Posts tagged: ai Archive

Blog and LLM Knowledge Base

2026-05-13T06:23:42Z

Dev #

Thinking about this tweet from Andrej Karpathy about LLM knowledge bases. I maintain a blog like this one at work, which is fairly well structured markdown files and I use markdown lsp for linking docs conceptually. It makes it easy for me to trace thoughts and search what I’ve written about. However I find that there’s a limit to what I am able to organize by myself… The hardest part of of a zettlekasten to me is breaking down a thought into the smallest most atomic unit so that I don’t end up duplicating thoughts in files and lose the real traceability. In my head I think there’s a mathematical way to solve this, some kine of embedding setup that’s beyond my actual comprehension. But I think I’m overlookking that agents are probably just good enough with raw markdown, at the scale that I keep notes, that I should start to move in the direction, at least at work, of making my working notse blog repo more agentically friendly: AGENTS.md file, using an agent with its own dumping ground for files, some pruning sessions, and maybe a simple chat interface where I open an agent with a skill for using myrepo as a knowledge base and see if that gives me any super powers for organization or problem solving.

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

The Irony of 10x

2026-03-26T06:00:50Z

Opener #

Agentic coding has been an exciting change to me over the last couple of months specifically. I’ve been using AI tools for a few years now but something really shifted with Opus 4.5 as well as the tools/harnesses getting better and more useful around the same time. I’ve been influenced by people like Simon Willison and Steve Yegge who have been on the forefront of agentic coding and vibe engineering since the dawn of ChatGPT and in my small circles of work I’m definitely on the bloodiest bleeding edge of the adoption of these practices. Are they the future? I don’t know - I tried to maintain a skeptical posture but until the bubble pops it’s looking like this is at least a direction the future of my line of work is going.

Danger

The tough part is mixing the new world of agentic coding with developers on wildly different points in the spectrum of adoption and maturity.

Here’s a short anecdote about what I mean…

In a new project that I’m on, I’ve refactored/reimplemented a lot of legacy code, producing about 140k lines of code, configuration, and docs in about 2 months. It’s an insane amount of “product” and I’ve done it entirely with agents. But I didn’t do it with a handful of vanilla chat sessions like “Hey Copilot, reimplement this API, no mistakes”. I’ve been building out my own process for using LLMs effectively. I have no actual idea besides my own experience if what I’m building is useful, but it feels pretty good - agents for planning, building, reviewing, testing, etc. My “harness” includes specialized agents and opinionated development workflows to try to ensure that code is never one-shotted into production.

The catch though is that I’m working with a handful of developers and they are not early-adopters or aggressive experimenters with these new agentic tools. Most of them are still in Steve Yegge’s stage 2 or 3 of AI coding as he’s outlined in this medium article introducing GasTown. They open a chat session, and say yes or no to Opus. I’ve been in-between stages 6 and 7 for a while now - managing multiple agentic sessions that themselves run specialized subagents primarily for context management, although I don’t feel quite ready for full blown stage 8 (agents running agents running agents). So this isn’t me saying that I’m “better” than the guys I’m working with, just the project we’re on together is now being touched by people on wildly different ends of the agentic coding spectrum, and the project itself is an experiment to me of living in the “New World”.

Finally the anecdote - I’m getting PRs from guys in stages 2 and 3, thousands of lines of code and config, that I know they are not necessarily experts in, but they’re producing that code with fairly vanilla practices. How do I verify it? I know my own prompts, I know what agents generated my code, I know the direction and prodding I gave as they’re building, but all I know in a PR review from someone else is the diff - not how they tested it (unless of course there’s tests, and there must be, but tests are only as good as the tests are…), not the steering they gave, not the manual UAT validation they did, or even the full intent… And ultimately I own the code they merge because chances are I outlast their contract with the team. It’s kind of a scary thing to review and accept… do I trust my agentic practices to validate their work? Cause I certainly don’t trust myself to validate it perfectly.

a note about contractors

I’m not implying anything about anyone who works contract, I think the facts is that contractors are typically dis-incentivized to really “own” something - an unfortunate consequence of the world.

This is about me, not everyone #

I want to make sure to reiterate the context of this post and my thoughts - it’s really about me. I’m not saying anything here is true for anyone else, I’m not making predictions about tomorrow, and I’m not a prominent FOSS developer or a high-profile ex-FAANG engineer. I’m not talking about every project under the sun. I work at a big company, not a fast-paced startup. I work on internal tooling, nothing that should even be seen outside our network. I’m also a self-taught developer, not a trained software engineer.

As far as AI goes, lately I’m mostly all in on Github Copilot CLI with a sprinkling of Opencode. I’m not using Claude Code or Codex which seem to have their own communities around plugins and usage.

And since we’re so focused on me right now, the important thing to keep as the backdrop for this post is my temperament and make-up. I’m definitely made a certain kind of way, not different than every single person necessarily, but different enough, from enough people, to not blend in with the 9-5 guy who can log out and not think about work until the morning (I’m not saying that’s a strength either, my boundary problems are for another time).

I feel an aggressive burden to solve problems and own those solutions, call it white-knighting or a savior complex if you want, but I’ve got enough of a reputation at Caterpillar now (and anywhere else I’ve worked) to become a go-to person for more than I think is necessarily appropriate. And with that burden I’ll bring results - if a problem hooks into my brain it. will. be. solved. I probably won’t do the best most clean-code solution right out of the gate, but I’ll do whatever I can to find a solution.

Who I was before agents #

That is who I was before agents made code easy to produce. My first boss at Caterpillar used the phrase “tenacious learner” to describe me in several reviews. I kind of rejected the description because I basically refused to believe that I was really any different, any harder of a worker, than my peers.

But with almost a decade of experience in the corporate world, and some adult perspective on my life, I think it’s accurate… I am a harder worker, to my detriment sometimes, than a fair number of people I’ve worked with**.

I’m not brilliant but I can focus for a long time in the right circumstances. The gift of perseverance (or the curse of not being able to let something go, depending on how you look at it) has led to blessing in my life in both reward and skill.

\*\*

There’s nothing wrong with it either, Cat’s somewhat noticed that work in my EOY reviews and I’m certainly not against “just doing your job”.

And then AI came along and with another set of the right circumstances catalyzed a new way to work.

Leaning into “Agentic Engineering” #

I jumped onto experimenting with AI coding tools as soon as they became available, but mostly I just tried vibe coding rather than using tools for real engineering work. I vibed up an API at one of my jobs that went into production way too early, with far too little validation, and it was scary to support it from then on out. I also did the meme, vibe-coded a TODO app, and threw that puppy into the internet without locking down my API endpoints… That was before agents were quite as useful as they can be now, but that experience along with a handful of other stepping stones (like learning some real actual fundamentals about security) began to give me confidence in using the AI as a tool, like my IDE is a tool, for producing solutions that take the form of code.

Note

As one-shot apps got better and better, and as I learned about scoping work more appropriately for agentic tools my confidence in them grew.

There’s quite a difference between “Claude make me a todo app, no mistakes” and scoping out a solution in natural language, with some technical guardrails, and having agents tackle the implementation methodically.

What actually changed #

What’s actually changed for me is quite a lot… I haven’t opened my IDE to seriously write code for months now. I’ve oscillated between Opencode and Copilot CLI, leaning moreso into Copilot since it’s an approved tool at work and as of mid-February is quite good. Mentally I’m approaching problems with a little more thought on the front-end than before because prior to agents I would think as I implemented. At the scale of work that I do, this was really fine - working on CLI utilities to solve simple problems, developing an iterative testing cycle for each problem that allowed me to move fast, and once I found a groove I was cooking. But now I don’t even need to find it, I open Opencode or Copilot CLI with my Planner agent, describe what I want to happen and have Opus or GPT scope out a plan for me. Usually there’s some back and forth on feature scoping, then I review a markdown file it produces, and once it looks decent enough to me I say “go” and it goes.

That works a lot better than I even care to admit because at the same time as I’ve been leaning harder into agents, I’ve been building my own harness of sorts - not a replacement for Copilot CLI or a competitor to Opencode, but moreso an opinionated workflow spine that I force agents into to give strict gates to the SDLC (software development lifecycle).

Problem Solving Workflows

Plan and implement is fine for a lot of things, and I do think it’s only getting better. My harness, mentioned a few times around here before, called Nexus, is a set of agents and rules that I want the code I’ll be responsible for to go through before it lands in production. That cycle isn’t too complex, and there’s only about 10,000 similar tools to Nexus on Github trending right now. I’ve thought about dropping my idea and picking up something more popular, like superpowers but at the moment I’m continuing to develop on and lean into my own idea here.

Mini post on Nexus

I keep saying a blog post is coming, but the high level of Nexus is that it’s a task tracker with a CLI that agents use to advance a ticket through a plan -> build -> test -> review -> verify -> merge lifecycle that is almost exactly how I would otherwise have solved a problem by hand. I think it needs work, I need to be harder on TDD methodlogies with agents, and work on verification gating a bit more (shoutout to showboat by Simon Willison) but overall it’s a system of thinking that I already participate in so I’m doing my best to farm out specific parts of my workflow to agents rather than trying to one-shot enterprise problems and solutions.

Who's doing the thinking?

I’ve noticed that as I’ve developed Nexus out though, I lean on the agents for more and more of my own thinking, and am trusting my problem solving process moreso than my actual problem solving abilities.

Hidden costs #

The cost of this increase in speed is a lack of familiarity - and the fallout of lack of familiarity is hard to express. There’s also many facets to it. For me, the first facet is that Nexus helps me move fast, but as I’ve leaned into it for more and more of the planning, I’m less and less familiar with the state of the code. I find myself asking my reviewer agents in fresh sessions often to explain it to me, and thankfully they’re usually consistent, but nonetheless I’m still not intimately familiar with the code. And on Nexus it’s not a big deal, that’s low stakes, it’s just me and my workflow.

I’m using Nexus + Copilot at work and that feels like higher stakes… I have my agents explain the status of our project and although they’re also somewhat consistent the thing that’s scary is that other people are working on that repo with me, and that’s where another layer of complexity manifests itself. If it’s just me and my Clankers, let’s go all day long, rebuild, ask questions, etc… but I have other developers I rub shoulders with now, and if they ask me a question what am I going to say? “Hold on, let me prompt my agent for you” - it’s LMGTFY on steroids. And the burden becomes if I feel like I can own and support what those other developers push into the repo.

Murky responsibility boundaries #

Why do I own their work? Well for the third time, this post is pretty self-centered and all about me, and my situation is that the other developers I work with presently are all contractors. Their work agreement with Cat could end at any second, for practically any reason. The incentive structure isn’t there for these guys who technically work for an agency… Their bonuses aren’t bigger (or even exist) if Cat performs well, there’s no extra vacation days in it for em (aye, contractors don’t get vacation days anyways), and not that it’s a problem, moreso just the nature of the world we’re in - but they’re basically mercenaries out to the highest bidder and I happen to know of multiple times where Cat lost a good person to a higher bidder.

So this isn’t really me trying to be negative about contractors at all, I’m here for a pay-check as well but Cat at least gives me SOME incentive to work hard with the goal of compensation regardless of how altruistic I feel in my own circumstances.

Incentive

Better ratings mean marginally better end-of-year salary increases, and I’ve received some other awards that certainly give me pause about jumping ship to another long-term place even when things can be crazy at Cat.

It’s more than just the contrator-ownership dilema, I’ve dug myself quite a hole over the last 8-10 years, gaining a reputation that I think many would appreciate, but for me only lately increases the stress. I don’t need to parrot every accolade I’ve ever received, that’s not the point, but to make the point as clear as I can - I have a lot of respect from quite a few people at Caterpillar. I’m blessed to have that reputation, and it’s not like I haven’t worked hard for it - but people talk about me in a way so flattering I feel like the main character in a fictional story sometimes.

In a fictional story I can check out the ending, hit up spark notes, or ask AI how it ends… but there isn’t an “end” in my real world scenario, there’s only tomorrow and I feel the pressure of not knowing what tomorrow holds now more than ever.

Danger

Being noticed is starting to feel more costly than rewarding…

Financial irony #

What’s the cost? It’s hard to get specific without writing a novel but here’s the TLDR - because I’ve been pretty good at what I do I’ve been able to do this type of work outside my normal 9-5 responsibilities and with that extra work has been some pretty great financial benefits. However with Cat changes, responsibility increases, and now owning code that others (and their clankers) write, the extra time I gained for myself is eaten-up and has been reclaimed by the mega-corp… “Exceeding expectations” every year just meant the bar is raised, the expectations are higher, the time-commitment requirement is higher, and as I’ve had to meet the requirements of both the new world and the curse of being noticed, I’ve lost the time for the extra work… For years I’ve realized the benefit of my own skills and drive, but the irony of agents (and a handful of other things) is that with the dramatic increase in expectations, not only on me but on those I work with and therefore their output, I don’t get to realize the benefits of my own gifts anymore.

Meaning and fatigue #

I feel very torn because the work I’ve been called into with Cat is good, I said in Cat Autonomy 2.0 that autonomy will save people’s lives. I love getting to participate in that mission, it’s the primary reason I didn’t jump ship to try to maintain the levels, and type, of work I was doing before… But in a few short months the mission is being drowned out by expectations and requirements that are so high I’m losing the grip on my own life.

Open questions #

That leads me to questions that I can’t answer, the question I ask daily now of “What about tomorrow?”. What will agents do for us tomorrow, what problems will be solved, what bugs will I create (by agents of course because I’ve never written a bug by hand in my whole life \s). If I stopped using agents would people still be impressed? Would it even matter?

Fin #

I’m certainly not anti-AI, it’s typing all my code. I’m not anti-collaboration, although I do wish I could work alone with just my clanker-army to worry about. I’m not sure what I am anymore though… AI has changed how I work, what I work on, and who I work with… Everything has changed in such a short period of time and like the ending of this post, it’s pretty jarring.

Danger

Death comes to us all - James Acaster.

Thanks for reading.

To Live In A World Without AI

2026-03-22T14:50:38Z

I’m finding lately that I wish we could go back to pre-ChatGPT… A world without a code-gen easy button, where “easy” was LSP autocomplete, where tools were at my fingertips rather than remote inference endpoints, and where I and everyone I worked with, was paid and judged based on what we could produce with our own 2 hands, even if ctrl + c and ctrl + v was sometimes a common set of keys… it sure beat answering “yes” to a robot every 38 seconds.

Oration on Agents and being crushed by responsibility

2026-03-16T20:18:23Z

The intention of this note is to become a blog post:

Raw Recording #

Okay, this is the oration for a blog post, maybe an open letter of sorts, as I am just wrestling with my new job and the responsibilities that either are put on me or that I’ve put on myself as the world is changing. And software development and software engineering and code writing is just changing, thanks to AI, thanks to agents and their introduction into the corporate environment. The internet is full of opinions and ideas that push and pull and prod people in different directions. And every one of those ideas has contexts, which is necessary to keep in mind when filtering the ideas. At a smaller startup where you can have a $200 a month Claude Code subscription for all of your engineers, that’s gonna yield certain results that’s going to encourage and invite certain people into that space to use those tools, the clawed tools, or Gemini or Codex. I work at a big company, Caterpillar. We have a GitHub co-pilot, but through co-pilot we have access to the kind of frontier large language models. From there, we have access to the co-pilot harness, whether it’s co-pilot CLI or the co-pilot integration with Visual Studio code, or the other IDEs like JetBrains, et cetera. Out there in the world, he has certain opinions from some people. You have people who are very pro agents like Steve Yeage, who has been a big influence to me over the last several months, as I’ve definitely jumped into the deep end of agentic coding. I haven’t even really opened my IDD in months, except for small things that I want to do by hand. enterprise grade work for the last three months or so and maybe a little longer has been fully written and produced by agents and i’ve been using AI to write code to some capacity or another since chad gpt came out really since like claude sonnet 3.5 but this thinking that i am doing is It’s really supposed to be about me, so this blog post I want to write is me processing thoughts on the expectation change that’s placed on developers in certain environments for certain types of work. I guess even that, like that sounds kind of vague. So I can only speak to myself. So I need to put a big annotation and plaster this blog post with this big caveat message that this is me, this is my experience, this is my position, and this is not a truth that is wide reaching. It’s not any kind of prediction on the, you know, the larger change happening in the world. This is just where I am at. So maybe for the post to make sense, I need to give a brief amount of history about me and my work. I’m not a trained software engineer, but I’ve been, I guess I’ve been a pretty impressive asset in my employment. I’ve been at Caterpillar for nine years, and in that nine years or over in the last several years, I’ve moonlit, moonlighted a little bit. And at every place I’ve done any other amount of work, I’ve just consistently impressed people. Now, this really just leads me to have more imposter syndrome because I know or I have learned what impresses people, And it isn’t necessarily excellence, although I try to do things well and I want to solve problems and maintain the systems that I build to solve those problems. But I don’t write the best code. I’m not the smartest developer. I don’t know the most. I get to know the things I work with and I learn a lot and I work fast because I have Not a super unique, but a rare ability to focus in, when the stars align, to focus and produce more than the next person. Not every next person, but a lot of the next people, a lot of the people I work with, I outperform. And that’s been true, or it has seemed true, from every review, pretty much every review I’ve ever received, that’s been true. And as I’ve had that just personal truth be reflected in my work, it’s been enhanced with AI because I’ve been an early adopter and even as a, even as somewhat of a skeptic, like I’ve tried to maintain a posture of skepticism. But since Opus 4.5, Even though I might talk like a skeptic, I am leaning into agents hard. And I am practicing my own harness engineering with my nexus project. And I’m using it to do engineering work at Caterpillar that’s going into production for us, which is like data platform type of work. It’s not custom, it’s not external customer facing. It’s internal tooling, which is what I mostly built. And the thing I’m really trying to wrap my head around, or I guess I’m trying to explain, is without AI, I was out producing a lot of people. And with AI, you can produce just a shit ton of code. And I’m trying to do that part well, but in my own gifts and curses of hyper focus and Wanting to meet expectations and solve problems, I’m jumping into the new frontier very quickly. But something I’ve noticed is that as I’ve produced a lot, people are just impressed. And they’re not necessarily impressed with the excellence. They’re impressed with how I talk about things. They’re impressed with how much I do, and they’re impressed with the output they see. But the result of a bunch of code isn’t all of the code itself. And I haven’t read all the code I’ve produced with agents. I’ve certainly tried to test most of it or had most of it tested. But as I’ve sought to do this and as I’ve been working in this capacity of using agents very heavily, I just feel kind of torn. in trusting the agents and in working with the agents in the ways that I think are appropriate, but also collaborating with other humans who are also using AI. But are they using AI in a trustworthy way? I don’t know, and I’m not sure how to validate it. But my managers seem to think I know how to validate it, and they seem impressed with the work I do, and they seem impressed with my understanding of AI and agentic coding. And so I feel all of this pressure to continue to be a top performer like I have been since before AI. I feel a lot of pressure to, I don’t want to coach people in how I do things, but I’m being asked to coach people and how I do things because other people I work with notice me. And I guess I’m almost, I’m just kind of tired of being noticed. Because being noticed is not resulting in me becoming rich. And money is not the most important thing. But man, I’d love to just make enough money to fuck off for the rest of my life and never look at a computer ever again. I would love… to not have to continue to maintain systems that I build or that other people build or to collaborate with people I don’t know, who might not even be long term. That’s the one thing about Caterpillar is I’m working with a lot of contractors. And nothing wrong with contracting, I’ve done that. Nothing wrong with being a contractor, but They could disappear at any moment, and all the code that they write, and even today, all the code that they prompt and have an agent right becomes my problem, my maintenance issue, my responsibility. And it just is, It’s just making me nervous to maintain and to understand and to continue to work under conditions that I can’t even tell if I’m making them for myself or if they’re real expectations. Like if I fell off the face of the earth or if I dropped agentic coding entirely and I just went back to coding by hand completely, would my output continue to impress people today? I don’t know the answer to that. But I certainly wouldn’t be solving problems as fast, but would I be creating less problems? Like I’ve solved a lot of problems in three months in my new role here with Forge. But I don’t know how many problems I’ve created. Only time will tell. And the thing is, the guys I work with who are contractors, they might not be around long enough for time to tell, but because I’m not rich yet, I probably will be long enough, or be around long enough. to experience those bugs and those problems that I don’t know about yet. And I am just feeling so much pressure to be the best and to be the one to solve problems. And I don’t know if I’m putting it on me or if cat’s putting it on me or if my project manager is putting it on me. I’m just really struggling with the amount of responsibility I feel. And I’m feeling frustrated about it because that amount of responsibility feels like it has scaled a lot with agents, with AI, but not my paycheck. And in fact, my paycheck has gone down quite a bit. And I’m not more relaxed. I’m making about half of what I’ve made over the last couple of years. And I’m not more relaxed. I don’t feel freer. I don’t feel like I have 50% of the responsibility that I did. Cal Autonomy awoke something in me for purpose, but that feels like it’s already starting to die. I’m already tired. I’m, I don’t know about burnt out, but I’m tired. I was reviewing some code that one of these contractors I’m working with. It’s actually only an intern. He’s not even a contractor. He’s an intern. And I’m reviewing a pull request of his. And damn, it’s a lot of code. And it looks somewhat useful. But I don’t know exactly how to test it. That’s not true actually, I do know how to test it, but I don’t own enough of our infrastructure at Caterpillar to test it easily. So do I trust him? Do I trust his prompt? Do I trust his harness? Do I trust the tests that I have already put in the code base to capture and block any breakages he might have introduced with this code for the code I’ve already been producing? It’s different when I’m intimately familiar with the code base. And I’ve been, I don’t know, definitely two other places at least where I had very intimate familiarity with the system and with the code and with the platform. But the one I’m on right now is new, and I built it. I built it with agents. I built it to replace a thing that was never even working. But I built it. It’s mine. It’s not somebody else’s. It’s not that we’ll document it. There’s a lot of documentation, but it’s all been agent produced, and even though I’ve reviewed it, it’s hard for me to organize it in the right way in a way that makes sense. I’m just feeling so frustrated that it seems like, as the tools are getting better, as developers are able to move faster, it feels like I am expected to move faster. but I’m not expected to make any more money. I’m not expected to be compensated anymore for the increase in velocity and responsibility. It’s almost like if I could produce, this metric is not gonna be real, but let’s just, for the sake of argument, make it up. If I can write 30,000 lines of code in a year and make $100,000, I feel fine about that, whatever that number means. But now I can actually produce 300,000 lines of code in a year. I’m definitely producing 10 times more code than normal. I don’t know if I’m doing 10 times more engineering, but I am making exactly the same amount of money. And so why did I do it? Why am I leaning into it so hard? Why am I trying to still be the front runner to be on the bleeding edge when caterpillar is not going to pay anymore? And I could go somewhere else. I could go to Avant. I could make more. It’s a system I’m familiar with. It’s even people I know. And there’s AI there too. But it’s a system I know. It’s a system I know how to support and maintain. And there’s not an expectation that I’m going to produce 10,000 lines of code this week. But right now at Caterpillar, it feels like I am expected to write 10,000 lines of code this week. And deploy it and maintain it and solve new problems. All in an incredibly political and hectic, chaotic environment where there isn’t an appropriate amount of leadership over me. I am the expert in AI in my group. And that feels retarded. I am the professional agentic engineer and that is just scary because I’m not good enough to be that. I’m not Steve Yeagie, I’m not Simon Willison, I haven’t been a developer for 20 years. I’m just a guy who learned some Python to solve some problems and happen to work faster than a lot of people and that reputation has really It’s treated me well for sure. And my disposition to problem solving has, up until recently, I felt like I had been, like I’ve been compensated well by my company, but then I’ve also been able to do more work outside of one job for me to feel the value of my own skills. But now with AI changing everything, Can I show you what’s on the back of my story? It feels like I’ve given up the personal gains. One second, Alia. So I’ve given up personally realizing the benefit of my own skill. And now I am like in a position where Caterpillar, and only Caterpillar, is reaping the benefit of my new scheme in agentic engineering. And I think that’s just ticking me off because I stayed with Catt to do something good for the world. And I think we can do something good for the world. But I’m tired.

Notes and Summarization #

Possible Titles #

Agents, Responsibility, and the Cost of Going Faster
The Irony of 10x
More Output, Less Upside
When AI Made Me Faster and Poorer
Being Crushed by Responsibility

Working Angle #

This should probably be a personal essay, not a manifesto about AI. The core tension:

agents made me dramatically more productive
that productivity increased the responsibility I carry
the financial upside I used to realize personally has gone down, not up

Section Stub: Opening Scene #

Prompt:

Start with a concrete moment. Maybe reviewing a massive PR from the intern / contractor side of the system.
What did I feel in that moment: confusion, dread, pressure, responsibility, skepticism?
What was the specific thought: “I might own all of this later, even if I did not really produce or validate all of it myself.”
Can I open with one sharp sentence that names the tension right away?

Possible opener idea:

“Agentic coding has made me faster than ever, but it has also made me feel responsible for more than I can honestly hold in my head.”

Section Stub: This Is About Me, Not Everyone #

Prompt:

Add a strong caveat early: this is not prediction, prescription, or universal truth.
Name my context: big company, internal platform work, political environment, mixed team ownership, frontier models via Copilot.
What about my temperament matters here? Hyperfocus, initiative, problem solving, pressure sensitivity, desire to meet expectations.
What would I want a reader to not mishear?

Section Stub: Who I Was Before Agents #

Prompt:

Briefly describe my work history and self-understanding.
I am not formally trained, so what made me valuable anyway?
What does it mean that people were impressed before AI too?
Pull in relevant threads from pages/blog/forge-ahead.md: prior projects, practical problem solving, being useful, learning systems by doing.
What gifts do I think God gave me that made me effective before this shift?

Questions to answer:

What kind of worker was I before agents entered my workflow?
What did I already know how to do well?
What part of my value came from focus rather than brilliance?

Section Stub: Why I Leaned Into Agentic Coding #

Prompt:

Explain that I did not stumble into this accidentally; I chose to go hard.
Why did it feel natural to me to adopt agents aggressively?
What about the mission at Caterpillar Autonomy made the leap feel worth it?
Pull in pages/blog/cat-autonomy-2-0.md: I stayed because the work felt meaningful, not merely because it was interesting.
Mention harness engineering / Nexus / experimentation if it helps show this was a deliberate investment.

Questions to answer:

Why did I trust this direction enough to reorganize my work around it?
Was I chasing novelty, usefulness, recognition, mission, or all of the above?

Section Stub: What Actually Changed #

Prompt:

Describe the practical change in my day-to-day work.
I am no longer mainly writing code line by line; I am directing, reviewing, testing, and integrating generated work.
What changed about the quantity of output?
What changed about how coworkers and managers perceived me?
What changed about the kind of questions people started asking me?

Questions to answer:

When did I realize this was not just autocomplete, but a new mode of working?
What does “enterprise-grade work fully written and produced by agents” actually mean in practice?

Section Stub: The Hidden Cost of Speed #

Prompt:

Write about the gap between output and understanding.
I can ship more, but I do not always know the code as intimately as hand-written systems I built before.
Testing is not the same as ownership.
Documentation existing is not the same as confidence.
What kinds of future problems does speed create?

Questions to answer:

What do I know I have not fully read?
What kinds of bugs or maintenance traps am I afraid I am creating?
What does it feel like to build something that works but still feels partially foreign?

Section Stub: Responsibility Without Clear Boundaries #

Prompt:

This may be the emotional center before the money section.
Describe the contractor / intern dynamic and why it intensifies the burden.
Their output can become my long-term maintenance problem.
The organization sees me as someone who understands AI and agents, so responsibility pools onto me.
Ask whether the pressure is explicit or self-imposed, but note that the body feels it either way.

Questions to answer:

Who is actually accountable when AI-assisted code breaks six months from now?
What does it mean to be treated like the local expert when I do not feel like one?
Why is being noticed beginning to feel costly instead of rewarding?

Section Stub: The Financial Irony #

Prompt:

State this clearly and concretely.
For years, I realized more value from my output because I worked two jobs concurrently.
Due to the new role, new responsibility, and the demands of agentic coding in this environment, I had to go down to one job.
So the irony is: the tools make me dramatically more productive, but I personally realize less economic upside from my own gifts and skills than I did before.
Keep this grounded and honest, not whiny.

Questions to answer:

What exactly have I lost financially?
Why does this feel bigger than “I want more money”?
Where do the gains of AI go when an individual contributor gets faster?
If I can now produce 10x more code, why does my life not feel lighter, freer, or better compensated?

Possible thesis sentence:

“Agentic coding increased the value I create at work, while decreasing the share of that value I personally get to realize.”

Section Stub: Mission, Meaning, and Fatigue #

Prompt:

Reaffirm the good: this is meaningful work, and that matters to me.
Tie back to autonomy, safety, and the reasons I stayed at Caterpillar.
Then tell the truth: meaning does not cancel fatigue.
What has already started to feel tired, dulled, or overdrawn in me?

Questions to answer:

What purpose did this role awaken in me?
Why does that sense of purpose already feel threatened?
How do I talk honestly about fatigue without pretending I am fully burned out if I am not sure?

Section Stub: Questions I Cannot Yet Answer #

Prompt:

Let the essay breathe here instead of forcing a neat conclusion.
Gather the real unresolved questions.
If I stopped using agents, would I still impress people?
Would I solve fewer problems or create fewer future ones?
Am I carrying real expectations or anticipated ones?
Is this making me a better engineer, or just a faster source of output and responsibility?

Section Stub: Conclusion #

Prompt:

End modestly, not with a manifesto.
I am not anti-AI.
I am not saying agents are bad.
I am trying to tell the truth about what this transition has cost me personally.
Land on the mismatch between speed, responsibility, and reward.

Possible closing direction:

The tools are real.
The gains are real.
The responsibility is real.
The redistribution of value is also real.
I am still trying to understand what faithfulness looks like inside that tension.

Material To Weave In #

My lack of formal software engineering background
My history of being useful by solving practical problems quickly
My tendency toward hyperfocus and over-responsibility
The shift from hand-written confidence to agent-mediated uncertainty
The contractor / intern maintenance concern
The drop from two incomes to one
The irony that Cat benefits more from my increased output than I do
The Christian framing around calling, providence, stewardship, and fatigue

Things To Avoid #

sounding like I am making universal claims about all AI use
sounding bitter without being specific
centering line-count metrics too much without translating them into responsibility
pretending I have fully landed on a conclusion
making this a complaint post about coworkers instead of a reflection on burden and change

My Thoughts on Beads

2026-03-03T05:00:47Z

Steve Yegge is a pretty well-known individual in the tech field, having been around for a long time at some of the larger companies. He’s making quite a splash in the agentic coding world as well. I’ve appreciated Steve’s posts and projects lately and wanted to put some thoughts on one called beads.

Beads #

Beads is an issue tracker with links - issues relate to and block each other, but agents use beads to keep track of information and dependencies without storing it in their context 100% of the time. It seems like a very popular and useful tool - but I am not using it, and that’s what I wanted to capture… why not?

The answer for me is about where the organization layer is for the developer. Beads exists in a single repo - it’s a system-wide CLI but you ‘bd init’ in a git repo, and beads uses the .git/ folder, worktrees, dolt, and some git hooks to operate within that git repo. Outside the repo, it takes another tool to tie together all the beads databases you might have.

For me, I’m hardly “in” a git repo anymore. My workflow is that when I have something to work on, I create a “workspace” (self-defined concept) which is just a folder on my filesystem where I check-out git worktrees from any of the repos related to the work I’m doing. Sometimes it’s 1 worktree from 6 repos, sometimes it’s 6 worktrees from 1 repo for parallel work…

So because I like to organize myself in this way, beads is already “out” for me. That’s the main reason - I don’t have any real technical issues with beads or any criticism, it just is designed for a workflow that is not how I work.

This is why I’m building Nexus, something I hope to be able to put out there “soon”. It won’t be as general-purpose as beads, but my goal with it is to be plug-and-play for any agentic harness (copilot cli, claude code, opencode, etc.). It’s a challenge thinking about it as a personal tool but also as a tool to share someday, but agentic coding is making it possible to make some cool shareable stuff and I’m excited for my own workflow-task-manager to mature and at least become something useful to me (it already is, but building the plane in the air makes it kind of hard to enjoy the plane).

Credit #

banner image from ChatGPT

💭 It’s rude to show AI output to people | Alex Martsinovich

2025-07-21T11:46:43Z

Here’s my thought on It’s rude to show AI output to people | Alex Martsinovich

Great read on LLM output, and the slop that’s already been dumped into the internet. I’ve become a pretty heavy AI user - mostly for code and then thoughts organization. With code (that matters) I review everything AI makes. When I ask gemma3 via ollama to organize my thoughts, I reiterate the importance in my prompt to not re-word things, but to just organize and then help me work through whatever I’m working through in the post.

I’ve generated a blog post or 2 with AI only, and it’s still relatively clear to me when AI writes vs when a human writes. This post does a good job of highlighting how 1. this is a hilariously old sci-fi problem (there’s nothing new under the sun, ammirite?) 2. how AI is a useful tool and 3. how in the age of AI we’re still actually just battling other humans…. but humans are simple creatures in a lot of ways - we take the path of least resistance the vast majority of the time, and LLM sloppy output that looks not shitty is certainly a low-resistance path to a post or commit… but was the post or commit worth anything?

without proof of thought - no.

Note

This is one of [[ my-thoughts ]]. I picked this up from Waylon Walker(https://thoughts.waylonwalker.com). It’s a short note that I make about someone else’s content online. Learn more about the process Thoughts

💭 Anthropic Is Bleeding Out

2025-07-19T10:44:56Z

Here’s my thought on Anthropic Is Bleeding Out

Ah… so things are rough on the business side according to Edward Zitron

An intere|ing note is that anthropic models make more more than half of all token related to programming through OpenRouter… so they clearly have marketshare but if they’re losing money on every single customer then oof… what does the future of vibe coding look like? And who can afford it?

Note

💭 AI INTERVIEWS ARE HERE!! SO I TROLLED ONE TO SHOW YOU… - You…

2025-07-17T10:41:04Z

Here’s my thought on AI INTERVIEWS ARE HERE!! SO I TROLLED ONE TO SHOW YOU… - YouTube

AI has provided a lot of value in my life for sure… but it’s still not my favorite piece of technology, and the nearish-term-future certainly has some dystopian qualities to it. AI interviews definitely are in that category and Josh does a good and comical job of highlighting the ridiculousness of it.

I envy no one looking for a job, and I double envy no one who has to deal with AI powered HR solutions…

Note

💭 OpenAI’s Windsurf deal is off — and Windsurf’s CEO is going to…

2025-07-12T10:48:05Z

Here’s my thought on OpenAI’s Windsurf deal is off — and Windsurf’s CEO is going to Google | The Verge

Welp I’m not upset the OpenAI deal fell through (I mean probably not great for those at Codeium, but I don’t want these big tech firms to keep buying out the competition… So even though I’m not upset about it - I can’t say going to Google is my favorite alternative… I’m a windsurf user at home but if the Google deal goes through I may drop windsurf entirely and stick to roo…

Look at me care about who owns my editor when I’ll just root myself deeper into vs**** #smh

Note

💭 Docker Brings Compose to the AI Agent Era | Docker

2025-07-11T11:49:41Z

Here’s my thought on Docker Brings Compose to the AI Agent Era | Docker

Seems like docker is leaning harder into compose - which is great for me as a heavy compose user. I had heard about some of the LLM enablements directly through docker desktop - for example taking fuller advantage of the host’s compute power without paying the penalty of the nvidia runtime or something… I can’t claim to have followed it all but I heard “run LLMs in docker directly for a bigger boost” and I’m in.

Note

💭 Notes – 06:11 Sun 22 Jun 2025 – David Bushell – Web Dev (UK)

2025-06-25T11:04:29Z

Here’s my thought on Notes – 06:11 Sun 22 Jun 2025 – David Bushell – Web Dev (UK)

I see that most of the tech world, at least in public, leans left… which is fine. I think politically I’d have leaned left in the 90s/00s if I had been old enough to pay attention and understand.

But the sold-out hatred for Elon and Trump is wild to me… Nothing makes the reality of propganda in 2025 more clear to me than the manufactured hate of 2 guys who the Left USED to love… I’m not like all in on either guy - but the politically driven hatred which has led to physical violence against both of these guys is insane….

Dave’s ending thought here “We’re entering a Misinformation Age” is funny to me… Misinformation has been around forever - the CIA has admitted to an absolutely hilarious amount of things that over the years the government (US Government) had denied and the news had been complicit in covering up - right up until it was just undeniable and finally “ok ya everyone we did do that mind control experiment - BUT WE’LL NEVER DO IT AGAIN”… “ok ya everyone we did know the Taliban has been on the receiving end of millions of US dollars every year but don’t worry about it”… But for everyone to NOW get all up in arms over 2 guys the news will not shut up about, is just comically out of touch with history to me…

Note

💭 Copilot Instructions

2025-06-23T11:14:02Z

Here’s my thought on Copilot Instructions

I have tried a few different cursorrules/windsurfrules with windsurf, but I didn’t even know you could give copilot custom instructions. I see python + fastapi on here, which I build with frequently at home, so I’ll look forward to seeing if the results are any better with a custom prompt using copilot

Note

💭 microsandbox/microsandbox: Self-Hosted Plaform for Secure Exec…

2025-06-19T10:46:03Z

Here’s my thought on microsandbox/microsandbox: Self-Hosted Plaform for Secure Execution of Untrusted User/AI Code

There’s a ton of AI related tools I want to play with, and there are several I run in docker pretty willy nilly. However, the MCP server stuff as of late has felt fast and borderline reckless so I like seeing a project seeking to make it easy to sandbox code… I get blocked by this idea cause I’ll want to setup a VM on a VLAN or something before just spinning up an mcp server idk if I should trust - but if I can do it with a library then that’s amazing!

We just have to hope the motivation to try this stays strong enough to get started

Note

💭 Exploiting Copilot AI for SharePoint | Pen Test Partners

2025-06-13T02:17:17Z

Here’s my thought on Exploiting Copilot AI for SharePoint | Pen Test Partners

In hindsight this vulnerability isn’t even that clever, hopefully the share holders of the companies that rushed to throw AI into everything will make it out ok

Note

💭 SearXNG | Open WebUI

2025-05-25T11:21:40Z

Here’s my thought on SearXNG | Open WebUI

I’ve been meaning to set searxng for a while - it comes out of the box with khoj but I want an instance for open-webui as I get serious about deciding on a self-hosted AI tool-chain solution. It look like the setup for open-webui is a touch more than “docker compose up” so it’ll take probably 15-20 minutes of dedicated time to try this out

Note

💭 🏡 Home | Open WebUI

2025-01-01T13:26:50Z

Here’s my thought on 🏡 Home | Open WebUI

I’ve been using openwebui at home, mostly with llama3.2, to do some basic things - meal planning and short script generation. One thing I want to do is generate shrotened versions of my favorite books to read to my kids. For example, the Narnia series. It’s already a kid’s book, but a 3 year old can’t stay focused for even a whole chapter. I’d love for AI to shorten up the narrative to keep the jist of the story but in about 1000 words or so… I haven’t gotten it perfect, but openwebui lets me upload documents (for example, a pdf of The Lion, The Witch, and The Wardrobe, and then use that for context… It’s not quite RAG I don’t think, and that’s on the list of ideas to play with to this end.

Note

💭 Assistants API overview - OpenAI API

2024-12-29T20:26:01Z

Here’s my thought on Assistants API overview - OpenAI API

OpenAI assistants seem like a sweet idea… I wonder if I could use one as a project manager that manages my issues and prioritizes my day like a scrum master might

Note