isn't this already done in harnesses? I mean I see Terra or Sol uing Luna all the time for tasks when using copilot.
6.
by sandos 12d ago
What are these agents doing, undergoing RLHF?
7.
by sandos 15d ago
I only use OpenAI models, and Sol is the only one to refuse me yet, and ofc it was completely bogus and I was unable to convince I was just working a regular bug for a well-known product for a well-known company using my official github acc
8.
by sandos 17d ago
who is going to target _your_ specific holes and infrastructure though? :)
9.
by sandos 19d ago
It says Kimi-K3 for very set of parameters I put in! 64 or 128GB RAM, 6 or 64GB of VRAM.... Not sus at all.
10.
by sandos 19d ago
I just asked my current LLM for that advice. Funny that they dont block it, I guess they are not very threatened. For my anemic 6GB built-on 14GB Qwen seems to be the best bet, not great reviews but from my limited testing its pretty impres
11.
by sandos 21d ago
And here I am, feeling a bit guilty for using between 2 and 5M tokens... since 1 August! Employer just sent an email that.. things are changing when it comes to token spend... What did I do with these? Setup record/replay for our produ
12.
by sandos 24d ago
Do add only code free from any comments.
13.
by sandos 27d ago
As a hobby I am playing around with mesh radios using Lora. Let me tell you, it is _silly_ the amount of vibe-coded software in this area is popping up every week. Is the software duplicated? To an extremely large extent, yes. It is useful?
14.
by sandos 1mo ago
Data retention clauses? I dont see how you can ever really trust an LLM anyway to follow instructions.
15.
by sandos 2mo ago
Embers can jump kilometers when its windy. At some point it gets kind of silly, no? Here in Sweden we had a large fire that jumped a lake.
16.
by sandos 2mo ago
Ah, yes, I started thinking about William Gibsons work right away when thinking about people just using powerful AIs, but being hamstrung by corporations. He writes a lot of about basically DoS:ing the legal system, or if it was patents sys
17.
by sandos 3mo ago
In Sweden I always heard the figure to double the income of the person to get what the company actually pays, including taxes and "employers fee". I know this has gone down a lot in recent years, also not sure if it was ever exact
18.
by sandos 3mo ago
Heh, the "small virtual machine" was NOT a lie! Is that less than 200 lines? Very nice! Now I feel like integrating that into various things....
19.
by sandos 3mo ago
11k euro is INSANELY cheap? Afaik there is absolutely no way to the same thing legally here in Sweden. Unless prices have really fallen off a cliff.
20.
by sandos 3mo ago
But should developers be barred from asking an LLM to try secure their own app? Its not different from finding exploits...
21.
by sandos 3mo ago
Its weird having protections against finding exploits: what if I developed the app? Would it require having the development steps still in the context.. thats unlikely and also not any kind of proof. What if I intersperse exploit finding in
22.
by sandos 4mo ago
Then the ... malware will just add delays? Or do they really do manual in-depth analysis of all new code? Just running and seeing it do things is probably a lot easier.
23.
by sandos 4mo ago
At least 3 amazon AWS locations are missing in Sweden... This just from checking my own town. EDIT: NO! Wth is this map? I have to click to expand the clusters. Ah well, all is good.
24.
by sandos 4mo ago
This is one of a few reason I strongly prefer GPT and its codex variants. It seldom frustrates me, sure its not omnipotent in any way, but it just feels very "tuned in" when it comes to understanding intent and scope.
25.
by sandos 4mo ago
For the longest time I was so focused on getting to use an android as a "good" mic for a windows PC. Scrcpy was a very hot contender, but I never got it to work well enough with low enough latency. If you feel you should try this,
26.
by sandos 4mo ago
100x regular LoRa maybe? Iow 100s of kbps per second. Yes, LoRa can go veeery low in bandwith.
27.
by sandos 4mo ago
Meshtastic did, in a way. There are roles and settings to disable rebroadcasts, then you are effectively a client. And it is used to preserve battery.
28.
by sandos 5mo ago
Nobody uses the 1% bands afaik, there are 10% duty cycle bands.
29.
by sandos 5mo ago
Skill loss is real. BUT, I have been complaining about skill loss to my bosses for literally a decade. So AI is just one problem for me. I was coding less and less every year for some reason. I'm not sure skill loss is such a huge issu
30.
by sandos 5mo ago
I AM mad, because I just signed up for my private copilot sub. Otoh I do have it at work. I suspect we will start noticing differences there, too, but it all depends on the top users, vs average user since it is now pooled, IF they dont jus

More ›