Hacker News: cjamsonhnhttps://news.ycombinator.com/user?id=cjamsonhnHacker News RSShttps://hnrss.org/hnrss v2.1.1Mon, 01 Jun 2026 19:27:08 +0000<![CDATA[Intuitions for Tranformer Circuits]]>Article URL: https://www.connorjdavis.com/p/intuitions-for-transformer-circuits

Comments URL: https://news.ycombinator.com/item?id=47484227

Points: 67

# Comments: 5

]]>Mon, 23 Mar 2026 01:06:35 +0000https://www.connorjdavis.com/p/intuitions-for-transformer-circuitscjamsonhnhttps://news.ycombinator.com/item?id=47484227https://news.ycombinator.com/item?id=47484227<![CDATA[New comment by cjamsonhn in "A Survival Guide to a PhD (2016)"]]>>I was smart enough to realize I'd never be a good researcher relatively early during my PhD and started preparing for job interviews

What are some things that made you realize that? Im considering going into research and certainly have some doubts

]]>Sat, 14 Mar 2026 21:19:13 +0000https://news.ycombinator.com/item?id=47381291cjamsonhnhttps://news.ycombinator.com/item?id=47381291https://news.ycombinator.com/item?id=47381291<![CDATA[Language Modeling, Part 5: Reverse Engineering LSTM Cells]]>Article URL: https://connorjdavis.substack.com/p/language-modeling-part-5-reverse

Comments URL: https://news.ycombinator.com/item?id=46893995

Points: 1

# Comments: 0

]]>Thu, 05 Feb 2026 00:31:32 +0000https://connorjdavis.substack.com/p/language-modeling-part-5-reversecjamsonhnhttps://news.ycombinator.com/item?id=46893995https://news.ycombinator.com/item?id=46893995<![CDATA[Language Modeling Part 4: LSTMs]]>Article URL: https://connorjdavis.substack.com/p/lanugage-modeling-part-4-lstms

Comments URL: https://news.ycombinator.com/item?id=46746034

Points: 3

# Comments: 0

]]>Sat, 24 Jan 2026 18:17:55 +0000https://connorjdavis.substack.com/p/lanugage-modeling-part-4-lstmscjamsonhnhttps://news.ycombinator.com/item?id=46746034https://news.ycombinator.com/item?id=46746034<![CDATA[Langauge Modeling, Part 3: Vanilla RNNs]]>Article URL: https://connorjdavis.substack.com/p/language-modeling-part-3-vanilla

Comments URL: https://news.ycombinator.com/item?id=46653774

Points: 1

# Comments: 0

]]>Fri, 16 Jan 2026 23:52:36 +0000https://connorjdavis.substack.com/p/language-modeling-part-3-vanillacjamsonhnhttps://news.ycombinator.com/item?id=46653774https://news.ycombinator.com/item?id=46653774<![CDATA[Language Modeling, Part 2: Training Dynamics]]>Article URL: https://connorjdavis.substack.com/p/language-modeling-part-2-training

Comments URL: https://news.ycombinator.com/item?id=46549435

Points: 1

# Comments: 0

]]>Fri, 09 Jan 2026 02:27:30 +0000https://connorjdavis.substack.com/p/language-modeling-part-2-trainingcjamsonhnhttps://news.ycombinator.com/item?id=46549435https://news.ycombinator.com/item?id=46549435<![CDATA[New comment by cjamsonhn in "Neural Networks: Zero to Hero"]]>Highly recommend this as well. Does a great job of helping you build intuition for why things like gradient descent and normalization work. Also gets into the weeds on training dynamics and how to ensure they are behaving properly

]]>Sun, 04 Jan 2026 18:38:41 +0000https://news.ycombinator.com/item?id=46490837cjamsonhnhttps://news.ycombinator.com/item?id=46490837https://news.ycombinator.com/item?id=46490837<![CDATA[New comment by cjamsonhn in "Rob Pike goes nuclear over GenAI"]]>> Because the motherfuckers behind public AI interfaces fine-tune them to be as human-like, as rewarding, as dopamine-inducing, as addictive, as possible?

And to think they dont even have ad-driven business models yet

]]>Fri, 26 Dec 2025 21:37:07 +0000https://news.ycombinator.com/item?id=46396550cjamsonhnhttps://news.ycombinator.com/item?id=46396550https://news.ycombinator.com/item?id=46396550