I'm seeing amazing result to with agents, when provided an well formed knowledge base and directed through each piece of work like its a sprint. Review and iron out scope requirements, api surface/contract, have agents create multi phase implementation plans and technical specifications in a share dev directory and to make high quality changes logs, document future consideration and any bugs/issues found that can be deferred. Every phase is addressed with a human code review along with gemini who is great at catching drift from spec and bugs in less obvious places.

While I'm sure an enterprise code base could still be an issue and would require even more direction (and opus I wont let touch java, it codes like an enterprise java greybeard who loves to create an interface/factory for everything), I think that's still just a tooling issues.

I'm not of the super pro AI camp, but having followed its development and used it throughout. For the first time I am actual amazed and bothered, and convinced if people dont embrace these tools, they will be left behind. No they dont 10-100x a jr dev, but if someone has proper domain knowledge to direct the agent, performs dual research with it to iron things out with the human actually understanding the problem space, 2-5x seems quite reasonable currently if driven by a capable developer. But this just move the work to review and documentation maintenance/crafting. Which has its own fatigue and is less rewarding for a programmers mind who loves to solve challenges and gets dopamine from it .

But given how man people are adverse...I dont think anyone who embraces it is going to have job security issues and be replaced, but here are many capable engineers who might due to their own reservations. I'm amazed by how many intelligent and capable people try llms/agents like a political straw man, there is no reasoning with them. They say vibe coding sucks (it does for anything more than a small throw away that wont be maintained), yet their examples for agents/llm not working is it can't just take a prompt and produce the best code ever and automatically and manifest the knowledge needed to work on their codebase. You still need to put in effort and learn to actually perform the engineering with the tools, but if it doesnt take a paragraph with no AGENTS.md and turn it into a feature or bug fix they are not good to them. Yeah they will get distracted and fuck up, just like if you throw 9/10 developers in the same situation and told them to get to work with no knowledge of the code base or domain and have their pr in by noon.

IIRC there is another raw vulkan library that just generated bindings as well and stayed up to date but that comes with its own issues.

Our child also got stuck in the canal during birth and there was a good 30 seconds where the midwife from the hospital was trying to encourage to doctor who was to step in to let here keep trying, my kid came out white and took the longest 30-60 seconds to take their first breath. Never experienced so much dunning-kurger all at once. I had read a few week before that about medical professionals talking about how ominous a quiet birth it and was just zoned out as that was exactly what happened and I could sense all the tension. Then people from children services start demanding umbilical cord because my fiance had failed for MJ on her first prenatal vist, she quit smoking as soon as we knew and never failed a test after wards. But it all felt like an extreme lack of compassion. Then I was ostracised because I didnt want to cut the cord while I just thought my kid was dead and these social workers are trying to insert themselves in the process and its all chaos for no reason. The only good thing was a nurse pretty much told them to fuck off and wait in a nice but check yourself kinda way.

But multiple times people cared about their own ego, or their perceived power than actually attempt to do a compassionate job.

Thats what convinced me they are ready to do real work, are they going to replace claude code...not currently. But it is insane to me that such a small model can follow those explicit directions and consistently perform that workflow.

I've during that experimentation, even when not putting the sql explicit it was able to craft the queries on its own from just text description, and has no issue navigating the cli and file system doing basic day to day things.

I'm sure there are a lot of people doing "adult" things, but my interest is sparked because they finally at the level they can be a tool in a homelab, and no longer is llm usage limits subsidized like they used to be. Not to mention I am really disillusioned with big tech having my data or exposing a tool making API calls to them that then can make actions on my system.

I'll still keep using claude code day to day coding. But for small system based tasks I plan on moving to local llms. Their capabilities have inspired me to write my own agentic framework to see what work flows can be put together for just management and automation of day to day task. Ideally it would be nice to just chat with an llm and tell it to add an appointment or call at x time or make sure I do it that day and it can read my schedule and remind-me at a chill time of my day to make the call, and then check up that I followed through. I also plan on seeing if I can also set it up to remind me and help to practice mindfulness and just general stress management I should do. While sure a simple reminder might work, but as someone with adhd who easily forgets reminders as soon as they pop up if I can get to them now, being pestered by an agent that wakes up and engages with me seems like it might be an interesting workflow.

And the hacker aspect, now that they are capable I really want to mess around with persistent knowledge in databases and making them intercommunicate and work together. Might even give them access to rewrite themselves and access the application during run time with a lisp. But to me local llms have gotten to the point they are fun and not annoying. I can run a model that is better than chatgpt 3.5 for the most part, its knowledge is more distilled and narrower, but for what they do understand their correctness is much better.

I have encounter a lot of your posts and that's what pushed me towards just tackling vulkan instead of using wgpu. I also encountered many of the same issues around the ecosystem. I think the main issue is there is just not enough dev time going into it or money. Even valoren, which I already knew of before learning rust from posts in linux/oss communities only has received 8k of funding, while offering the closest to an AA experience.

But I don't think its that reasonable to expect the ecosystem to just have a batteries included performant general rendering solution, idk if any language has that? I know there is bgfx, which might be the closest thing but I assume also has its own issues. So I don't really think its the graphics part holding things back, as ash is a great wrapper around vulkan and maps 1-1 with a little bit of improvements (builders for structs, not needing to set stype for each struct, easy p chaining).

The main issue I encounter is all around the lack of dev-time and the tendency for single developers and for small single purpose crates. Most of my friction is around lack of documentation, constant refactoring making that lack even worse, and this causing disjoint dependency trees. So many times have I encountered one create using version x.x of one crate that depends on x.y version of another then the next being on z.x of another dependency and then another still needing z.y. This normally wouldn't be that big of an issue, except the tendency to constantly introduce refactoring and breaking changes meaning I end up having to fork and fix these inter-dependencies myself and cant just patch them.

But this all just circle back to there just isn't much dev time going into them. It also seems the "safety" concerns and rust just not allowing some things causes devs of many crates chasing their tails with refactors trying to work around these constraints. But it does get quite tiresome having to deal with all of these issues. If I was using c++ I could just use sdl/glfw, imgui, vma and vulkan and they would all be up to date with each other. In rust I need winit, imgui bindings, imgui-winit, imgui-vulkan, raw-window-handle, ash and vma bindings. And most of these are all using different versions of each other and half of them have breaking changes version to version.

I'm pickign rust up by porting over a bytecode vm, so I kinda need to use some raw pointers. It would gaslight me about the risks and how it would be irresponsible to help me as it could lead to possible violations of the integrity of user data.

I had to explain to the AI that it is a personal project that has no users data, the only risk was the program crashing and it was a personal project that would only affect me. It still would try to revert or tell me other solutions, I finally just went and read up on it elsewhere.

What made me appreciate scheme was watching some of the SICP lectures (https://www.youtube.com/watch?v=2Op3QLzMgSY&list=PL8FE88AA54...) and the little schemer to learn more. I also read some of the SICP along with it, though I put it down due to not having the time to work through it.

Scheme is interesting and toying with recursion is fun, but the path a mentioned above is only really enjoyable if you are looking to toy around with CS concepts and recursion. You can do a lot more in modern scheme as well, and you can build anything out of CL. But learning the basics of scheme/lisp is can be pretty dry if you are just looking to build something right away like you already can in a traditional imperative language. But it is interesting if you are interested in a different perspective. But even RS7S scheme is still far from the batteries included you get with CL.

I personal found the most enjoyment using Kawa scheme, which is jvm based and using it for scripting with java programs as it has great interop. I used it some for a game back end in the event system to be able to emit events while developing and script behaviors, I've also used it for configurations as well with a graphical terminal app, I used hooks into the ascii display/table libraries then kawa to configure the tables/outputs and how to format the data.

I've moved to combining all my data into single files, but sometimes it also seems to have issues with them as well even if they are under the upload size limit, I assume that is due to how many characters are in them, and it will just brick the whole GPT until the offending file is removed.

The part I have issues with is having it actually use the data, it will quote/summarize data it found in the knowledge base and return where it found it if it can, but I can never make it do more than that. Ideally I want it to contextualize the data it finds in the knowledge files and prompt itself or factor it into a response, but anytime it accesses the knowledge base I get nothing more than a paraphrased response of what it found and why it may be applicable to my prompt.

I have tried to work on one where I uploaded various documentation and spec sheets, wrote detailed instructions on how to search through it. Then described how it should handle different prompt situations (errors, types of questions, quotes from the documentation). It is able to search through the provided knowledge and provide quotes and responses with it, but it at no point gives a coherent response, so it basically always functions like a more intelligent search feature. Putting that it should re-prompt itself with the knowledge extracted and rationalize/elaborated on it doesn't seem to do much either, though it did provide some improvement.

Being able to inter-opt with java is great, as I can wrap scheme procedures in functional interfaces and use them as drop in replacements for java functions, as my event system was already based on predicates and consumers for event handling.

I've worked through some of the SICP and have always wanted to get more into scheme/lisp, but the barrier of starting a full project in it always kept me from getting much hands on experience. Its been quite enlightening actually getting to work with a form of REPL driven development and getting my hands dirty with coding some scheme, having access to the JVM means I can do practically anything with it, with out needing to bootstrap tons of code, and using it in a project with a large scope lets me solve real-world problems with it vs just toying around which has been what most of my scheme/lisp experience was before.

Opinions on property rights aside there is no lack of land to explore and enjoy.

Aside from there there are also state parks and forests, though the states define their own terms of use and enjoyment around them.

Bruce Eckel is fairly well known for his original Thinking in java/c++ books, and I find his style of writing great. He has a knack explaining topics in an easily digestible thorough way, while not being completely dry. Its a 1200 page book, and can double as a reference, but actually explains the when, how and why in thoughtful writing enough to double as a great book for learning and a reference when needed. I knew a lot about java and learned a lot of little things about topics I already knew, and it breaks down each piece of the language so you can easily gloss over things you already may know from other from other languages.

I would say with someone new to coding it can be bad and good, as a lot of times it glosses over things, or can be slightly incorrect as it makes assumptions (or more so just answers in a more general context, and when asked to elaborate, or challenged on specifics it will reformat/improve it's answer, but without knowing you need to do so, I could see it easily see it providing half-baked foundational knowledge. You can ask it "x" and it will give a answer, but then if you ask it I am trying to do "y" with "x" and isn't "z" an issue or area of concern with its answer it will reformulate the information provided as its original response was flawed, but if you don't know exactly what the "y" you want to do is, or the "z" being foundational knowledge to challenge it on, you can easily get a whole wall of text that is out of context with what you are actually trying to learn.