Kaoru's Blog

Securing GPG and SSH Keys with YubiKey

Sat, 14 Feb 2026 16:00:00 +0900

Introduction

How do you manage your GPG and SSH keys?

Saving private keys in cloud storage like Dropbox?
Copying and reusing the same key across all PCs?
Creating separate keys for each PC but losing track of which key is where?

If this sounds like you, that’s weak. You are vulnerable.

Figure 1: You should be seriously concerned.

Like many of you, I used to struggle with managing GPG keys (needed for pass or sops) and SSH keys (essential for accessing other machines or ordering coffee) as my device count increased:

2026 Week 6 - Weekly Reading

Sun, 08 Feb 2026 14:47:00 +0900

State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490 - YouTube

RLVR (Reinforcement Learning with Verifiable Rewards): We still haven’t seen the limits of Scaling Laws here.
if you learn more, you forget more (No Free Lunch): LLMs also tend to forget previously learned information as they learn new things.
Is the dream of AGI dying?: The dream of achieving AGI with a single model is dying; instead, we are moving towards realizing AGI through the collaboration of multiple specialized agents.
rasbt/LLMs-from-scratch: Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

2026 Week 5 - Weekly Reading

Sun, 01 Feb 2026 20:12:00 +0900

Highlight of the week: I chipped my front tooth (3rd time in 5 years)

Best Ways to Build Better Habits & Break Bad Ones | James Clear - Huberman Lab

How to form good habits and break bad ones.
Standard advice that sounds familiar but is solid:
- Good habits: Make it obvious, attractive, easy, satisfying
- Bad habits: Make it invisible, unattractive, difficult, unsatisfying
Personally, I wanted to break the habit of mindlessly checking LINE News, but I couldn’t delete the LINE app itself. Inspired by this podcast, I looked it up and found this method to hide just the News tab. (Game changer!)

Tweet from Andrej Karpathy

A few random notes from claude coding quite a bit last few weeks.

Coding workflow. Given the latest lift in LLM coding capability, like many others I rapidly went from about 80% manual+autocomplete coding and 20% agents in November to 80% agent coding and 20% edits+touchups in…

2026 Week 4 - Weekly Reading

Sun, 25 Jan 2026 18:58:00 +0900

The Best Way to Read a Book (That Nobody’s Doing)

A video where Jeremy Howard explains his close reading workflow with Solveit.
The core idea: “load a lot of relevant context first, then read chapter-by-chapter while chatting with an AI, carrying context forward each time”.
It feels close to what I’ve been doing with gptel for papers/books, but the explicit chapter-level context handoff was new to me.
- (Related) Past post: A refined Emacs LLM setup

2026 Week 3 - Weekly Reading

Sun, 18 Jan 2026 18:58:00 +0900

I haven’t been updating my blog for a while, so I’m restarting with a lightweight weekly memo: things I read/watched this week that I enjoyed.

Onoguchi Snow Dome

A collector’s site showcasing their snow globe collection with photos.
The early-Heisei-web vibe is oddly charming.

The security paradox of local LLMs - Quesma Blog (HN)

Local LLMs feel “more secure” than cloud LLMs at first glance, but it’s not that simple.
The example attack prompts are fun to read (especially Attack #2).

Love Generation filming locations (map)

A fan-made location guide for a classic 90s drama I got hooked on again via Netflix.
Kimura Takuya is way too cool.

To Those Who Think Data Scientists are Becoming Obsolete

Sun, 31 Aug 2025 20:31:00 +0900

Figure 1: At 27, I have worries I can’t tell anyone.

Introduction: The Problem with the “AI Replaces Experts” Narrative

In recent years, we hear everywhere that specialized white-collar jobs will be taken by AI.

I’ve been deeply immersed in Data Science (DS) since my undergrad—through work, research, and hobbies. Lately, tech-illiterate family members and friends with no programming experience have started asking me (without any malice), “Are you still doing programming?” or “Can’t AI do everything now?”

(NixOS) Fixing "Could not start dynamically linked executable" When Launching mcp server via uvx

Wed, 02 Jul 2025 11:16:00 +0900

As shown under another article, I’m using uvx to run some mcp servers needed by mcp.el. However, when I first started them on NixOS, I encountered an error. After consulting Gemini for solutions and a root cause analysis, here are my notes.

Symptoms and Error Message

Checking the *Mcp-Hub/ buffer with mcp-hub-view-log, the following error appeared:

[stderr]  Could not start dynamically linked executable: /home/bk/.cache/uv/archive-v0/unI5q9QapqXHm9fPXna4G/bin/python
[stderr]  NixOS cannot run dynamically linked executables intended for generic
[stderr]  linux environments out of the box. For more information, see:
[stderr]  https://nix.dev/permalink/stub-ld
[jsonrpc] D[14:56:49.461] Connection state change: `exited abnormally with code 127

Solution: Enable nix-ld

Adding the following to your nix configuration resolved the issue for me:

A Refined Emacs LLM Environment with gpt.el & mcp.el

Wed, 02 Jul 2025 11:16:00 +0900

Introduction

The integration of large language models (LLMs) into Emacs has rapidly evolved over the past few years.

My first encounter with an LLM inside Emacs was in November 2022, through AI-powered code completion with copilot.el. I still vividly remember how impressed I was at that moment.

Shortly after, ChatGPT was launched, and tools like ChatGPT.el made it possible to interact with AI within Emacs—though, at that time, the features were few, and I didn’t really feel the advantages of using LLMs in Emacs.

How to Use mermaid.js with ox-hugo

Sat, 21 Jun 2025 22:07:00 +0900

Introduction

This blog is managed with the workflow shown below.

  graph LR
  org-mode -- <a href="https://ox-hugo.scripter.co/">ox-hugo</a> --> markdown -- <a href="https://gohugo.io/">Hugo</a> --> HTML

When you want to include a diagram like the flowchart above, what methods come to mind?

Surely, you’re not thinking about making diagrams in PowerPoint or similar tools, exporting them as images, and… calling it a day?

That’s a D-tier move

Story of How My Hoff Solutions Site Climbed Search Rankings After (?) Optimization 🚀

Sat, 03 May 2025 13:28:00 +0900

TL;DR (Summary)

This is a solutions site for the end-of-chapter exercises in A First Course in Bayesian Statistical Methods / 標準ベイズ統計学 (Japanese edition). I started it during my Master’s degree and still update it, albeit at a very slow pace.

Recently, after just a few updates, the site went from not appearing in Google search results at all to ranking highly. Here’s a memo on what I did 📝.

Motivation for Improvement: “Doesn’t Show Up in Search Results at All”

I put a lot of effort into this site during my Master’s, second only to my thesis, and I thought there would be decent demand from readers of Hoff, “A First Course in Bayesian Statistical Methods”.

Guide to Setting Up SKK on Doom Emacs & Mac

Sat, 12 Apr 2025 13:52:00 +0900

What You’ll Learn in This Article

What SKK is
How to install and configure AquaSKK
How to set up SKK (ddskk) in Doom Emacs
Challenges in adopting SKK and their solutions
Impressions and evaluation after using SKK long-term

What is SKK?

SKK (Simple Kana to Kanji conversion program) is a type of Japanese input method editor (IME). Unlike typical IMEs, it features explicit switching between input modes and unique operations for selecting conversion candidates. Key characteristics include:

The Convenience of Emacs' Indirect Buffers

Fri, 28 Mar 2025 09:36:00 +0900

When using Emacs, have you ever thought, “I wish I could edit just a part of this file in another window…” or “I’d like to compare the same content with different display modes…”?

I certainly have, many times. That’s when I discovered and was impressed by the Indirect Buffer feature!

This post explains the basic concept of Indirect Buffers, how to use them, and some points to keep in mind.

Anki Super Vocab: My Quest for the Ideal Vocabulary Learning Card

Mon, 17 Mar 2025 10:56:00 +0900

Introduction

Are you an Anki user? Anki is a powerful memorization tool that utilizes a spaced repetition system (SRS). I’ve been a long-time Anki user, and it’s become indispensable, especially for language learning.

Recently, while using Anki for English vocabulary learning, I had an idea for the “perfect” card type – one with features I’d always wished for. I’ve now realized this ideal card type and made it publicly available on GitHub. This article introduces the “Super Vocab” card type, explains the development process, and, for Anki/Emacs users, describes how to create cards using anki-editor.

Startup error in Emacs: "Project … cannot be read"

Sun, 23 Feb 2025 18:42:00 +0900

The Problem

The other day, when I started Emacs as usual, the following message appeared in the minibuffer:

Project <Project Name> at <Project Path> cannot be read.

Emacs then became a piece of software that simply displayed a white screen.

I had recently changed the path of this project, so I suspected that might be the cause. I tried modifying/deleting the project path in projectile-known-projects-file, but the situation didn’t improve. I also tried commands like doom sync and doom upgrade, but none of these resolved the issue.

Troubleshooting "Server julia-ls starting exited… Do you want to restart it? (y or n)" in Julia/Emacs

Wed, 29 Jan 2025 16:00:00 +0900

The Problem

So, I recently fired up Emacs to do some Julia coding after a while, and I ran into this annoying message:

Server julia-ls:50759/starting exited (check corresponding stderr buffer for details). Do you want to restart it? (y or n)

It kept popping up (even if I pressed y), and LSP just wouldn’t start.

It took me a bit of time to figure this out, so I’m leaving this note here for anyone else using Julia in Emacs.

Published the Source Code for the Hoff Exercise Solutions Site

Tue, 28 Jan 2025 16:18:00 +0900

Published the source code for: Example Solutions for Hoff’s Exercises

GitHub

This site was published exactly two years ago and has been updated gradually since then.

Previously, I had written on the site asking people to contact me via email if they found any mistakes or typos, but ultimately, I never received any such contact in two years 😭

(I did receive thank-you emails from researchers overseas, which made me very happy.)

Advent of Code 2023 Day 1

Tue, 07 Jan 2025 12:07:00 +0900

Yo, Happy New Year, peeps!

Last year, I was on that grind, tackling the 2023 Advent of Code bit by bit. So, I’m droppin’ my solutions and some quick notes here, ya dig?

I’m coding all this in Python, straight up.

I’m thinkin’ of droppin’ the whole thing on GitHub when I’m done, but no promises on when that’ll be, ya feel me?

Aight, let’s jump into Day 1’s solution, no time to waste.

Fixed Ugly Math Formulas in Anki After Emacs Upgrade

Sun, 24 Nov 2024 15:41:00 +0900

Introduction

Hey everyone!

I regularly use anki-editor to create Anki cards. But after upgrading Emacs, my math formulas started looking really messed up.

Specifically, I had these two problems:

In the desktop Anki browser’s preview for each field, the formulas weren’t rendering as formulas. They just showed up as plain LaTeX code.
Formulas were being displayed as PNG images. (They used to be rendered inline with MathJax.)

Since I rarely edit cards directly in the desktop Anki app, the first issue was somewhat tolerable. But the second one was really bugging me. The image resolution of the formulas was lower, and they didn’t match the size of the surrounding text.

Updated Hoff/AFCBSM Exercise 10-4

Tue, 19 Nov 2024 20:32:00 +0900

Introduction

I have updated Answers of exercises on Hoff, A first course in Bayesian statistical methods (added 10-4), so I’d like to briefly touch upon its content in this blog post as well.

First, let me briefly explain the relationship between this problem and the textbook content. I previously wrote an article titled Hoff/標準ベイズのM-Hアルゴリズムがworkすることの証明でつまずいた話 (A story about stumbling on the proof that the M-H algorithm in Hoff/Standard Bayes works), and in Hoff (2009), the proof that the M-H algorithm works in Chapter 10 is done in the following steps:

Fixing Reference Display Issues with ox-hugo

Tue, 12 Nov 2024 14:40:00 +0900

The Problem

I was trying to use Org Cite Citations and the following code to display my references:

#+print_bibliography:

But only the “References” header showed up – the actual list of references was missing.

I checked the generated Markdown (.md) files in my content/ directory, and the reference list was there in HTML. However, when I looked at the final HTML files in the public/ directory, the references were gone.

This suggested the problem was happening during the Markdown-to-HTML conversion (the md -> html part of org -> md -> html). After some digging, I found an issue on GitHub about the same problem, which led me to the solution.

A Story About Stumbling on the Proof that the M-H Algorithm in Hoff/AFCBSM Works

Mon, 11 Nov 2024 00:00:00 +0000

Introduction

Regarding the proof presented in Hoff (2009) and Hoff et al. (2022), Section 10.4.2 “Why does the Metropolis-Hastings algorithm work?”, I encountered a point of confusion, so I decided to write about it to organize my thoughts.

Hoff’s Proof Flow

In Hoff (2009) and Hoff et al. (2022), the proof explaining why the Markov chain generated by the M-H algorithm can approximate the target distribution \(p_0\) proceeds as follows:

My Master's Thesis Hit arXiv!

Sat, 05 Oct 2024 00:00:00 +0000

Whassup, peeps! It’s been a minute since my last post (shoutout to the one person probably reading this, you the real MVP!).

My Master’s thesis, with some fresh updates, just dropped on arXiv. Check it:

Paper: Babasaki, K., Sugasawa, S., McAlinn, K. and Takanashi, K. (2024). Ensemble doubly robust Bayesian inference via regression synthesis. (arXiv:2409.06288)

So, this paper, it’s all about takin’ this ensemble method called Bayesian Predictive Synthesis (BPS) that Professor McAlinn cooked up, and flexing it into the world of causal inference, specifically for estimatin’ Average Treatment Effects (ATE). We’re callin’ our new method “doubly robust Bayesian regression synthesis”. If you wanna get into the nitty-gritty, peep the paper, ya dig?

Exercise 0.4.4

Tue, 04 Apr 2023 00:00:00 +0000

Question

a: Make up a nonmathematical function that is one to one but not onto.
b: Make up a mathematical function that is one to one but not onto.

Answer

a: “The student number of” from Keio University students to integers.
b: \(f(x) = 2^x\)

Exercise 0.2.2

Sun, 05 Mar 2023 00:00:00 +0000

Question

Explain Why one of these statements is true and the other is false:

\begin{align*} (\forall \text{ man } M)(\exists \text{ woman } W) \mid W \text{ is the mother of } M \\ (\exists \text{ woman } W)(\forall \text{ man } M) \mid W \text{ is the mother of } M \end{align*}

Answer

The first statement is true, and the second is false because every man has a mother but there doesn’t exist a woman who is the mother of all men.