0xb0565e486

I keep seeing Astra make beautiful 3d stuff online, yet when I feed it some old school RuneScape assets (even tried with some very detailed guidelines) and asked it to generate some new plausible assets it failed horribly.

I think there’s still something really off with current (frontier) models when it comes to creating “novel” stuff? Even 2004 style graphics..

Or am promoting it wrong?

I have! Testing it on a few projects that I ended stuck on a specific set of problems.

So far it seems better at understand what my intent/the scope of the change I want to make?

Hopefully not just a fluke! I really like working bit by bit through iterations rather than doing a mega plan and executing. I tried muse 1.3 today and it was the worst offender of this. Just spamming extra validation/inferring behaviour I didn't (not yet) want.

When I was a kid my mom had a daycare and had access to a program where she could buy older and used desktops at a discount.

I loved installing different operating systems on them and try to customize it. Better window management, better animations, nicer colours etc.

Of course, the results were always marginally better or worse than the stock version of that distribution.

Seems as this is the case for Omarchy here as well.

Regardless of what instructions I put it to try to be terser and all, I still find that the output is of much lower quality.

Even if it uses little words, I often find the reasoning that is quite challenging to understand rather than it was if it was a regular human or colleague that would just explain me something.

Any tips here?

I really wanted to Lisp as a main programming language, and sometimes I still do.

I just find readability such a hurdle regardless of how long I used it. I didn't find that it ever became as natural as the other group of programming languages.

I find a procedural style of programming so much easier to reason about, both when writing and reading.

Either way, I'm really happy I took some time to learn it and use it a little at some point.

Aren't you paid to think?

No idea, seems like a much better fit :shrug:

I love how beautiful and creative the website header is while being a fragment of what it does. Fantastic work!

Lovely 3D header animation! Fits incredibly well with the posts content :)

I think you may be over-reading into this a bit. Seems to me as just a familiarity thing?

Programming in English = X Programming in French = Y

You're used to X and now it's Y. So it feels weird. English speakers are used to X and it's X. So it feels normal.

Somewhat related since this about "workflows" and not cloud function, but are there any practical benefits to cloud functions other than the fact that it's cheaper for the providers as they don't have to run an entire bespoke runtime for every customer/application?

I have spending the last ~4 weeks writing a triple store!

I wish this came out earlier, there are a few insights in there that took me me a while to understand :)

Isn’t this more of a modal usage thing than the actual data model?

Isn’t the slack data model presented here totally possible with hipchats actual data model?

I did not know you could patent data structures like that.

I think the website looks lovely! The style gives it a lot of personality.

I'd love a truly new OS, but I just don’t know what it would look like at this point? "New OS" ideas tend to converge on the same trunk.

Building a hobby OS taught me how little is just "software". The CPU sets the rules. Page tables exist because the MMU says so. Syscalls are privilege flips. Task switches are register loads and TLB churn. Drivers are interrupt choreography. The OS to me is just policy wrapped around fixed machinery.

I’m using QEMU virt machine, so no hardware for the time being.

Would love to boot on a physical machine eventually though! If you have suggestions, happy to hear them :)

It’s getting there, but it does not handle permissions so you either have to add a bunch of rules through the database (such as RLS on Postgres) or define a permission schema.

Trying to see how far inference can go given that queries usually specify this information (ex: where(r => r.author == $SESSION.AUTHOR_ID)).

Lately, I’ve been exploring a few interconnected ideas:

Local-first web applications with a compiled backend – After eight years working on web platforms, the conventional stack feels bloated. The client already defines what it wants to fetch or insert. Usually through queries. So why not parse those queries and generate the backend automatically (or at least, the parts that can be)?

Triple stores as a core abstraction – I’ve been thinking about using a triple-based model instead of traditional in-memory data structures, especially in local-first apps. Facts could power both state and logic, and make syncing a lot simpler.

Lower-level systems programming – I’ve mostly worked in high-level languages, but lately I’ve been writing C libraries (like hash maps) and built a minimal 32-bit bare-metal RISC-V OS.

It’s all still brewing, but I think these ideas tie together nicely. What if the OS didn’t have a file system and just a fact store? Everything could be queried and updated live, like a Lisp machine but built on facts.

Some other things I’ve been playing with:

A jQuery-like framework and element factory - You can pass signals that automatically updates the DOM.

A Datomic-like database on top of OPFS - where queries become signals that react to new triples as they enter the system. Pairs well with the framework above.

That’s some crazy high numbers for something that has yet to be shown/proven to be effective for its most common use case.

Very cold and insensitive management style in my opinion.

If a project I was working on for multiple years and I believed in was shut down and my manager told me “welcome to earth” and “our lives don’t matter” I would one hundred percent leave.

Yes, the markets are cold and ruthless but it doesn’t mean you have to be.

I would argue that in most cases where performance isn’t a constraint, the first algorithm that comes to mind is probably the most optimal choice. He even says:

> About 80% of the candidates go for the naive solution first. It’s easiest and most natural.

The “naive solution” will be easier to understand and maintain. Why make it harder if it doesn’t add value?

Why would you have a child when you could feed an updated dataset for 9.99$ to re-train your large language model from infancy to adulthood?

The Bible

I much prefer to have my remains buried as god has intended.

What do you mean? Where are you from? It’s totally possible for “the youth” to own a house from my perspective.

Yes I did.

Cant comment for the “tech circle” you’re referring to, but from where I’m from “the environment” was a topic taught in school from the get go and taking a stance that we aren’t “destroying the planet” was pretty much out of the question and frowned upon.

I specifically remember one of my teacher saying they wouldn’t have children because the future was going to be unbearable for them.

I’m bullish on life.

Why do pessimists think they know it all?