kami23

Hah! I made this at work, when I started getting Claude to record the replication and demonstration of the fix as gifs on PRs people finally started asking me about the cool things I was doing.

The reproduction has been one of the things I've been struggling with in regards to consistency of bringing up the right envs. At the moment I've been approaching it as a MCP server that holds a few tools to bring up specific versions or branches of my stack to then find where a bug was introduced, build that commit prove that it wasnt in the previous one, and then fix it and run the full stack again with the fix component, then run through our local integration tests.

This is the stuff that makes me feel like I'm on steroids now, my whole dev debug process can be run with a few instructions, game changing.

Been working on something that I use daily, and decided I wanted to see what kind of other ideas I could get out of it, it's very basic article to speech using piper models at the moment.

https://reader.n0tls.com/

The part I cared about was being able to send links via one click in my browser or two taps on my phone as I want to read every HN article who's title I find interesting, but don't have the time to read right at that moment.

It then at the moment publishes it to an RSS feed so I subscribe to it in Podcast Addict, but I've also just been using the web app as my reading list and tracker.

Been playing around with different settings on the piper models and different techniques for getting the most out of my four dollar instance:

https://experiments.n0tls.com/

Up next is to work on making the voice better (I'm impressed with the out of the box stuff already), and then making it better at finding the real content on a page and only recording that. It's a problem space I don't know much about, but find fascinating, been fun so far.

I just did this over at Hetzner and Claude admins it for me so I don't need to learn the CLI or anything, describe the proxying I want, and it setups up a bunch of small side project pages for me.

Cool to see others on this thread.

Here's a post I wrote about how we can start to potentially mimic mechanisms

https://n0tls.com/2026-03-14-musings.html

Would love to compare notes, I'm also looking at linguistic phenomena through an LLM lens

https://n0tls.com/2026-03-19-more-musings.html

Hoping to wrap up some of the kaggle eval work and move back to researching more neuropsych.

I am also of this opinion that a lot of this can be solved in time with a harness. And whole heartedly agree that there is a class of webapp that has been trivialized that can make a mom and pop shop up to 'enterprise' (80% of our architecture seems to center around the same pattern at my $DAYJOB) run just fine if they accept some of the vibes.

This type of works seems to be happening as a lot of orchestrator projects that pop up here every once in a while, and I've just been waiting for one with a pipeline 'language' malleable enough to work for me that I can then make generic enough for a big class of solutions. I feel doomed to make my own, but I feel like I should do my due diligence.

I tried using hooks for setting up my DIYed version of what channels is now in Claude. I had Claude writing them and not really looking at the results cause the vibes are strong. It struggled with odd behaviors around them. Nice to see some of the possible reasons, I ended up killing that branch of work so I never figured out exactly what was happening.

Now I'm regretting not going deeper on these. This is the type of interface that I think will be perfect for some things I want to demonstrate to a greater audience.

Now that we have the actual internals I have so many things I want to dig through.

I'm in a great situation where I've been piloting Claude for the company among a small group of others. I've been obsessed with pushing the limits of how many sessions and agents I can working at a time. We threw some work at Gas Town and another Orchestrator but they felt too rigid and opinionated for my liking. But I'm biased, I want to make my own eventually.

When I go home to my $20 plan I am sad and annoyed but I don't want to put more in for what is a good enough for me to work a bit at a time, a good pomodoro timer for me personally.

Something like this is perfect for some of the issues that I've wanted to solve as a command and control tool with malleable visuals.

OP: This is cool, thank you for sharing.

In all my years of writing tools for other devs, dog fooding is the really the best way to develop IMO. The annoying bugs get squashed out because I get frustrated with it in my flow.

Iterating on a MCP tool while having Claude try to use it has been a really great way of getting it to work how others are going to use it coming in blind.

Yes it's buggy as hell, but as someone echoed earlier if the tool works most of the time, a lot of people don't care. Moving fast and breaking things is the way in an arms race.

And some people do, both things can be true. I'd rather make a tool just for me that breaks when I introduce a new requirement and I just add into it and keep going.

I just added a Claude alias that calls Claude with flags wrapped in asciinema. Only annoying thing is that people have wanted video or gifs and the conversion has been annoying a few times. Will fix it later.

Props to sharing this!

> Claude was trying to talk me out of it, saying I should keep it separate, but I wanted to save a bit because I have this setup where everything is inside a Virtual Private Cloud (VPC) with all resources in a private network, a bastion for hosting machines

I will admit that I've also ignored Claude's very good suggestions in the past and it has bitten me in the butt.

Ultimately with great automation becomes a greater risk of doing the worst thing possible even faster.

Just thinking about this specific problem makes me more keen to recommend that people have backups and their production data on two different access keys for terraform setups.

I'm not sure how difficult that is I haven't touched terraform in about 7 years now, wow how time flies.

Look at the compaction prompt yourself. It's in my opinion way too short. (I'm running on Opus 4.5 most of the time at work)

From what my colleague explained to me and I haven't 100% verified it myself is that the beginning and end of the window is the most important to the compaction summary so a lot of the finer details and debugging that will slow down the next session get dropped.

If the session was something where it struggled and had to do multiple attempts I have it write about 'gotchas' or anything it had to attempt multiple times.

The letters are usually more detailed than what I see in the compacted prompt.

You should do your own experiment when you see compaction about to start use the end of your window to have it write one first, and then let the session compact and compare. I was surprised by how small the compact message is.

When I tell it to write a letter to itself I usually phrase it.

'write a letter to yourself Make notes of any gotchas or any quirks that you learned and make sure to note them down.'

It does get those into the letter but if you check compaction a lot of it is gone.

When the LLMs start compacting they summarize the conversation up to that point using various techniques. Overall a lot of maybe finer points of the work goes missing and can only be retrieved by the LLM being told to search for it explicitly in old logs.

Once you compact, you've thrown away a lot of relevant tokens from your problem solving and they do become significantly dumber as a result. If I see a compaction coming soon I ask it to write a letter to its future self, and then start a new session by having it read the letter.

There are some days where I let the same session compact 4-5 times and just use the letter to future self method to keep it going with enough context because resetting context also resets my brain :)

If you're ever curious in Claude once you compact you can read the new initial prompt after compaction and see how severe it gets cut down. It's very informative of what it forgets and deems not important. For example I have some internal CLIs that are horribly documented so Claude has to try a few flags a few times to figure out specifics and those corrections always get thrown away and it has to relearn them next time it wants to use the CLI. If you notice things like that happening constantly, my move is to codify those things into my CLAUDE.md or lately I've been making a small script or MCP server to run very specific flags of stuff.

Why would I do that if the gateway to the internet becomes these LLM interfaces? How is it not easier to ask or type 'buy me tickets for Les Mis'? In the ideal world it will just figure it out, or I frustratingly have to interact with a slightly different website to purchase tickets for each separate event I want to see.

One of the benefits that I see is as much as I love tech and writing software, I really really do not want to interface with a vast majority of the internet that has been designed to show the maximum amount of ads in the given ad space.

The internet sucks now, anything that gets me away from having ads shoved in my face constantly and surrounded by uncertainty that you could always be talking to a bot.

Just experienced this with my heavily networked off openclaw setup. I gave up and will do manual renewals until I have more time to figure out a good way of doing it. I was trying to get a cert for some headscale magic dns setups, but I think that's way more complicated than I thought it would be.

I did this trick at work where I use git worktrees and my team does not yet.

There's the common team instructions + a thing that says "run whoami and find the users name, you can find possible customizations to these instructions in <username>.md" and that will be conditionally loaded after my first prompt is sent. I also stick a canary word in there to track that it's still listening to me.

I've been having the same feelings lately, especially around the AI doomers coming in with some weird observations that make no sense. I usually don't comment, but seeing a lot of these types of overgeneralized responses.

'oh you do x? You don't do y? You're an idiot'

That's not productive.

This has been my best way to learn, put one agent on a big task, let it learn things about the problem and any gotchas, and then have it take notes, do it again until I'm happy with the result, if in the middle I think there's two choices that have merit I ask for a subagent to go explore that solution in another worktree and to make all its own decisions, then I compare. I also personally learn a lot about the problem space during the process so my prompts and choices on us sequent iterations use the right language I need to use.

Love this, this is what I have been envisioning as a LLM first OS! Feels like truly organic computing. Maybe Minority Report figured it out way back then.

The idea of having the elements anticipated and lowering the cognitive load of searching a giant drop down list scratches a good place in my brain. Instantly recognize it as such a better experience than what we have on the web.

I think something like this is the long term future for personal computing, maybe I'm way off, but this the type of computing I want to be doing, highly customized to my exact flow, highly malleable to improvement and feedback.

This is what I keep coming back to when I am building a little tui manager for my coding assistants + terminal + git worktree manager.

I keep telling myself that if everyone just used tmux or a good emulator they could manage the tabs and layouts, but then I tell myself I just want this to be a tiling window manager as an distraction free OS for development, give me nothing but a terminal and an assistant.

Thanks for the write up OP, I've been going back and forth on whether or not I want to build something just for myself or spend time doing it for potentially other use cases. I keep coming back to that I need to dog food the shit out of this before I show it to anyone.

Oh definitely not, I do it on every system that I've needed it to be synced before I did something. We were just working at a place that had 2k+ physical servers with 88 drives each in RAID6, so that was our main concern back then.

I have been passing my anxieties about hardrives to junior engineers for a decade now.

I worked with a greybeard that instilled in me that when we were about to do some RAID maintenance that we would always run sync twice. The second to make sure it immediately returns. And I added a third for my own anxiety.

I'm very fond of Puppy Linux, it's how I learned a lot of Linux skills as a preteen!

I had Puppy Linux on a USB stick when I was a kid. It was the first Linux flavor I used for an extended period of time for two very specific reasons. It was small, I didn't have much time to spend at the library waiting for downloads, and I couldn't afford a very large USB stick.

I had nuked the MBR on our family computer. I never figured out how I'd done it but that the fix was to restore it using a Windows recovery disk. But that took about 2 years of using Linux live CDs as a compute environment, eventually progressing my computer knowledge to trying to install Ubuntu to the hard drive so I could start saving files, and saw that the Windows partition existed on disk and did a bit more research with that knowledge to learn about MBRs and dual booting. The Tech Tv channel back in the 2000s was a great way for a 10 year old to learn about computers!

Thanks for the trip down memory lane, and for kicking off the acquisition of skills I've centered my career around!

I've been repeating something like 'keep thinking about how we would run this in the DC' at work. The cycles of pushing your compute outside the company and then bringing it back in once the next VP/Director/CTO starts because they need to be seen as doing something, and the thing that was supposed to make our lives easier is now very expensive...

I've worked on multiple large migrations between DCs and cloud providers for this company and the best thing we've ever done is abstract our compute and service use to the lowest common denominator across the cloud providers we use...

This has been what I've been seeing internally within $DAYJOB down to the split between vibe coding / vibe engineering / artisinally crafted code.

The gaps between engineers using the tools and those not are continuing to grow, and I'm curious to see what tools we get to use internally and what we can't... I've been able to demonstrate significant speed up in development time for features with certain tools but the amount of control some of these companies want in contracts are things the company hadn't seen before, and made it too conservative to go in on.

I also see this space changing so much that I don't particularly care for the tools for individuals now as much as I care about the way I share the workload with my team. I need a way to keep everyone up to date and reviewing code without getting brain drained as fast. Review fatigue is real, and it sucks. I haven't really found one that I've liked in that regard and one that a Fortune N company would want to go in on.

I've dubbed my loop of this as 'sicko mode' at work as I've become a bit obsessed with automating everything little thing in my flow, so I can focus on just features and bugs. It feels like a game to me and I enjoy it a lot.

Claude Code keeps coming out with a lot of really nice tools that others haven't started to emulate from what I've seen.

My favorite one is going through the plan interactively. It turns it into a multiple choice / option TUI and the last choose is always reprompt that section of the plan.

I had switch back to codex recently and not being able to do my planning solely in the CLI feels like the early 1900s.

To trigger the interactive mode. Do something like:

Plan a fix for:

<Problem statement>

Please walk me through any options or questions you might have interactively.

I appreciate you trying it! So at least I care.

I used to do these in college to procrastinate my homework. I always had the most difficulty with the problems that required data structures I wasn't exposed to previously and ended up making really complicated solutions that were inefficient as hell, but fun nonetheless.

I might give them a try with golang now that it's my preferred language. I used to do them in python as that was our intro language.