kator

I learned BASIC in 1981 on the TRS-80 Model I with 4k RAM (later upgraded to a massive 16k), and it wasn't long before someone showed me EDTASM [1], and I was hooked. My parents couldn't believe me when I asked them to buy "How to Program the Z80" by Rodnay Zaks [2]. They were shocked, it was the first time I asked them to buy a technical book. I was so excited I couldn't sleep for days. I was in High School, and I would carry the book around and read, use the lookup tables, and write Z80 assembly in my spiral-bound 8.5"x11" notebook.

One three-day weekend, I slept about 4 hours while I was disassembling the BASIC ROM to find nuggets I could call in my Z80 programs, and also to learn so much more about the video, the cassette interface, interrupts, etc. My notebook looked like something a madman had scribbled on the walls of an asylum to everyone around me, but to me it was perfectly organized, and I used it to reference things I would use later in my programs. I got to the point where I could read the hex dumps and "see" the op codes in my mind; people would look over my shoulder while I was debugging, and I'd be explaining what the code was doing, but I hadn't disassembled it yet. I didn't even realize I was doing that until one of my classmates pointed it out: "Hey, disassemble that. I'm not following you, how the heck do you see that?"

And all of this because my science teacher had that TRS-80 Model I sitting in the corner of his classroom gathering dust; nobody knew what it was, I didn't either, but I had to figure it out. I've always had this bug, I don't believe in magic, I want to know how everything works, it's been an insatiable thing all my life, that first machine set me on a course of learning for the rest of my life.

Thanks, Mr. Kruzan, you changed my life. I'll never forget you. Oh, and thanks to all the people who created the Z80, the TRS-80 Model I, and paved the path for me to learn computers and never stop learning.

We are truly building on the shoulders of giants; every generation, more giants. It's humbling to look back and think about it forty-five years later...

[1] https://www.trs-80.org/edtasm.html

[2] http://www.z80.info/zip/zaks_book.pdf

PS: Pretty sure the memcpy examples have a stack bug.

That was 2001 today Joel seems to think this is the future: https://hash.ai/

https://archive.is/zDhfv

Years of practice and “This won’t kill me” thinking have helped a lot. Also I watch every single talk I’ve given which is brutal but it lets me improve but also re-affirm I didn’t die…

[1] https://www.karlbunch.com/random/change/

Professional public speakers...

Maybe some do, but I've never needed it. Often I actually find public speaking easier than small groups. In a small group my brain is trying to "model" what each person is thinking about my talk, as the groups get larger that becomes impossible and I tend to relax and let go. I also find the energy in a larger setting is a useful feedback mechanism. I might toss a small joke out and see if the audience is engaged, or I will ask a question and get a show of hands. The more I engage the calmer I feel and the more enjoyable the experience is for me and my audience.

True, that said, I downloaded and compiled Perl in 1987 from comp.sources.misc, even back then, things moved at light speed compared to health and medical.

I think the most interesting thing is that it took 15 years for people to apparently take this seriously. And another 40 to recognize its impact. The original paper[1] was from 1982...

Having been in software development for 45 years, I find this crazy. Maybe it's because in our world, it often takes a month for something to spread from "interesting" to the new technology of the day, or the new way of doing things.

[1] https://www.jneurosci.org/content/jneuro/2/3/284.full.pdf

Reading your link IMHO in today's world I would set a basic rule, if you're touching >20% of a Java codebase you should refactor to Rust. With AI-Native development practices it's worth the SDE time to refactor, replace the underlying subsytem and reduce your fleet by 50% or more.

Might keep this in mind if you go to tindie:

URGENT

Tindie has not paid me for the last month. They take your money for your order and then require me to ship product to you, but they do not pay me. Temporarily, please place your order through the tattlersolutions.com website. See below

Next up, you must login to ID.me to use AI and providers will be required to retain every session for analysis.

Most security code scanning I am aware of does AST parsing of actual code before analysis; the comments won't even make it to the LLM. That said, embedded strings could cause this type of false denial, but even so, the errors would be raised in the pipeline for human-in-the-loop security analysis. If anything, it might get a faster reaction in some environments because it causes faults in the analysis pipeline.

Oh, and while you're at it, 640K ought to be enough for anybody.

oh good point got that backwards… OMG my fax brain didn’t even think about it.

The sad thing is I feel trapped on all sides of the debate, I wrote a book about LLMs and human creativity (spoiler Humans win for a long time) but I was going to do it as a blog series, instead I published https://www.amazon.com/dp/B0GXCSY4W8 because I felt at least I might get a bit back for literally 100’s of hours of my life I poured into the book and my editor and friends who read and provided reviews.

And I push a lot of open source code including a ton for the SWGEmu project, but now I’m of mixed mind to stop pushing anything public. I can’t decide, am I talking out of both sides of my mouth, it’s a confusing time to navigate for sure.

Yet, this shift made me re-evaluate the open source code publishing. Prior to that, I have been positive about free and open software, and considered this to be the default mode for work such as kefir. I did not require any justifications from myself to publish something. Now, however, I feel more and more that the main beneficiaries of my unpaid work are companies scraping the internet to train large language models. Currently accepted status quo in this area goes against my own intentions in licensing this work under GNU GPLv3. Publication has ceased to be the "null hypothesis" for me, and requires explicit mental justification which I am not able to provide.

I feel this pain, one of my small donation driven sites has been destroyed by crawlers who just ignore robots.txt and burn the site into the ground.

Sort of jokingly I proposed an update to the "spam fax" law:

https://www.karlbunch.com/random/website-protection-act/

All this said I'm more concerned about Automatic Content Recognition (ACR) on smartTV you buy in the store and never even realize it's phoning home with everything you watch...

I recently had my donation-driven site ruined by bots, it's a constant battle. I (jokingly) proposed we should amend the fax spam law to take this into consideration:

https://www.karlbunch.com/random/website-protection-act/

555 gigabytes of bandwidth in a week! We're paying more for egress than compute and storage now. I've tried robots.txt and finally gave in and started setting up aggressive WAF rules.

I wonder whether the author deliberately avoided ontology? That's what comes to mind when I read this. The age-old debate between taxonomy and ontology.

I don't think the mirror is a warning. It could be the most valuable thing LLMs can do for us. I recently dug very deep into my own use of LLMs and why I was getting different results than a lot of my peers. I think the primary difference is I look at them as a tool that can extract what's below my verbal mind, deeper in my experience and intuition. Everyone points LLMs outward, "go build this", "go write that", but they miss the most valuable thing, "help me get this idea out of my head".

I recently wrote a book about this, turns out the mirror is more interesting when you stop looking past it and look into it. And even more when you realize LLMs are not humans, they have strengths, and weaknesses, but so do the humans using them. If used carefully you can achieve much more together than either part can alone.

models that aren't over-optimized for it.

But how do you know the model was over-optimized for it or just really good?

Those who fail to study history (or live through it) are doomed to repeat it.

SPECint and SPECfp went through this exact movie: benchmark, saturate, retire, replace, repeat. The treadmill is the product.

I don't have the solution just noticing the pattern.

I tried tmux so many times, could not commit the sequences to memory, but then zellij was just out of the gate, easy to "discover," and then I started writing plugins (rust wasm), and I even submitted a PR which got accepted to support background colors in panes/tabs.

I am a monthly donor, I think it has the right balance of community plus the lead dev has a vision, opinionated but open to inputs, and focused.

Interesting, I just did this on some stuff I have personally handwritten, but the writing is about LLMs, and it says 25% AI generated, but because I used terms that are "x% likely AI" generated.

This is not a very useful test; it basically means a person has to "ban" terms like "user experience" and "tireless" etc. because these are "Nx more likely to appear in generated content".

I think it's a bit sad that we often say people "Sold out". Sometimes, I agree, but often, I point out that until the lady at the grocery store stops asking me for money when I walk up to check out, I need to pay my bills to eat.

I contribute to open source projects, but none of them to date could support me buying much more than a beer. If one took off such that I could "live" off it, I would be happy to leave my current job and dive all in. Until then, I just keep plodding along.

I put together a quick audit to check for "early landing" messages[1] using jq, ripgrep, and the messages[2] flagged in the stop guard script.

I have noticed a trend in these sessions asking more and more about calling it a day, "it's getting late," and other phrases. I sort of assumed it was some kind of "load shedding" on Anthropic's side.

My audit of 80 sessions was interesting. Sorry, I won't share details, but I recommend you do the same.

[1] https://gist.github.com/karlbunch/d52b538e6838f232d0a7977e7f...

[2] https://gist.github.com/benvanik/ee00bd1b6c9154d6545c63e06a3...

Fascinating, I thought I was losing my mind. Claude CLI has been telling me I should go to bed, or that it's late, let's call it here, etc, and then I look at the stop-phrase-guard.sh [1] and I'm seeing quite a few of these. I thought it was because I accidentally allowed Claude to know my deadline, and it started spitting out all sorts of things like "we only have N days left, let's put this aside for now," etc.

Just this morning I typed:

    STOP WORRYING ABOUT THE DEADLINE THAT IS MY JOB

The "calculator ruined the world" argument was actually studied to death once the panic subsided. Large meta-analyses of 50 years of data show it was mostly a non-problem. Students using calculators generally developed better attitudes toward math and attempted more complex problems because the mechanical drudgery was gone.

The only real "catch" researchers found was timing. If you introduce them before a kid has "automaticity" (around 4th grade), they never develop a baseline number sense, which makes high-level math harder later on.

It's a pretty clean parallel for LLMs. The tool isn't the problem, but using it to bypass the "becoming" phase of a skill usually backfires. If you use an LLM before you know how to structure an argument or a block of code, you're just building on sand.

Peak irony: OpenAI's Codex project is being DDoS'd by its own creation.

They made code generation essentially free, and now they're drowning in AI-generated PRs that take more time to review than it would take to just write the fix themselves. So they've moved to invitation-only contributions. They built the tools that removed code as a barrier, and now they're living with what that actually means: quantity goes up, quality goes down, and every PR still needs a human with architectural context to evaluate it.

The bottleneck has completely flipped. It used to be finding someone to write the code. Now anyone can dump a 200-line PR in seconds, but the review time from someone who actually understands the codebase is the most expensive thing in the repo. That context can't be generated, it has to be earned.

They built a firehose and their own team can't drink from it.

Some users are moving to local models, I think, because they want to avoid the agent's cost, or they think it'll be more secure (not). The mac mini has unified memory and can dynamically allocate memory to the GPU by stealing from the general RAM pool so you can run large local LLMs without buying a massive (and expensive) GPU.

I use Terminus with Zellij and keep about 8 sessions going with a combination of Claude and Codex, and once in a while, Gemini. It's great when you're sitting in a docotor's office lobby bored out of your skull and when you get back to your desk you just join the session and it's all right there.