Nothing to be said
title
152334H
152334H Personal Blog
Latest posts
Some Notes on Alignment Blogs
These are quick self-notes I wrote this week, while scrolling various AI Alignment related posts online. There are no insights here, just mundane reading & paraphrasing to ensure I ‘get it’.
Notes on POIS
Post Orgasmic Illness Syndrome ( POIS ) is, allegedly, a Rare Sexual Disorder whose sufferers experience some combination of flu/fever/fatigue/irritability/aphasia/… rapidly after any orgasm.
Personal ratings of how models criticise me
Instead of a garbage monthly post, let’s look at how different models respond to my garbage.
Things worth doing
Most of my life has been dedicated towards intellectual excellence, but continuing to do so would be a mistake.
Are exploits free?
For various reasons, I had to develop a kernel n-day exploit last week.
Truncated Thoughts on 2025
It has been, overall, a pretty shit year.
Time To Think
I don’t remember what it’s like to think with more than 60 seconds of context.
Calculating the Cost of a Google Deepmind Paper
Recently, GDM released a great paper titled, Scaling Exponents Across Parameterizations and Optimizers , in which they conduct over 10,000 LLM training runs to obtain optimal hyperparameters under different regimes. After reading it (it was great), I wanted to test my understanding of the paper by tallying up all experiments conducted within, calculating the total compute cost it would take to…
DeepSeek Core Readings 0 - Coder
Paper summary: 1.3B to 33B LLMs on 1/2T code tokens (87 langs) w/ FiM and 16K seqlen. Strong effort in constructing pretraining data from Github from scratch, with repository-level samples. Evals beat OSS code models solidly + GPT-3.5 a bit; Coder-7B > CodeLlama-33B often. They don’t spend much effort on Instruction tuning. They commit a continued pretrain of DeepSeek LLM -> Coder: I believe…