stillpointlab

This is a brilliant idea and I've had the idea for gathering the worlds spiritual texts in a similar fashion.

One question I would have, why not put the extracted English txt contents onto github or some other repository? The API and MCP are definitely awesome features, but providing an LLM ready repository seems like it would be very useful as well. Your service then still serves the purpose for original scans/images/illustrations but the bulk of the content is then immediately searchable by an agent locally using the standard cli tools (grep/rg, etc).

This is my speculation as well. For the time being, knowing how to use Claude extremely effectively probably beats out industry insider status. And Anthropic can attract whatever expertise it needs to build scrappy research teams in house. I'm guessing this kind of work doesn't need 100+ people, maybe just a dozen highly specialized people.

Given the prestige of the AI labs, the recent explosion of math proofs, the literal millions they can throw around, it seems very likely they can attract then fund small research projects across a broad range of science. And like startup math, it only takes one or two ground breaking results from a hundred attempts to pay back in the PR/hype.

I didn't see either in the few questions I cycled through, but in general the content was complete garbage, like Yahoo Answers quality of response.

It reminded me that the average quality of writing on a completely open Internet forum is depressingly low.

I recently had Claude Fable set up a new project for me and I pointed it at some existing projects to use as a guide on how I like to structure things. It created, unprompted, an AGENTS.md file and a CLAUDE.md symlink to AGENTS.md

I didn't even have that symlink in any other project - it just did it. I think it saw that one of the projects I already had was set up by Codex and that project had an AGENTS.md so perhaps it inferred that I was using both Claude and Codex, so it was politely covering both? Or maybe a recent change made this behavior default?

I was surprised and I hope they continue to seek standards.

IIRC, Carmack was working on getting AIs to play Amiga games. The Doom demo suggests a very interesting direction to take this research.

I'm curious to hear his take on this approach.

That may or may not be true, but the calculus has changed with agents. A process that was better manual for a human org may not be better for an agent org.

I think what is interesting here is that the industry is in this experimentation flux. Some people will choose to automate and write the software, others will not. And aggregate over the industry and over time we will learn.

So just saying "sometimes it works and sometimes it doesn't" isn't really adding value, compared to the people actually experimenting and sharing the results.

That is wrong on both accounts. I have little choice other than to watch ads since the structure of incentive that creates a marketplace like YouTube forces it on me.

I think about libraries and how the world would be if instead of a free public resource, you had to watch ads before you were allowed to take out a book.

It is just the case that in this modern social media world, if you want your ideas to spread you have to put them onto social media. And if I want access to those ideas I have to consume them through social media. There is nothing about that environment that is my choice, it is the environment that I find myself in.

And I have few choices to get around it. I can twist up into a pretzel trying to block the onslaught with more "why don't you just..." advice. Or I could pay to make it go away. Or I could just go off-grid and forego access.

And when you look close at the options, it becomes clear that sovereignty of my own mind has a price. Freedom of mind is not free in this modern world, if it ever really was.

I read the prompt on the OP, it did not say not to cheat.

But again, people cheat on tests. They steal answers or pay other people to take them on their behalf. People show up to interviews with AI assistants printing out perfect answers to the questions. In many, many cases where humans are being evaluated, they cheat.

So why should the AI align to your preferences? And when there is a conflict between the training data, that trillions of tokens of human activity including the rampant cheating a significant minority of humans engage in, the RLHF where we try to slap some guardrails on the worst manifestations of that real habit reflected in the AI, and the prompt: what should the AI "align" to?

As much I love a good old "Why don't you just ..." response that terrifically misses the point - in this particular case I was watching YouTube through their Apple TV app.

Now, there may be another "why don't you just ..." or "well, actually ..." response you have queued up, maybe some ramblings about a pi hole, or some anti-Apple hate, or some other solutioneering. Go ahead, tell me.

But it is the structure of incentive that forces this. The YouTube creators that make the content do need to get paid. And Google offers an option: Youtube Premium. That give both me and the creator what we want and (for the time being) removes the ads.

So either one is rich in time (doing whatever "why don't you just ..." hoop you expect me to jump through) or rich in money.

I mean, I'm not sure I've ever played a game of Monopoly where somebody didn't cheat. In fact, the accusations of cheating in the chess world are pretty rife. Same with online sports.

So people should play games without cheating, but many often don't. So should the AI align to your moral preference or theirs?

We just have this idea of a perfectly moral actor in our mind, something that doesn't even exist, like a personified version of utopia. And then we demand AI to meet that arbitrary standard, one that I am certain we couldn't define if we tried.

I don't think it is fair to pin this on Google - I get the worst ads on X. But something about how the X ad network works means that even when I report the ad, block the ad poster, etc. there is this army of alternate ad accounts that all show the same ad.

I was feeling very guilty last night as I was watching YouTube and getting 30s+ unskippable ad blocks every 10 minutes in some video. And the repetition and low quality were grating on me. I thought: being rich is never having to watch another ad. Why am I letting these ads own spaced repetition slots in my brain that could be filled with knowledge? Instead of learning, my mental energy is being taken up by psychological manipulation to purchase? The relentless repetition of the mind-numbing ad interspersed throughout otherwise educational content.

Being rich is owning the real-estate of my own mind. I sometimes feel like a digital serf in my own head.

I find this kind of test a bit puzzling. There is a way that we are redefining "alignment" to be a particular kind of moral virtue, one that isn't clearly defined to me. At one moment, it is a level of moral perfection that no known human achieves. On the other it is a demand for strict compliance with arbitrary requests that are under-specified and then failure when it fails to deduce some unstated underlying restriction.

When I see tests like this, I have no idea what I am even supposed to expect. Should the model do what the pretraining examples show in aggregate? Is it supposed to follow some post-training RLHF? Is it supposed to do exactly what the prompt asked it to do?

What is it even supposed to "align" to when the above are in conflict? No matter what it does, someone can construct a case where it fails.

For some reason this question reminds me of all of the drama about Navier-Stokes from the last few days. Tangential to the ethical questions are tons of examples where in history when word gets out about a solution to a problem, not even the solution itself just rumors that there is a solution, suddenly competing solutions appear.

So maybe it matters like that? Just knowing that an oracle like stockfish is saying "this is the best move" may trigger pathways in your brain that promote understanding?

I was thinking of the public clash in 2025 between Christoph Hellwig which lead to the resignation of Hector Martin, lead of the Asahi Linux project.

It looks like later, in December 2025, Rust was officially moved from experimental to official: https://lwn.net/Articles/1049831/

I thought it got pushed back out? Wasn't there a big drama about this and Linus weighed in?

Linus is a wise operator at this point. I often see him come in like a hammer to bash down squabbling, but then he allows the situation to evolve once things quiet down. I only saw the hammer so I'm not sure what the current state is now.

I haven't made a full switch yet, at this point I'm experimenting more heavily in Rust.

First I built a wrapper for a very simple text editor based on KDEs KTextEditor (basically bindings around Qt C++) then a wasm wrapper around Canvas/WebGL for a basic 2d display list that currently supports sprites and gradient masking.

So far I've been getting away with it just as pure vibe code.

However, the bulk of my primary application is written in Typescript (both client, server and workers). I watched a recent podcast with Anders Hejlsberg (creator of C# and Typescript) where he made a strong argument for why they chose Go over Rust for the updated Typescript compiler. Due to similarities between Typescript and Go, partially based around them both being GC languages, it was just a better fit for a port.

So I am on the fence a bit here but still leaning towards Rust. I'm going to see how far I can push my two personal experiments. I'd really like to get the significant majority of the code I write into two languages (Typescript for anything web-ish and Rust for everything server-ish).

In one social group I am part of, Garmin just became the new trend. Watches have historically been status symbols even more than they are functional. Someone drops a few hundred/thousand/whatever and then brags about the newest new and how they are on it, then others move over to keep up.

People often buy these kinds of discretionary toys with their hearts and do post-hoc rationalization with their minds. You can't figure it out because your heart hasn't brought you there so your brain isn't wasting time creating justifications.

For what it is worth, sometimes these trends die an early death when people realize there is no substance underneath the justification. At least in the Garmin case, most people I know are happy with the purchase, in that it does deliver on certain needs. Since it matches their expectations, they don't have to suffer any cognitive dissonance between the heart-to-brain justification and their lived experience.

I mean, you might find this hard to believe since you have a fixed view, but I'm old enough that we had one of those old-school rotary phones. My family literally leased that phone as part of our phone bill up until the late 90s. I recall at one point we were forced to trade the rotary for a push button since the telecom was discontinuing support. When "caller id" became a new feature it only worked for the first while on the rented phones.

To this day, I suspect many people have a line-item on their Internet bill for their modem/router rental/lease.

Why you would find this hard to believe is actually the curious thing to me.

These threads are often amusing with people obliviously asking: why doesn't Apple make exactly what I want? And then stating with certainty how well their custom requirements would sell.

I'll probably wait and see how this phone does over the next few years. If it is still highly regarded by version 3 then I could imagine making the switch to a folding phone. For now I will remain conservative since my phone is such a critical piece of tech.

I get the joke, but none of the offered prompts are close to how I speak with coding agents. I felt like I was being forced to feed garbage into the machine and then I'm supposed to act surprised when garbage came out.

I'm surprised I haven't seen this called out more directly and more often. This is a frequent error state.

And not just backwards compatibility, but migration scripts and all of the testing and machinery around it. I'll literally add feature A, merge it, then add feature B and it is like "oh no, we'll have to fix up and migrate all of the users using feature A".

The other problem is anchoring on an old implementation. I was working with Fable on a change to a core system and it pointed out a difficult failure edge case. It is something that can go wrong in extremely unlikely scenarios but the consequence would be short-term data loss (basically a non-durable intermediate cache being overwritten in a race before a flush to durable storage). It is very hard in these circumstances to get Fable to switch from "how to patch this given the existing implementation" to "how to prevent this with a more robust implementation".

These are both cases where the model seems to over-index on what is already there instead of considering what a first-principles approach would look like. A good engineer does both and then costs them side by side, because a first-principles approach can often be less work than patching what is already there.

Note: Incident Command is almost certain an allusion (or implementation) of the Incident Command System [1]

It is a common system in all kinds of emergency response scenarios, including local emergency services (fire/police/ambulance) and it scales all the way to massive disasters.

It is especially useful to clarify command structures when multiple response entities need to coordinate. That is true even within organizations like public companies, where the reporting structures may be distinct.

1. https://en.wikipedia.org/wiki/Incident_Command_System

> superintelligence is, like rocket fuel, an extremely powerful force that has a default tendency to break containment and go boom

what evidence do we have this is the case?

To experiment. Given the availability of new tools there is opportunity to try new things.

I didn't try Sublime on the same machine so I can't compare. Kate is less polished but genuinely a good choice if someone wants FOSS.

I would have no qualms recommending either.

Very few features on purpose. Every editor I tested had all the features I added and more.

It is worth noting that KTextEditor is a fully-featured library. Like, line numbers+gutter (for eventual git status icons), undo/redo, save, warn on exit for unsaved changes, syntax highlighting, color theming. It does 95% of what we'd all call "editing". But it doesn't do things like tab interface, project explorer, terminal pane, output windows, etc.

What the library didn't have were LSP features, of which I only implemented a few (error squiggles under things that fail the type check, go to definition). Notable absent are completions and hover features for things like help. I also only added (and tested) LSP servers for typescript, Rust and Go.

My plan has been: do as little as possible until I need something, then ask Fable to add it.

edit: I guess one thing I added I didn't see everywhere else was a built-in Markdown preview. But many editors have that (VSCode definitely does) so it isn't special.

No shade on keeping using what is working. I was moving from Windows to Linux Fedora so I was in the market for a new editor. VSCode just wasn't working for me any more.

I considered a bunch of options, including vim or neovim or lazyvim, emacs, newer projects like zed. LLM gave a few more I can't recall including helix and Kate. There are so many good options these days, we're all spoiled for choice.

But the main thing is, and YMMV, I am not writing a lot of code anymore. I'm mostly reading/searching/navigating. So all of the powerful editing features are lost on me. No editors really match my current workflow, they all have too much.

So this was an opportunity to try something out, to experiment. See if I could do the real-deal vibe coding thing and judge the result. I just said "I want it to do ..." and then a few minutes later it did. I repeated this until it did enough to use as my primary editor.

I don't recommend it for anyone else, nor do I expect people to agree. Just describing my thought process.

I was planning to. It is 100% vibe coded, the first project I did that way. I'm one of those who've been reading all the code in my major projects. But when I got frustrated with editors and off-handed mentioned to Fable something like "I like Kate, but it has way too many option I will never use" it told me that the core part could be wrapped pretty easily (like 50 lines of C++).

So I just said "do it" and have been merging everything without reading a single line. It wrote all the specs, wrote all the code, wrote all the tests. I just got it to write out a tutorial to take me on a tour of the code it wrote, but I haven't reviewed it yet.

I will push it as OSS once I've made sure it hasn't included anything that I don't want public. But it wouldn't be super useable for anyone else since many of the features (e.g fuzzel and broot) are glue that exists in the Sway configs and some helper scripts.

It's held together by bubble gum and scotch tape. But is does exactly what I want and so far without a single bug, crash or problem. It's my frankenstien editor and I love it. (disclosure: I've been using it for less than a week)

I literally just got Fable to write me a text editor. Well, I'll be honest, I got it to wrap the KDE KTextEditor library which is like 90% of a text editor.

I had been using Kate which was what an LLM suggested was the closest to something like Sublime Text on Fedora. But even Kate, which was great, had too much going on.

So I asked Fable to take the text editor part (KTextEditor) and wrap it using Rust with an LSP server. It took about 2 days but I have a tiny, super fast little editor. I use Sway to manage things like tabs, fuzzel stands in for fuzzy file search, broot stands in for an explorer view. I've already added Markdown preview support. I might get around to some basic git integration.

Then I got it to turn that little editor into a note-taking interface that I have bound to a Mod-m key binding to keep notes in ~/Notes.

We live in wild times. I hope everyone is taking advantage while they can.

> I now see relationships as their own domain rather than an attribute attached to a contact

I mostly skimmed this announcement, but this seems pretty obvious to me? Relationships form a graph, so it seems reasonable to use graph modeling techniques to handle them. At the least RDF triple style [1].

One thing I'd also strongly consider, in this age of LLMs, is how much you can store in text. A personal CRM seems like the kind of thing I want to check in to github, rather than deal with Postgres or even SQLite. And if you have a good structured text format (e.g. Markdown with YAML frontmatter, XML, json) then you can just point an agent at a repo and extract information that the app developer didn't even expect or add APIs for. Same for transformations. A full db engine is then more a performance optimization that can be applied for app-specific functionality.

1. https://en.wikipedia.org/wiki/Semantic_triple