AndyNemmity

5 ms·

31.
by AndyNemmity 5mo ago
Yeah, I think we're in a phase honestly where you shouldn't use anyone elses skills, and you should instead point your stuff at a repo with skills, have it really read it, and then ask what of value there is to potentially rewrite
32.
by AndyNemmity 5mo ago
I run a lot of a/b testing. But I'm not sure showing it actually communicates all that much. Since these are non deterministic systems, even showing you an a/b test from when i made the decision a month ago, doesn't real
33.
by AndyNemmity 5mo ago
Define obviously validation? What is the signal that tells you one is reasonable vs another? I find the only way to do that is to look at it, if it passes some visual tests, try it, and then a/b test if it's any better than withou
34.
by AndyNemmity 5mo ago
I watched all the Alphago games live, I've watched analysis of so many Alphago games. I think one of the particulars about Go is how hard the player base took it. Far harder than chess did. Far harder than Starcraft did (although argua
35.
55 Hours of Codex /Goal: What a Port Task Teaches You About Autonomous Loops(vexjoy.com)
2 points by AndyNemmity 5mo ago|0 comments
36.
by AndyNemmity 5mo ago
For sure, this is the pattern I use. And I wish I could make even more deterministic. Maybe I can, but it can also be a bit challenging to sort.
37.
by AndyNemmity 5mo ago
Yeah, that's how I do skills. If I can make a script, I do. Everything that can be deterministic should be. https://github.com/notque/vexjoy-agent
38.
by AndyNemmity 5mo ago
That's what my system does. It uses a workflow if one already exists, if not, it just creates one on the fly from the primitives. https://github.com/notque/vexjoy-agent I would prefer that be deterministic though.
39.
by AndyNemmity 5mo ago
Yeah, I Blind A/B test everything, and a lot. But I don't expect anyone to every use my stuff. It's complicated as hell. But it's for me, and it works without me having to remotely think about the complexity. I love that
40.
by AndyNemmity 5mo ago
This is why I created the /do router, to route to all skills. I also have anti rationalization, progressive context discovery etc. I only make it for me, so it's a bit complex and targeted towards me, and what I do, but it's
41.
Richard Dawkins (AI) Refutes Richard Dawkins (Human) on AI Conciousness(vexjoy.com)
3 points by AndyNemmity 5mo ago|0 comments
42.
by AndyNemmity 5mo ago
An LLM cannot be conscious. A submarine cannot swim.
43.
by AndyNemmity 5mo ago
Good for you, but I'm not sure what you're showing! :)
44.
Cloud Skills Are Still Just Skills(vexjoy.com)
1 points by AndyNemmity 5mo ago|1 comments
45.
by AndyNemmity 5mo ago
Wrote this on Claude Code's Ultraplan and Ultra review, including the newly released Claude Security. It's a departure for Anthropic releasing skills to us to use, versus gating them and hiding what they do. It feels important to
46.
by AndyNemmity 6mo ago
There is no evidence of any of that. He was paid to work on it. That stopped, he continued to work on it in the hopes he could find someone who would hire him to work on it. That wasn’t true, no one has funded it. So due to the economic sys
47.
by AndyNemmity 6mo ago
Why is it the responsibility of the person working for free? Why is it never the responsibility of the people using it? If anyone cares enough they will. People didn’t care enough to pay, so maybe no one cares enough to fork and be the new
48.
by AndyNemmity 6mo ago
Well said, accurate framing.
49.
by AndyNemmity 6mo ago
The reason for something to exist is not to be used. He was paid while doing it, and that pay stopped, and he kept doing it. Now he wishes to stop. The reason for something to exist is someone finds joy doing it. Especially when they are un
50.
by AndyNemmity 6mo ago
It can still be forked. There is no salting the ground here. If you maintain the project and have for a long time, and you wish to stop, you can stop. If no one cared enough to support the project, why does anyone care enough now? It all so
51.
by AndyNemmity 6mo ago
Skills are not ignored if you use a router in front of them, and they are actually called. The problem is the base harnesses don't call them aggressively enough. Not that they don't work.
52.
by AndyNemmity 6mo ago
I mean, skills also include calling python scripts. That's determinism. Anything that can be deterministic, should be
53.
by AndyNemmity 6mo ago
I'm actually in the process of switching all of my agents to sonnet, and going to try to drop down to medium. I used up 1/3rd of my context in less than a day. I am working diligently to do whatever I can to lower token usage.
54.
by AndyNemmity 6mo ago
Sorry, in that case I misunderstood max to mean the subscription, max 20. I am on xhigh.
55.
by AndyNemmity 6mo ago
The same thing I've been doing all the time, now has used up 1/3rd of my week in one day on max20. So yes, for the same tasks, usage runs out faster (currently)
56.
by AndyNemmity 6mo ago
It's unsurprising when this is the first day that tokens have been crazy like this. All of us doing crazy agentic stuff were fine on max before this. Now with Opus 4.7, we're no longer fine, and troubleshooting, and working throug
57.
by AndyNemmity 6mo ago
I'm at 35% :(
58.
Adaptive Thinking Doesn't Change Quality. It Changes Variance(vexjoy.com)
1 points by AndyNemmity 6mo ago|0 comments
59.
by AndyNemmity 6mo ago
The biggest issue I have with these systems is, I don't want a blanket memory. I want everything to be embedded in skills and progressively discovered when they are required. I've been playing around with doing that with a cron jo
60.
Tell the Model What to Do, Not What to Avoid(vexjoy.com)
1 points by AndyNemmity 7mo ago|0 comments

More ›