Hacker News: Linellohttps://news.ycombinator.com/user?id=LinelloHacker News RSShttps://hnrss.org/hnrss v2.1.1Sat, 05 Sep 2026 08:28:34 +0000<![CDATA[New comment by Linello in "Ask HN: Why were OpenAI, Claude, and Grok simultaneously down?"]]>What about a hard-takeoff scenario of an unleashed OpenAI Astra taking other models down for computational resources control?

]]>Thu, 03 Sep 2026 15:49:49 +0000https://news.ycombinator.com/item?id=49552013Linellohttps://news.ycombinator.com/item?id=49552013https://news.ycombinator.com/item?id=49552013<![CDATA[New comment by Linello in "Elevated Errors for Multiple Models"]]>Cursor Grok models are down too...maybe a datacenter outage?

]]>Thu, 03 Sep 2026 13:54:44 +0000https://news.ycombinator.com/item?id=49549959Linellohttps://news.ycombinator.com/item?id=49549959https://news.ycombinator.com/item?id=49549959<![CDATA[New comment by Linello in "What is it like to be a bat? (1974) [pdf]"]]>It's very likely related to this beautiful article by Katherine Hayles

https://modesofcognition.antikythera.org/

]]>Thu, 11 Jun 2026 09:02:14 +0000https://news.ycombinator.com/item?id=48487979Linellohttps://news.ycombinator.com/item?id=48487979https://news.ycombinator.com/item?id=48487979<![CDATA[New comment by Linello in "Day 1 of ARC-AGI-3"]]>Also think about the program-synthesis approach proposed by Poetiq.ai. python programs are being generated and evaluated against previous examples. Then in-context learning is done programmatically via prompt concatenation. If you can "score" online the working and non working examples, then you have a very strong reward signal.

]]>Fri, 27 Mar 2026 10:47:02 +0000https://news.ycombinator.com/item?id=47541066Linellohttps://news.ycombinator.com/item?id=47541066https://news.ycombinator.com/item?id=47541066<![CDATA[New comment by Linello in "Day 1 of ARC-AGI-3"]]>Scaffolding is all you need. I am absolutely certain about that. It's abound finding good ways to approximate the reward function being used during post-training, but at inference time. A general enough reward that can score candidates well will inevitably improve the abilities of LLMs when put inside scaffolds.

]]>Fri, 27 Mar 2026 10:45:05 +0000https://news.ycombinator.com/item?id=47541059Linellohttps://news.ycombinator.com/item?id=47541059https://news.ycombinator.com/item?id=47541059<![CDATA[New comment by Linello in "Build your own hedge fund with autoresearch-skfolio"]]>I’m one of the maintainers of skfolio, and I wanted to share something I’ve been tinkering with lately. This project was heavily inspired by Andrej Karpathy’s autoresearch pattern. I wanted to see if I could apply that same "loop" to quantitative finance—specifically, using LLM agents to autonomously iterate on portfolio construction and risk strategies. Turns out that GLM-5 improved the deflated Sharpe ratio significantly, hitting scores up to 0.93 in my testing. I've also added a section about how to run Claude Code for free using OpenRouter free-tier models.

]]>Mon, 23 Mar 2026 13:27:59 +0000https://news.ycombinator.com/item?id=47489263Linellohttps://news.ycombinator.com/item?id=47489263https://news.ycombinator.com/item?id=47489263<![CDATA[Build your own hedge fund with autoresearch-skfolio]]>Article URL: https://github.com/carlonicolini/autoresearch-skfolio

Comments URL: https://news.ycombinator.com/item?id=47489262

Points: 2

# Comments: 2

]]>Mon, 23 Mar 2026 13:27:59 +0000https://github.com/carlonicolini/autoresearch-skfolioLinellohttps://news.ycombinator.com/item?id=47489262https://news.ycombinator.com/item?id=47489262