Tech on the art of simplicity

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

Sun, 28 Jun 2026 00:00:00 +0000

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

I keep a few Claude Code and Cursor sessions running through the day, and at some point I started actually watching the token counter instead of ignoring it. The thing that struck me wasn’t how much it climbed. It was what it climbed on. Most of the budget wasn’t going on the model reasoning about my problem. It was going on plumbing: ls, grep, reading the same file three times to rebuild a call path, scrolling a 400-line cargo test dump to find the one assertion that failed.

The Harness Engineering Cheat Sheet

Tue, 28 Apr 2026 00:00:00 +0000

The Harness Engineering Cheat Sheet

I have a Cursor problem. Or maybe an “agent session” problem more generally — I keep five or six chats open, each with its own slice of context, each in a slightly different state. We have nice skills. We have nice tools to fetch the right context. We have a code-review skill that drops inline comments on a PR. But I still lose track of which window is doing what, and occasionally paste the wrong thing into the wrong session.

I Built an AI Ultrarunner to QA My App

Wed, 11 Mar 2026 00:00:00 +0000

I Built an AI Ultrarunner to QA My App

I’ve been building a race tracking app for ultrarunners. It lets you manage your race calendar, track gear, write journey entries, and plan your season. The problem? I have no users yet. Just me.

That means nobody is poking around, trying weird things, or stumbling into edge cases. The app has unit tests, CI checks, type checking — all the usual stuff. But there’s a whole class of bugs that only surface when someone actually uses the thing. Opens the browser, clicks around, creates records, and goes “wait, that doesn’t look right.”

Smile every time you sudo

Thu, 18 Feb 2021 00:00:00 +0000

If you want your terminal prompt to look like this every time you sodo, please continue reading. note: i did on ubuntu so for other flavor, file path might be slightly different

step 1. download this amazing ascii art text and move to /etc/sudoers.d/sudoers.lecture

❯ curl -O https://caferock.org/chris/groot.txt
❯ sudo cp groot.txt /etc/sudoers.d/sudoers.lecture

step 2. edit /etc/sudoers (or /etc/sudoers.d/privacy)

Defaults lecture = always
Defaults lecture_file = /etc/sudoers.d/sudoers.lecture

Let’s see this magnificent art <3

# reset timestamp
❯ sudo -k 
# some sudo command
❯ sudo -l

Cheers!

nvm to manage node version

Sun, 25 Oct 2020 00:00:00 +0000

NVM is Node Version Manager is a tool to manage multiple node.js version.

list installed version

❯ nvm ls

 v8.17.0
-> v12.9.0
 system
default -> 12.9 (-> v12.9.0)
node -> stable (-> v12.9.0) (default)
stable -> 12.9 (-> v12.9.0) (default)

list available version on remote (is version 14 available? If so it will list)

❯ nvm ls-remote 14
 v14.0.0
 v14.1.0
 ...
 v14.13.1
 v14.14.0

Install the version that you would like

❯ nvm install 14

Set version 14 (or another version) as default

nvm alias default 14

Switch to version 8 (or another version)

❯ nvm use v8
Now using node v8.17.0 (npm v6.14.8)

Cheers!

Install and use fzf

Mon, 13 Jul 2020 00:00:00 +0000

fzf is a general-purpose command-line fuzzy finder. Let’s install and make you 10 times more productive!

install fzf

# clone into ~/.fzf
git clone --depth 1 https://github.com/junegunn/fzf.git ~/.fzf
# run install command
~/.fzf/install

output: answer “y” to enable auto-completion and key bindings

Cloning into '/home/naoko/.fzf'...
remote: Enumerating objects: 101, done.
remote: Counting objects: 100% (101/101), done.
remote: Compressing objects: 100% (94/94), done.
remote: Total 101 (delta 4), reused 28 (delta 2), pack-reused 0
Receiving objects: 100% (101/101), 180.78 KiB | 1.83 MiB/s, done.
Resolving deltas: 100% (4/4), done.
Downloading bin/fzf ...
 - Found in $PATH
 - Creating symlink: bin/fzf -> /usr/bin/fzf
 - Checking fzf executable ... 0.20.0 != 0.21.1
 % Total % Received % Xferd Average Speed Time Time Time Current
 Dload Upload Total Spent Left Speed
100 637 100 637 0 0 1750 0 --:--:-- --:--:-- --:--:-- 1750
100 1127k 100 1127k 0 0 991k 0 0:00:01 0:00:01 --:--:-- 2597k
 - Checking fzf executable ... 0.21.1
Do you want to enable fuzzy auto-completion? ([y]/n) y
Do you want to enable key bindings? ([y]/n) y

Generate /home/naoko/.fzf.bash ... OK
Generate /home/naoko/.fzf.zsh ... OK

Do you want to update your shell configuration files? ([y]/n) y

Update /home/naoko/.bashrc:
 - [ -f ~/.fzf.bash ] && source ~/.fzf.bash
 + Added

Update /home/naoko/.zshrc:
 - [ -f ~/.fzf.zsh ] && source ~/.fzf.zsh
 + Added

Finished. Restart your shell or reload config file.
 source ~/.bashrc # bash
 source ~/.zshrc # zsh

Use uninstall script to remove fzf.

For more information, see: https://github.com/junegunn/fzf

Let’s play

Key Binidngs

ctl + r: fuzzy find thorugh your command history
ctl + t: fuzzy find through the current directory

`**` trigger command + tab

It understand context and list only relative stuff. eg:

Pretty JSON in Sublime on Linux

Thu, 14 May 2020 00:00:00 +0000

I’ve been using JSONLint for every JSON prettifier needs. This site is super because it even format invalid JSON. For example, standard JSON prettifier won’t prettify JSON with key or value enclosed in single quote but this site does but having your local editor do quick prettifier if you have valid JSON or willing to quickly fix with find and replace.

Sublime is my go to text editor - taking note, prettify JSON etc. So let’s do that real quick.

Stanza - A Python NLP Library for Many Human Languages

Sun, 12 Apr 2020 00:00:00 +0000

I tested out Stanza. English tokenizer and definately works. I ran quick test with Japanese lang and output was somewhat unexpected.

import stanza

# japanese "ja", for english model "en"

stanza.download("ja")
nlp = stanza.Pipeline("ja")
doc = nlp("皆さんおはようございます！　ご機嫌いかがですか？")

for i, sentence in enumerate(doc.sentences):
 print(f"===== Sentence {i+1} tokens =====")
 print(*[f"word: {word.text}\t upos: {word.upos} xpos: {word.xpos}" for word in sentence.words], sep="\n")

The output is:

Useful Ubuntu Keyboard Shortcuts

Sun, 08 Mar 2020 00:00:00 +0000

See Useful keyboard shortcuts

Addition to the above:

Super + Space: Change input keyboard
Ctl + Q: Close an application window
Super + ↑ : Maxmize the application window.
Super → or ← : Move the application window to left | right


Cheers!

Kafka CLI kaf

Sat, 25 Jan 2020 00:00:00 +0000

Kafka is super popular and powerful data streaming platform but I’m surprised by lack of easy cli tool.

To find out version, there is no such thing as kafka --version you do:

$ find /opt/kafka/libs/ -name kafka_\* | head -1
/opt/kafka/libs/kafka_2.13-2.4.0-test-sources.jar

where 2.13 is scala version and 2.4.0 is kafka version.

One tool that I played with is lovely called kaf, “Modern CLI for Kafka”.

to list topics, you do:

sudo bin/kafka-topics.sh --list --bootstrap-server localhost:9092

whereas with kaf you do:

Docker Image / Container Cleaning

Sun, 25 Aug 2019 00:00:00 +0000

Unused Docker image and container can quickly take up your disk spaces.

Let’s check disk usage

Summary of data used. You can add -v flag for detailed info.

$ docker system df
TYPE TOTAL ACTIVE SIZE RECLAIMABLE
Images 87 1 63.39GB 63.17GB (99%)
Containers 1 0 0B 0B
Local Volumes 69 1 2.732GB 2.683GB (98%)
Build Cache 0 0 0B 0B

Remove unused data

The following command will remove:

all stopped containers
all networks not used by at least one container
all dangling images (see below fo dangling images)
all build cache By default, volumes are not removed to prevent important data from being deleted.

$ docker system prune
WARNING! This will remove:
 - all stopped containers
 - all networks not used by at least one container
 - all dangling images
 - all dangling build cache

Are you sure you want to continue? [y/N] y
Deleted Containers:
...
Total reclaimed space: 13.91GB

Options:

Create and Extract tar.gz on Ubuntu

Wed, 07 Aug 2019 00:00:00 +0000

Create tar.gz archives

tar czf new-tar-file-name.tar.gz file-or-folder-to-archive

c - create new archive z - compress the archive using gzip f - use archive file

Extract archives

tar -xzf tar-file-name.tar.gz

x - extract the archive. z - uncompress the archive using gzip. f - use archive file.

Let’s do it.

I only want to compress 3 files. When I extract archive what I want to see is files in the directory.

$ tree file_dir/
file_dir/
├── file1.txt
├── file2.txt
└── file3.txt

$ cd file_dir
$ tar czf files.tar.gz .
tar: .: file changed as we read it

# moved files.tar.gz to different directory
$ tar -xzf files.tar.gz 
$ ls
file1.txt file2.txt file3.txt file_dir files.tar.gz

Cheers!

Install GraalVM and run ptyhon with debugger

Sun, 14 Apr 2019 00:00:00 +0000

What is GraalVM?

GraalVM is a high-performance, embeddable, polyglot virtual machine for running applications written in JavaScript, Python, Ruby, R, JVM-based languages like Java, Scala, Kotlin, and LLVM-based languages such as C and C++.

Here is the Official doc link

Hmm… Okay, I have to see it.

Let’s install

Below is the way I installed GraalVM Community Edition on Ubuntu 18.04. For other platform, the official doc installation guide is here.

# update this number to latest version from here: https://github.com/oracle/graal/releases
version=1.0.0-rc15
wget https://github.com/oracle/graal/releases/download/vm-${version}/graalvm-ce-${version}-linux-amd64.tar.gz
tar -xvzf graalvm-ce-${version}-linux-amd64.tar.gz
# clean up
rm graalvm-ce-${version}-linux-amd64.tar.gz
# to wherever you want. 
mv graalvm-ce-1.0.0-rc15/ ~/bin/graalvm
# if you want to make it permanent, put this in your bashrc 
export PATH=$HOME/graalvm/bin:$PATH

Now that your graalvm/bin in your path, you’ll get the GraalVM versions of those runtimes.

Get my public IP address via CLI

Thu, 04 Apr 2019 00:00:00 +0000

How to determine my public IP address assigned by the ISP via CLI

Use dig command

dig +short myip.opendns.com @resolver1.opendns.com

Use 3rd party

curl ifconfig.me

Cheers!

Fingerprint of PEM ssh key

Wed, 03 Apr 2019 00:00:00 +0000

The problem

I was on AWS and needed to select SSH Key pair. It has been a while that I needed to create one that might require SSH Key pair. So to be sure I got the correct key, I needed to compare fingerprint listed on Key Pair list. I do have pem key with name contain aws so I want to generate finger print for that key.

How to solve it

openssl pkcs8 -in path/to/private_key_file -nocrypt -topk8 -outform DER | openssl sha1 -c

or use AWS tool

Get started with Cython

Sun, 24 Mar 2019 00:00:00 +0000

Cython is designed as a C-extension for Python. The developers can use Cython to speed up Python code execution. Performance gains are most significant in CPU-bound programs.

Here is some cpu-intense code in python

# cat compute_it.py 
import datetime
import math


def main():
 t0 = datetime.datetime.now()
 do_math(num=30_000_000)
 dt = datetime.datetime.now() - t0
 print("Done in {:,.2f} sec".format(dt.total_seconds()))


def do_math(start=0, num=10):
	pos = start
	k_sq = 1000 * 1000
	while pos < num:
		pos += 1
		math.sqrt((pos - k_sq) * (pos - k_sq))


if __name__ == '__main__':
 main()

Took 7.39 sec

list changes of specific line(s) with git

Tue, 19 Mar 2019 00:00:00 +0000

If you only need to know who was the last person to change the line, git blame will do

To see who modified the code on line # 85 and next 5 lines

git blame ${path-to-the-file} -L 85,+5

output would look like this

f1f66980053 (Naoko 2019-02-25 17:32:25 -0700 85) text = re.sub(r'{([^\n]+?)}', r'{0[\1]}', template["text"])
50ccaf0f3b8 (Naoko 2019-02-16 10:44:54 -0700 86) template["text"] = text.format(template_vars)
f373fd241db (Tom Bocklisch 2018-07-10 14:33:18 +0200 87) except KeyError as e:
f373fd241db (Tom Bocklisch 2018-07-10 14:33:18 +0200 88) logger.exception(
390792a0008 (Tom Bocklisch 2018-11-16 12:51:21 +0100 89) "Failed to fill utterance template '{}'. "

But often the above information is not good enough.

How to fork and keeping your github repo in sync

Tue, 19 Feb 2019 00:00:00 +0000

Forking a Github Repository

Just click “fork” button on upper right corner

black tea

Making a Local Clone

$ git clone https://github.com/YOUR_USERNAME/YOUR_FORK.git

Adding a Remote

# list current configured remote repository for your fork
$ git remote -v

# add upstream
$ git remote add upstream https://github.com/ORIGINAL_OWNER/ORIGINAL_REPOSITORY.git

# list again to see upstream is added
$ git remote -v
origin https://github.com/YOUR_USERNAME/YOUR_FORK.git (fetch)	
origin https://github.com/YOUR_USERNAME/YOUR_FORK.git (push)
upstream https://github.com/ORIGINAL_OWNER/ORIGINAL_REPOSITORY.git (fetch)
upstream https://github.com/ORIGINAL_OWNER/ORIGINAL_REPOSITORY.git (push)

Working in a Branch

Create and checkout a feature branch. Personally, I like to set github issue number followed by summary of issue. This notation GH-${issue_num} will create a link to the issue on github.

Install Python on Ubuntu

Mon, 15 Oct 2018 00:00:00 +0000

Check the latest version here At the time of writing, 3.8.0 is the latest and 3.8.5 has release candidate.

Also make sure you have sqlite3, libbz2-dev and libffi-dev are installed

sudo apt-get install libsqlite3-dev libbz2-dev libffi-dev

version=3.8.5
wget https://www.python.org/ftp/python/${version}/Python-${version}.tgz
tar xzvf Python-${version}.tgz
cd Python-${version}
# Linux (or any Unix-like system), the default prefix and exec-prefix are /usr/local.
# thus you should be able to omit --prefix here
# --enable-optimizations option for significant speed boost (10-20%) but much
# slower build process
./configure --prefix /usr/local --enable-optimizations
make
sudo make install
# OR if you want to skip creating the python link then:
sudo make altinstall

in case you want to remove and re-install it again cause some software was missing before installation

How to setup custom domain in Blogger via Namecheap

Mon, 03 Sep 2018 00:00:00 +0000

The Goal:

You got blog going with Blogger and now ready to setup custom domain!

How to setup custom domain?

First, you will need to register your domain. To register your domain, you want to go through domain registrar. To become the domain registrar, you will need ICANN Acrediation so you want to go ahead register through registrar instead of becoming one.

Now there are many of them out there. The popular ones are likely GoDaddy. I see Google is providing this service too. I decided go to with NameCheap.com for a few reasons:

Elasticsearch Access

Wed, 22 Aug 2018 00:00:00 +0000

The problem:

You installed Elasticsearch on server. You can run curl localhost:9200 and all looks good but the access is denied from outside when curl <server-ip>:9200.

How to solve it

So first thing first. Elasticsearch do need to listen to ip you are accessing. To make it listen to all, you can simply change / add network.host: 0.0.0.0 to /etc/elasticsearch/elasticsearch.yml and restart elasticsearch server. Try curl <server-ip>:9200 and works? That’s great. Your server is configured / ready for port 9200.

Intercept Python Logging

Thu, 26 Jul 2018 00:00:00 +0000

The problem:

I need to ship specific log record and had formatter written in python. It is pretty complex transformation.

I thought of using Logstash but I then need to either convert this python logic or write a plugin to use already written python parser. Plus I need to install logstash… I wanted a simpler solution

How to solve it

Use custom python logging Handler and Filter!

import logging

messages = []
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)


class ListenFilter(logging.Filter):

 def filter(self, record):
 """Determine which log records to output.
 Returns 0 for no, nonzero for yes.
 """
 if record.getMessage().startswith('dont: '):
 return False
 return True


class RequestsHandler(logging.Handler):
 def emit(self, record):
 """Send the log records (created by loggers) to
 the appropriate destination.
 """
 messages.append(record.getMessage())


handler = RequestsHandler()
logger.addHandler(handler)

filter_ = ListenFilter()
logger.addFilter(filter_)

# log I want
logger.info("logme: Howdy!")


# log i want to skip
logger.info("dont: I'm doing great!")

# prints ['logme: Howdy!']
print(messages)

Cheers!

ML Platforms

Sun, 08 Jul 2018 00:00:00 +0000

The problem:

When you start small Machine Learning team with a few projects, your experiment is done via Jupyter Notebook and maybe the notebook is in github. The notebook might contain a method to download data so it can be reproducible but it is getting harder and harder to track various experiments.

We also need to make sure models does not train on corrupted / skewed data and only high quality model are pushed to production. Currently these processes are manual and not centralized nor has unified common tool.

Fix screen resolution after 18.04 LTS upgrade on System 76

Tue, 05 Jun 2018 00:00:00 +0000

The problem:

Ubuntu prompt me that 18.04 LST is available so I clicked Upgrade to initiate upgrade. When upgrade is completed, the system was rebooted and came back with lowest screen resolution you can ever imagine.

Fix

In summary…. This is only good for System 76 system. Looks like driver update was necessary. Anyhow…

So upgrade was completed

$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 18.04 LTS
Release:	18.04
Codename:	bionic

and looks like I lost lsb module. so restoring that…

Python Project Install - develop vs install & setuptools vs pip

Mon, 05 Mar 2018 00:00:00 +0000

The problem:

I don’t understand the difference between setup.py develop and setup.py install
I don’t understand the difference between setup.py develop and pip install -e [dir]
I don’t see the changes to my code when I import my code

The difference between `setup.py develop` and `setup.py install`

In short, you want to run setup.py develop when you are editing code because when you run setup.py install, it will copy your code into site-packages thus if you want to test your latest code you will need to install (copy) again. On the other hand, with develop, it creates a link to your source code so that when you import your code, it is your latest code.

Upgrade vagrant to latest

Tue, 06 Feb 2018 00:00:00 +0000

The Problem:

I love vagrant. I am so dependent on testing my ansible before I run against mutable boxes. Yes, I’m still living in place where Ops provision VM and the box stays mutable. It is okay though… change should be coming. Anyhow, I installed vagnrat on my new Ubuntu machine (ubuntu-17.10) but the version was 1.9. It was working fine until I upgraded the box to debian/stretch64. It says that The box 'debian/stretch64' could not be found. And noticed that it was trying to look in the old repo https://atlas.hashicorp.com instead of this https://app.vagrantup.com.

Upgrade from Debian Jessie to Stretch

Mon, 05 Feb 2018 00:00:00 +0000

The Problem:

I was running Debian Jessie (debian 8) and the package I needed to install was in debian testing repo. So I added testing repo and installed the package (libsnappy1v5) and I got scold by my super sysadmin boss: “Naoko, don’t make a FrankenDebian!”

The Solution:

The package I needed is now available in Stretch (debian 9) stable so I need to upgrade to it.

First, I need to remove testing repo I’ve added

Remote Edit with PyCharm

Tue, 30 Jan 2018 00:00:00 +0000

The Problem:

I have powerful Linux machine and semi-powerful mac-book pro. I have need of editing code on Linux box from mac-book.

PyCharm has deploy method for pushing code but that requires sftp to setup etc. I just want to use ssh to do it… I’ve done rsync method but sometimes manual sync was needed. I want more transparent method…

The Solution:

The easiest and most effective way for me was to mount linux directory via sshfs. It was amazingly easy. Assuming that you have established ssh connection with ssh-key with the remote machine. All you have to do was to install osxfuse and sshfs with brew.

No Space Left Error when pip install

Tue, 26 Dec 2017 00:00:00 +0000

The Problem:

Got

OSError: [Errno 28] No space left on device

When your home directory where your virtual environment located has more than enough space…

Why?:

The culprit is likely that your /tmp directory do not have enough space for some reason. During the pip installation, pip will use temporarily directory to perform what is necessary to perform installation (e.g. download source etc). Thus if you do not have enough space in /tmp that package installation requires then you will get disk space error.

Secret Management with Vault

Sat, 28 Oct 2017 00:00:00 +0000

The problem:

Where do you store your config/deployment management script? I’ve done 2 ways:

Put along with application/repo.
Put all the config/deployment script in one repo separately from apps.

Each approach has its own Pros and Cons and I prefer the first approach which IMO is simpler but creates some duplication which is okay…

I have some shared secrets that I encrypt with Ansible Vault but when I had to copy and pasted this secret on a few repos I started to look for better way to do this.

Tech on the art of simplicity

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

The Harness Engineering Cheat Sheet

The Harness Engineering Cheat Sheet

I Built an AI Ultrarunner to QA My App

I Built an AI Ultrarunner to QA My App

Smile every time you sudo

step 1. download this amazing ascii art text and move to /etc/sudoers.d/sudoers.lecture

step 2. edit /etc/sudoers (or /etc/sudoers.d/privacy)

Let’s see this magnificent art <3

nvm to manage node version

list installed version

list available version on remote (is version 14 available? If so it will list)

Install the version that you would like

Set version 14 (or another version) as default

Switch to version 8 (or another version)

Install and use fzf

install fzf

Let’s play

Key Binidngs

** trigger command + tab

Pretty JSON in Sublime on Linux

Stanza - A Python NLP Library for Many Human Languages

Useful Ubuntu Keyboard Shortcuts

Kafka CLI kaf

Docker Image / Container Cleaning

Let’s check disk usage

Remove unused data

Create and Extract tar.gz on Ubuntu

Create tar.gz archives

Extract archives

Let’s do it.

Install GraalVM and run ptyhon with debugger

What is GraalVM?

Let’s install

Get my public IP address via CLI

How to determine my public IP address assigned by the ISP via CLI

Use dig command

Use 3rd party

Fingerprint of PEM ssh key

The problem

How to solve it

Get started with Cython

list changes of specific line(s) with git

How to fork and keeping your github repo in sync

Forking a Github Repository

Making a Local Clone

Adding a Remote

Working in a Branch

Install Python on Ubuntu

How to setup custom domain in Blogger via Namecheap

The Goal:

How to setup custom domain?

Elasticsearch Access

The problem:

How to solve it

Intercept Python Logging

The problem:

How to solve it

ML Platforms

The problem:

Fix screen resolution after 18.04 LTS upgrade on System 76

The problem:

Fix

Python Project Install - develop vs install & setuptools vs pip

The problem:

The difference between setup.py develop and setup.py install

Upgrade vagrant to latest

The Problem:

Upgrade from Debian Jessie to Stretch

The Problem:

The Solution:

Remote Edit with PyCharm

The Problem:

The Solution:

No Space Left Error when pip install

The Problem:

Why?:

Secret Management with Vault

`**` trigger command + tab

The difference between `setup.py develop` and `setup.py install`