arcanemachiner

8 ms·

1.
by arcanemachiner 4d ago
You don't need human input. Any coherent input will do the trick. You can, for example, hook it up to a logging system and have it fix errors as they occur on your platform.
2.
by arcanemachiner 4d ago
Banks create money when issuing a loan. This is how fractional reserve banking works. They lend money they don't have (most of). This is institutionalized fraud, and it's been standard operating procedure for centuries.
3.
by arcanemachiner 6d ago
You can chain a lot more commands together with this technique than with a single Edit tool call.
4.
by arcanemachiner 6d ago
They're talking about writing a file with a harness-native Edit tool. They're saying the agents aren't doing that, but are using ad-hoc methods of writing the files. (My agents seem to prefer see these days.)
5.
by arcanemachiner 6d ago
> you’d want to be using GLM 5.3 Flash right now for most things agentic That was yesterday. I think the crown currently belongs to DeepSeek Flash v4.1 for the next few days or weeks.
6.
by arcanemachiner 8d ago
Is "normies" finally a mainstream word? Just wondering how out of the loop I am here.
7.
by arcanemachiner 9d ago
Maybe they should rebrand again. It took a few renames to get the stink off of Blackwater (private military contractor).
8.
by arcanemachiner 11d ago
"Show me the incentive, and I will show you the outcome."
9.
by arcanemachiner 13d ago
Must be system prompts and tool instructions guiding the agents differently.
10.
by arcanemachiner 14d ago
You can always use JSONC or JSON5.
11.
by arcanemachiner 15d ago
You have tried Qwen 3.8 27B before coming to this conclusion, I hope? It's an incremental improvement over 3.6, but I mostly want to make sure you didn't just try running some old junker before coming to this conclusion.
12.
by arcanemachiner 16d ago
How would this scam work? Why would someone pose as a locksmith specifically?
13.
by arcanemachiner 16d ago
So was ChatGPT back in 2021.
14.
by arcanemachiner 17d ago
If it makes you feel any better, I have just upvoted every submission made by your account.
15.
by arcanemachiner 17d ago
I thought they just did this? People were using some loophole to use their Chat sessions to power their Codex usage after their Codex quotas had run out.
16.
by arcanemachiner 18d ago
That's kinda how I'm thinking about this in terms of licenzing my upcoming projects. The code is just going to be laundered anyways, so why bother putting on a license like AGPL that might just hinder adoption?
17.
by arcanemachiner 19d ago
Anthropic rents a lot of capacity from xAI, so I don't think they're in a position to do the same.
18.
by arcanemachiner 22d ago
It also costs more to buy the same computer you already have, which is not a great thing.
19.
by arcanemachiner 25d ago
Jeez, I thought I could get away with q8_0 KV cache. Guess not.
20.
by arcanemachiner 25d ago
/model claude-opus-4-6[1m]
21.
by arcanemachiner 26d ago
Ah, I forgot. I thought there was a free version. I haven't used PowerAmp in like a decade.
22.
by arcanemachiner 26d ago
What, so people don't like OTel, but they like Jaeger, which implements an OTel spec? (I'm a noob to this subject, if that wasn't obvious.)
23.
by arcanemachiner 26d ago
So what's the alternative then? (Genuine question, not hypothetical snark.)
24.
by arcanemachiner 26d ago
I'm fairly ignorant as to whether Motorola, as a company, will make a good steward for this kind of arrangement. Does anybody have any inside baseball to share in terms of whether we can trust this arrangement to last longer than, say,
25.
by arcanemachiner 26d ago
The PopSci book you're thinking of is "Thinking, Fast and Slow" by Daniel Kahneman
26.
by arcanemachiner 26d ago
Use Aurora Store to download apps from Play Store without an account (very trustworthy app IMO, has been around and working for years): https://auroraoss.com/aurora-store
27.
by arcanemachiner 27d ago
All of my non-work AI coding is that open-source, so I'm happy to feed my data into the machine. It's a win for me: my code goes into the training data, and my sessions are fed into future training data, making the model stronger
28.
by arcanemachiner 27d ago
Reminds me of this old WKUK skit: https://m.youtube.com/watch?v=KivCRqfFcqY
29.
by arcanemachiner 27d ago
You should out an emdash in there just to mess with people.
30.
by arcanemachiner 28d ago
> back then for a nickel you could go to a paper vending machine and buy a paper. But it would just open up and you could literally grab all of them if you wanted. I miss that world.

More ›