Jeremys Awesome Blog

Kilo AI CLI Coding Agent

I’ve been working with Kilo CLI for the last day. So, this is first impressions. This is coming from working with Codex CLI for a few months. The following post comes from this experience. In this case I’m using GPT-5.5 medium in both Kilo and Codex.

The Kilo AI CLI has some cool features. I really appreciate the ability to connect with your existing provider. I like the auto-model routing. The interface is pretty cool, with a lot of themes to choose from. There is a persistent status section in the CLI that shows usage, context available, tokens used, modified files. It’s really cool.

However, I’ve consistently run into issues since attempting to use Kilo. And it’s not with the interface or the functionality of the software. It’s with the underlying tooling. Something about Kilo causes the LLM to do things that I don’t want it to do. Here’s an excerpt from my AGENTS.md file.

When the user pastes review feedback, follow the Review Feedback section below. Treat the feedback as a claim to verify, explain the conclusion, recommend the smallest correct fix, and ask before implementing unless the user explicitly asks for changes.
...
## Review Feedback

When addressing merge request or code review feedback, treat each comment as a claim to verify, not as an instruction to apply automatically.

- First confirm whether the reported problem is real by reading the relevant code paths and, when practical, reproducing or reasoning through the behavior.
- Validate suggested solutions before implementing them; they may be incomplete, too broad, or aimed at the wrong boundary.
- If the issue is valid, choose the smallest correct fix that changes the fewest lines and preserves existing intended behavior.
- If the issue is not valid, explain why with concrete code references or behavior.
- If a minimal fix could hide an inconsistent state, distinguish the legitimate case from the suspicious case and handle each explicitly.

That’s pretty clear about what I expect. My expectation is no changes unless it’s verified with me. Or unless I tell the agent to explicitly fix it. Over the last day I’ve given feedback to the agent and it continues to ignore my instructions, opting instead to make a change without verifying it with me. That’s frustrating.

Kilo Modes

Kilo comes with several modes out of the box. There’s a Code mode, a Plan mode, an Ask mode, a Debug mode, and an Orchestrator mode. Now, this frustration could be my fault for using the tool wrong. I started in “Code mode”. Perhaps, code mode, means that Kilo is going to strongly tell the LLM to write code, even if it contradicts my instructions. It could be my fault for not starting in ASK mode and then flipping to CODE mode. But… seriously. I’m not going to sit here switching modes all day when I expect the tool to be smart enough to follow instructions. If the tool is in code mode, and the instructions are “ask before coding” then the tool should ask. The mode should be inferred from the prompt. If the prompt is a question, then assume an answer is wanted. If the prompt is, fix this issue, then assume that a combination of debug+code is wanted. If the prompt is, implement this feature, then assume orchestrator+code is likely wanted.

Maybe that’s just me. But switching between modes seems like busy work for me.

Kilo Tooling

I have RTK set up on my machine and integrated into my AI agents. The tool exists as rtk and is callable by the AI agents. However, the code I was writing was in php and I don’t have php installed on my system. I do have it running in a container running on my system though. I have explicit instructions in my AGENTS.md file telling the Agent to look at local machine notes. My local machine notes tell the Agent all about how this current system uses Fedora and Podman for containers.

Today, Kilo failed hard on a few things:
– It did not use rtk. It attempted to use rtk php and when that failed it assumed rtk did not exist and didn’t use it anymore.
– It did not read my local notes. And because it did not read my local notes (despite being told to in the AGENTS file) it failed to validate my code using php
– It tried to use docker multiple times even though the local notes tell it to use podman

Working Verdict

I really like the concept of Kilo. Super cool interface, super cool themes, super cool status window, an auto-model selector. Kilo has a lot of good stuff going for it.

I’ve worked with Kilo and attempted to use it. It’s been an incredibly frustrating experience. Kilo is not reliable and it doesn’t listen.

Despite all the cool, and the hype, when the rubber meets the road Kilo is not up to snuff. Kilo is not up to snuff because Kilo ignores explicit directions.

My plan is to use Pi.dev. I’ve heard that it will work with the Kilo gateway. I probably won’t come back to Kilo, but who knows, they might update the tool to be more reliable. If they do, I’ll try it again.

Remember Stack Overflow?

It’s one of those places that I used to spend a lot of time on. But then the community got all pretentious and it became more important to users to show you all the ways your question had already been answered instead of just answer it.

About the time I started really disliking StackOverflow, ChatGPT came around and… well it doesn’t yell at me for my questions.

Truenas SCALE MineOS Server Not Starting

This post is to serve basically as long term memory for me. I keep forgetting how to do this, and the original post I used to refer to is gone. (Well, at the time of this writing it is). That said, WaybackMachine is awesome and I was able to access the archived original post there.

However, for ease, I’m just going to repeat the steps here because it’s easy enough to do.

Install the right Java Version

As of this writing, the TrueNAS Scale MineOS App, does not come with Java 21 pre-installed. Every time the MineOS App is updated, Java 21 is blown away. So, you have to re-install it in the MineOS App Container.

Step 1 – Access the MineOS App Shell

Accessing the TrueNAS Scale MineOS shell

Step 2 – Run apt update

Step 3 – Run apt install openjdk-21-jdk

Start The Servers

Ok – now that you’ve updated the java version, you can go in and start your MineOS servers.

Visit the Server page
Select the correct profile from the Server Actions box. (i.e. 1.21.4)
Select the correct runnable jar from the Java Settings box (i.e. minecraft_server.1.21.4.jar)
Push the big green Start button
Wait a bit and you should see the Server status checkbox flip from a red “Down” to a green “Up”

Here’s a couple screenies to help you get your bearings.

PHPStorm Classic Mode

So, one of the most annoying things IN MY LIFE has been when JetBrains started adding all of this random AI Assistant autocompletes to their IDE’s. I know I’m not alone, because there are even tickets like this one, How to Shut Off All this AI Crap?.

It’s not that I’m anti-AI – I’m anti-Intrusion. When I’m driving the car, I get to drive the car, not the AI Assistant. Unfortunately, the way JetBrains is adding AI into it’s editor feels very intrusive. And, also unfortunately, it’s a slog to go through all of the IDE settings to try and find every single little thing that needs to be turned off.

Luckily, you don’t need to! JetBrains has added some files that will allow you to selectively use AI tools. My favorite is dropping a .noai file in your project root. It’s easy to do, navigate to your project root in Powershell and run touch .noai. That’s it, all you need is an empty file to disable all the AI stuff in PHPStorm.

Now you can get back to PHPStorm Classic mode. LFG? Molten Core? … oops wrong classic mode.

But what about Junie?

That’s a valid question. Like I said, I’m not anti-AI. I actually like the coding agents, I’ve found them extremely helpful in certain circumstances. So, I wondered, will .noai disable all coding agents? Nope! It won’t. As of this writing .noai will disable the AI Assistant features in the IDE – but you can still use the Junie agent window. I don’t know if that’s an oversight or not, I actually recently heard that the Junie agent window is going away as they consolidate underneath the AI Assistant chat window.

The Better Junie Option (imo)

The better option, in my opinion, is to just use the Junie CLI. It’s currently in beta, and appears very usable from my testing. It’s integrated with your current JetBrains AI pricing model, so if you pay to use Junie within your IDE, then you are paying to use the CLI tool as well. I’ve found that I can open my own custom terminal, or use the built in IDE terminal, and open Junie via CLI. This gives me the best of both worlds, a powerful IDE and a powerful coding agent. I can use both when I want.

Not Quite .noai

Despite .noai working in most circumstances, I’m noticing little bits of AI crop back up. For example, inline Code Completion. I’m wanting classic intellisense, but JetBrains is giving me inline completion using cloud and local models. I had expected this to not happen, and indeed thought it wasn’t happening. But today, as I’m coding, I noticed phpstorm giving me ghosted AI completion hints. So, I navigated to phpstorm settings Editor > General > Code Completion > Inline and unchecked the Enable inline completion using language models: option.

.noai seems to have no effect on this setting

Capturing Animated GIFS on Linux with Spectacle

OK. Honestly, I’m just dropping a quick one in here because it’s necessary. You can easily capture Animated GIFS, Screen Recordings, Animated PNGS on Linux with KDE Spectacle.

Background

Here’s the deal. Windows 10 be complaining that it’s End-of-Life and it won’t run on my laptop anymore and they need me to update my hardware. You know what OS doesn’t need me to update my hardware? That’s right, it’s Linux. So I went ahead and dropped Fedora KDE Plasma Desktop on my laptop and it’s working great. I used Kubuntu years ago and I’m super impressed at how far Linux has come along. There’s pretty much a Linux alternative to all the tools I use, not to mention that a lot of the tools I use are Linux native anyways.

The one thing that I couldn’t quite replicate is ShareX on Linux. I used ShareX on Windows to do the following things:

Capture screenshots
Capture screen recordings

And that’s it. I don’t use any sharing features. Try as I may, I couldn’t find something that would allow me to do that easily in KDE. I tried Flameshot, and it’s pretty cool, but I found it’s not quite as good as what I’m used to with ShareX. My experience is that opening flameshot basically took a screenshot of what’s on my screen and then tried to give me a selection tool. But the way it worked was it offset all my screens so just opening Flameshot made everything on my screen look different.

There’s also a very promising ShareX fork called SnapX. It looks pretty cool, basically like ShareX but for everywhere! However, it’s still very much in active development and I didn’t feel like trying to beta test the tool.

Solution

Enter Spectacle. Yeah, I know what you’re thinking. That’s just a simple screenshot tool, it’s not going to check off both your boxes. But, here’s the thing. It does check off both boxes. The docs don’t tell you about it, nothing published seems to say anything about it, but the ability to capture both a screenshot and a screen recording is legit right there in the interface for Spectacle. (I’ll go ahead and drop in a super grainy picture of my monitor to illustrate).

See – it’s right there. Spectacle gives you the opportunity to take a screenshot or a screen recording, right there with sweet rectangular selection, window, or fullscreen options.

How can you save animated gifs or png’s though? You just need to set that as the output format in Spectacle’s settings. The way I accessed that is by clicking the Options button just to the right of “New Recording” (see above uber grainy screenshot of monitor). From there you can navigate to Video Saving. Right next to the Filename text field there is a file type dropdown. Your options are WebM/VP9, MP4/H.264, Animated WebP (better than GIF), and GIF (compatible, but inefficient). Sweet right?

Important Note: Screenshots are stored in your Pictures/Screenshots folder and screen recordings are stored in your Videos/Screencasts folder.

Here’s a list of the keyboard shortcuts available to use.

KDE Spectacle – Screenshot & Recording Keyboard Shortcuts

Global keyboard shortcuts available in KDE Spectacle for screenshots and screen recording.
Action	Global Shortcut	Global Alternate
Capture Active Window	Meta+Print
Capture Current Monitor	Shift+Print
Capture Entire Desktop	Meta+Shift+Print
Capture Rectangular Region	Meta+Shift+Print
Capture Selected Window	Meta+Ctrl+Print
Launch Spectacle	Print	Meta+Shift+S
Launch Spectacle without capturing
Start/Stop Region Recording	Meta+Shift+R	Meta+R
Start/Stop Screen Recording	Meta+Alt+R
Start/Stop Window Recording	Meta+Ctrl+R

Proof

A Screen recording taken with Spectacle

Final Thoughts

While it was frustrating to find this tool, and that it doesn’t seem to be documented main stream – it’s awesome that it exists and it’s built in to KDE! Props to the Spectacle team.

Get the word out, you can screen capture and screen record right from Spectacle. The tool is awesome 💥

The Morning Everything Went Wrong (and Still Worked Out)

LOL – first day of school for the kids today.

Obviously, I slept in. Planned to leave by 8:20 but woke up at 7:30. We’re out of bread. And my son was going to make a sandwich for lunch. That’s ok, I have time to run to the store and back.

Head out to the car. It’s not unlocking with the button, which is kind of odd. Open it up with the key. The light selector is turned to running lights… oh boy. I forgot to check the knobs after taking the toddler out of the front seat yesterday. I tried to turn the car on. Nothin.

I searched through two of our emergency kits. I have two sets of jumper cables, odd brands of band-aids, flashlights, batteries, foil blankets, cigarette lighter air compressors… but nothing to help me jump my car.

Not a problem, there’s a dude chillin in his truck two houses down. I walked over, he seemed annoyed but drove over to help me jump my car. I open the hood. I can’t find the battery, neither of us can. Where in the heck is the battery on this transit?!

Oh… it’s underneath the drivers seat. You have to literally take the drivers seat out to reach the battery. I apologize to my neighbor for wasting his time. It’s 7:58.

I figure I’d search the internet for how to jump a Ford Transit. Oh – there are specific jumper terminals in the main engine compartment. I found them. Turns out our transit is missing the very noticeable red jumper cable label that points directly to the terminal on most other Transits. It must’ve broken off. Had it been there my neighbor and I (first time I met him by the way) would’ve found it.

Great I found the terminals! But, my neighbor is gone…

I called my brother-in-law, he lives a few minutes away. Asked if he could come jump my car. He could. It would take him a bit to head over. My son probably isn’t going to get his sandwich.

Brother-in-law is here, I glance at the clock. It’s 8:15. Alright, I hope the car jumps. It totally does. Whew! I thank my brother-in-law profusely. He takes off, it’s 8:19. The kids get in the car. We’re out of there by 8:20am and I get the kids dropped off at school on time. My son doesn’t have a sandwich.

Yes, our kids are homeschooled. But for highschool they have a brand new co-op they are going to. They get 1 day of in class lessons from the teacher a week, and the rest of the week is homeschooled.

If only I would’ve “jumped” on a post from my boss 6 years ago 🤣

Anyways, after this whole situation I bought one of these Wolfbox things. Gonna be totally prepared next time. You never know when your toddler will leave your car lights on.

https://amzn.to/3JW5tYe 👈 (this is totally an affiliate link. 😆)
💡Heads up: I may earn a small commission if you buy through this link – at no extra cost to you!

Fixing ShareX Hotkeys: The Print Screen Conflict with Logi Options

Somewhat recently, I found that my ShareX hotkeys were not working. I tried a number of different things: re-installing ShareX, updating ShareX, and resetting the ShareX hotkey settings. Nothing worked.

And so, since I had more important things to do, I just left the situation alone and manually triggered the ShareX features via the menu when I needed them.

This morning, I attempted to use a ShareX hotkey again. However, I noticed something I’ve never noticed before—a little Logi Options logo appeared on my Windows taskbar for a brief moment. And that got me thinking… what am I doing to trigger Logi Options to show up?

So obviously, I opened up Logi Options. In the settings for my keyboard, I discovered that Print Screen is a configurable key. It was set, at some point, to Screen Capture. While that sounds good—it’s what Print Screen is supposed to do—this actually isn’t good! Screen Capture tells Logi Options to take a screenshot for you. It behaves almost like a native Windows Print Screen, and so it’s hard to notice.

In order to fix your ShareX hotkeys, you need to choose the Print Screen option. This tells Logi Options to get out of the way and treat a Print Screen press like normal.

RocketChat is the Worst

Ok. I’ve tried it. I’ve attempted to run a RocketChat server on digital ocean for a few years now. My goal was to provide an easy to use alternative to our families group text messages. Group text messages are terrible, they are unorganized, and often times people are left out of the conversation or added into the middle and left wondering “what the heck is going on?!”

I’d remembered looking into RocketChat years ago and decided to give it a try. My family was excited to try this new approach to communication. Right off the bat there was an issue. One of our family members attempted to change their profile picture via the RocketChat app. It didn’t work. So they decided not to use the program, just like that. I’ll say it again, in their first experience with RocketChat, it didn’t work, so they would not use it.

The rest of my family was more forgiving of the program. However, they too had issues. RocketChat was unreliable. My family members would find that they’d been logged out of the app, and so were not getting notifications. They’d have to log back in periodically. That’s annoying for them. So, I lost a few more users.

In addition to these issues, the RocketChat server will randomly stop. The result is that the app and web site will respond with Bad Gateway… Or just not load. I have to then ssh into the server to attempt to restart the service. Sometimes a simple reboot does the trick, other times I have to literally navigate to the RocketChat server directory and manually install npm packages. That’s crazy!

This morning I woke up early to work on a personal project I’ve been excited to work on. However, I spent all of my time trying to figure out how to get RocketChat up and running again. And, after a couple of hours of work, RocketChat still won’t start, and I’ve run out of the time that I had to work on my personal project, I spent it all on RocketChat. That’s incredibly frustrating.

So. I’m pretty much done with RocketChat. I won’t recommend anyone use it. It’s not worth the time investment required. It’s not user friendly. It’s the worst.

Hawaiian Macaroni Salad

Reproducing this recipe from the website “inspired2cook.com” which doesn’t seem to exist anymore. Here’s a link to the WaybackMachine if you want to see the original.

Ingredients

These will be used throughout the recipe

2 Cups Whole Milk (not low-fat because it makes the dressing too thin)
2 Cups Mayonnaise (not low-fat because it makes the dressing too thin)
1 Tablespoon Brown Sugar
1 pound Elbow Macaroni
1/2 cup apple cider vinegar
4 green onions, sliced thin
1 large carrot, peeled and grated
1 celery rib, chopped fine
Salt & Pepper

Make Dressing

Whisk the following in a bowl:

1 ½ cups milk
1 cup mayonnaise
1 Tablespoon brown sugar
1/2 teaspoon salt
2 teaspoons pepper

Cook Pasta

Bring 4 quarts of water to boil in a large pot. Add 1 tablespoon salt and the pasta and cook for about 15 minutes.

Make Salad Step 1

Drain the pasta and transfer to a suitable sized bowl
Mix 1/2 cup Apple Cider Vinegar into the pasta
Cool Pasta – I usually cover it and put it in the fridge for about an hour
Stir about half the dressing into the pasta
Cool Completely – I usually cover it and put it in the fridge overnight

Make Salad Step 2

Add the following ingredients to the cooled salad and stir

the remaining half of the dressing
4 Green Onions, sliced thin
1 large carrot, peeled and grated
1 celery rib, chopped fine
1/2 cup milk
up to 1 cup mayonnaise (play around with this, could be you don’t want any, depends on how you want the consistency of the salad)
Salt and Pepper (to your liking, I usually grind bunch of pepper over the top so that the top is nicely speckled ?)

GPG Signing on Windows with an SSH Key

Summary

I just experienced an issue which took me a day to figure out. So, as per normal, I’m going to document it here so that in the future I don’t have to bother looking it up!

My problem: every time I try to commit using git I am asked for my SSH key passphrase. However, I’ve ensured the SSH agent is running. I can pull from the remote, push to the remote, and do anything with the remote repository without needing to insert my SSH key multiple times. However, when I try to commit via git, I’m asked for my passphrase on every commit. Why is this a problem? Because when I rebase a 150+ branch of commits I’m having to enter my passphrase 150 times consecutively. That’s unsustainable.

As you probably know, Git Supports signing your commits with an SSH key these days. Christopher, over on dev.to has a great write up that can get you started.

Still, even armed with Christopher’s information, I couldn’t figure out how to fix the problem I as running into.

Since this post is meant to help guide me in setting this up from scratch (AGAIN) I’m going to go through all the relevant steps. But if you are just looking for the solution to the problem then skip to here.

Setup

Everyone’s setup is a bit different, but I’m going to document the relevant portions of mine.

Environment

Clean Install of Windows 11
Git
Note: When installing I select the “Use Git and optional Unix tools from the Command Prompt”
PowerShell Core
posh-git

If you want to automatically start the ssh agent, which I’d recommend, you can set the service to start automatically.

Restart my computer and I’m good to go.

SSH Setup

After my environment is setup, I set up SSH. This involved creating a new key, adding it to the SSH Agent, and making sure the agent was running. I run all these commands from an instance of Powershell Core (installed above).

Generating a New SSH Key

Since this is a new machine, I generated a new SSH key using the steps on GitHub

Basically the step is, use the following command, accept the default location, and use a secure passphrase. (1Password offers an excellent password generator if you need inspiration. I love the memorable password feature, XKCD ftw)

ssh-keygen -t ed25519 -C "[email protected]"

Adding the SSH Key to the SSH Agent

Adding the key to the agent is simple enough. You just need to run a command and give it the location of the private SSH key you generated in the step above. In my case that’s as simple as:

ssh-add c:/Users/YOU/.ssh/id_ed25519

Start the SSH Agent automatically.
Refer to: Starting the SSH Agent service automatically.

Now I would either typically restart my computer, or close the current PowerShell Core window and open a new one.

Adding the SSH Key to GitHub

If you want to sign commits with GitHub then you need to upload your SSH key 2wice (that’s a clever way of writing twice). One upload for your “Authentication Key” which will be for access to the repo. The second upload for your Signing Key. You can do both at https://github.com/settings/keys just click “Add Key” in the SSH Keys section and upload one for Authentication and then click “Add Key” again and upload one for Signing.

Signing Git Commits

Once again, referring to Christopher’s post. I do the following, making sure to switch out the path to my pub SSH key:

git config --global gpg.format ssh
git config --global user.signingkey ~/.ssh/id_ed25519.pub
git config --global commit.gpgsign true
git config --global tag.gpgsign true

This actually comes from Ahmad’s post on StackOverflow

git config --global push.gpgSign "if-asked"

Solving the Problem

Ok. Now the problem. You’ve successfully connected to Github and cloned your repo. Now you start making commits and find that you have to insert your SSH key passphrase over and over, despite the SSH agent running.

It took me a long time to figure out, but the problem is likely that you have two versions of SSH on your machine. The one that comes with Windows by default, and the one that came with Git for Windows when it was installed.

To solve this problem, tell Git For Windows to “Use external OpenSSH” when installing Git. The following comes from this answer on Stack Overflow by Ajedi32.

If you are using Windows’ native implementation of OpenSSH with the native ssh-agent Windows service, make sure that git for Windows was configured to use that SSH implementation when you installed it:

If you used the bundled OpenSSH installation, git will default to that and will not use any keys imported into Windows’ native ssh-agent service. You need to select “Use external OpenSSH” instead when prompted.

If you did not make this selection when installing, you should be able to fix that by just running the installer again.

Ajedi32

And that’s that!

Previous Versions of this article said the following. This is no longer what I recommend.

You’ll remember we selected the “Use Git and optional Unix tools from the Command Prompt” option. This option will add the referenced Git and Unix tools to your systems Path. The SSH Agent that is registered and used for Authentication is the one that comes installed with windows. The one that Git is using for signing, is the one that comes with git.

The order of the PATH variables will determine which version of the ssh-agent is used. The first version encountered will be the one that windows uses. So, to solve this problem, make sure the C:\Program Files\Git\usr\bin is above the %SYSTEMROOT%\System32\OpenSSH\ in your system path.