marsh_mellow

7 ms·

1.
NanoClaw solves one of OpenClaw's biggest security issues(venturebeat.com)
46 points by marsh_mellow 7mo ago|28 comments
2.
Benchmarking GPT-5 on 400 real-world code reviews(qodo.ai)
72 points by marsh_mellow 1y ago|80 comments
3.
Kubernetes Will Solve YAML Headaches with Kyaml(thenewstack.io)
4 points by marsh_mellow 1y ago|0 comments
4.
Matter vs. Force: Why There Are Two Types of Particles(quantamagazine.org)
2 points by marsh_mellow 1y ago|0 comments
5.
Improving Deep Learning with a Little Help from Physics(quantamagazine.org)
10 points by marsh_mellow 1y ago|0 comments
6.
by marsh_mellow 1y ago
p-value of 7.9% — so very close to statistical significance. the p-value for GPT-4.1 having a win rate of at least 49% is 4.92%, so we can say conclusively that GPT-4.1 is at least (essentially) evenly matched with Claude Sonnet 3.7, if not
7.
by marsh_mellow 1y ago
I don't think the absolute score means much — judge models have a tendency to score around 7/10 lol 55% vs. 45% equates to about a 36 point difference in ELO. in chess that would be two players in the same league but one with a cl
8.
by marsh_mellow 1y ago
Good point. They said they validated the results by testing with other models (including Claude), as well as with manual sanity checks. 55% to 45% definitely isn't a blowout but it is meaningful — in terms of ELO it equates to about a
9.
by marsh_mellow 1y ago
From OpenAI's announcement: > Qodo tested GPT‑4.1 head-to-head against Claude Sonnet 3.7 on generating high-quality code reviews from GitHub pull requests. Across 200 real-world pull requests with the same prompts and conditions, th
10.
Determining IaC ownership – a tag-based approach(token.security)
5 points by marsh_mellow 1y ago|6 comments
11.
Expanded language support in Amazon Q Developer(aws.amazon.com)
1 points by marsh_mellow 1y ago|0 comments
12.
by marsh_mellow 2y ago
Very cool! Could this work for detecting nearby drones?
13.
Computer Use API Documentation(docs.anthropic.com)
13 points by marsh_mellow 2y ago|2 comments
14.
by marsh_mellow 2y ago
Anthropic blog post outlining the research process: https://www.anthropic.com/news/developing-computer-use Computer use API documentation: https://docs.anthropic.com/en/docs/build-with-claude&
15.
The U.S. AI Safety Institute stands on shaky ground(techcrunch.com)
2 points by marsh_mellow 2y ago|0 comments
16.
Developing a Computer Use Model(anthropic.com)
3 points by marsh_mellow 2y ago|1 comments
17.
AI Agents: A Comprehensive Introduction for Developers(thenewstack.io)
1 points by marsh_mellow 2y ago|0 comments
18.
How We Generated Millions of Content Annotations(engineering.atspotify.com)
3 points by marsh_mellow 2y ago|0 comments
19.
Open-source tool for LLM document processing pipelines from UC Berkeley(github.com)
1 points by marsh_mellow 2y ago|0 comments
20.
by marsh_mellow 2y ago
To tag on to this, what are the most useful capabilities besides code generation?
21.
by marsh_mellow 2y ago
This is great. Is there any work being done to make something similar part of the browser API?
22.
by marsh_mellow 2y ago
There's an open source version of this as well: https://github.com/Codium-ai/pr-agent
23.
PR-Agent — extension that adds AI chat to code reviews on GitHub(chromewebstore.google.com)
24 points by marsh_mellow 2y ago|15 comments
24.
by marsh_mellow 2y ago
They list seven different use cases in this technical blog: https://harrison.ai/news/reimagining-medical-ai-with-the-mos... I'd interpret it as a foundation model in the radiology domain
25.
US Government Sets Out to Improve Internet Routing Security(infosecurity-magazine.com)
41 points by marsh_mellow 2y ago|8 comments
26.
Pinot for Low-Latency Offline Table Analytics(uber.com)
1 points by marsh_mellow 2y ago|0 comments