amarble

4 ms·

1.
by amarble 1mo ago
I did something similar in March using Opus 4.6 (iirc) on google's "Parameter Golf" challenge, "a challenge to train the best language model that fits in a 16MB artifact and trains in under 10 minutes on 8xH100s, evaluat
2.
Evaluating AI Agents as Products(marble.onl)
1 points by amarble 1mo ago|0 comments
3.
Evaluating AI Agents as Products(marble.onl)
2 points by amarble 1mo ago|0 comments
4.
Updating IP Regulations for AI Distillation(marble.onl)
1 points by amarble 2mo ago|0 comments
5.
Updating IP Regulations for AI Distillation(marble.onl)
2 points by amarble 2mo ago|0 comments
6.
by amarble 2mo ago
They also indicate they have a 276B A12B version, but it doesn't seem the weights are available. This might actually be able to fit in 128GB when quantized to 2 bits or so which makes it interesting.
7.
LLMs are still just low code / no code software(marble.onl)
5 points by amarble 3mo ago|1 comments
8.
LLMs are still just low code / no code software(marble.onl)
2 points by amarble 3mo ago|0 comments
9.
There is minimal downside to switching to open models(marble.onl)
407 points by amarble 3mo ago|330 comments
10.
by amarble 7mo ago
The point of a discussion site is to hear what other people think and get different perspectives. Just getting an LLM's insightful, well-thought-out response isn't really a big draw, if one is looking for that, there's a pret
11.
AI doesn't replace white collar work(marble.onl)
67 points by amarble 7mo ago|104 comments
12.
by amarble 8mo ago
As a counterpoint, I also tried writing something with Claude last weekend: a Google docs clone[1]. I spent $170 on Anthropic API credits, and got something that did mostly what I asked but was basically useless. It seems that for simple in
13.
by amarble 8mo ago
This weekend I tried what I'd call a medium scale agentic coding project[0], following what Anthropic demonstrated last week autonomously building a C-compiler [1]. Bottom line is, it's possible to make demos that look good, but i
14.
I paid $170 and all I got was this demo(marble.onl)
3 points by amarble 8mo ago|1 comments
15.
I paid $170 and all I got was this stupid demo(marble.onl)
1 points by amarble 8mo ago|0 comments
16.
Task-free intelligence testing of LLMs(marble.onl)
69 points by amarble 9mo ago|22 comments
17.
Task-free intelligence testing of LLMs(marble.onl)
1 points by amarble 9mo ago|0 comments
18.
Intelligence is not just about task completion(marble.onl)
1 points by amarble 9mo ago|0 comments
19.
If You Meet ET in Space, Kill Him (2024)(nautil.us)
3 points by amarble 9mo ago|0 comments
20.
Intelligence is not just about task completion(marble.onl)
3 points by amarble 9mo ago|0 comments
21.
Show HN: Gen AI Writing Showdown(writing-showdown.com)
2 points by amarble 9mo ago|0 comments
22.
Ifrro member Kopinor signs agreement on newspaper content for AI in Norway(ifrro.org)
2 points by amarble 9mo ago|0 comments
23.
Comparing language model performance on creative writing transformations(writing-showdown.com)
2 points by amarble 9mo ago|0 comments
24.
by amarble 11mo ago
Since we're sharing, I have a "claude" command that lets me get quick answers but also saves the conversation and outputs an identifier so in the rare case I want a follow-up, I can ask a question with the ID to continue the
25.
Eminembench(marble.onl)
1 points by amarble 1y ago|0 comments
26.
Promptware Attacks Against LLM-Powered Assistants in Production(sites.google.com)
1 points by amarble 1y ago|0 comments
27.
Managing LLM application performance through code standards(marble.onl)
2 points by amarble 1y ago|0 comments
28.
by amarble 1y ago
A linter for LLM application code (including prompts), i.e. programs that do LLM calls, scanning code issues related to security or performance: https://github.com/kereva-dev/kereva-scanner
29.
Catching Claude Cheating(marble.onl)
1 points by amarble 2y ago|0 comments
30.
Catching Claude Cheating(marble.onl)
1 points by amarble 2y ago|1 comments

More ›