I think I will pick up work for Arris (Java-no-build) on my vacation.
More jdk tooling support and some nicer ergonomics.
That will sadly bloat the lines of code, but I think better ergonomics are worth it.
(Still single file tho)
I think I will pick up work for Arris (Java-no-build) on my vacation.
More jdk tooling support and some nicer ergonomics.
That will sadly bloat the lines of code, but I think better ergonomics are worth it.
(Still single file tho)
It is time to globally outlaw centralized for profit compute.
It is time to globally outlaw centralized for profit compute.
Hey Game devs!
I'm looking for an Audio programmer who might be able to answer some questions about modifing voice chat in real time so players might sound like non human, distorted, etc. Potentially a small paid job!
#gamedev
A thing I wanted was an option inside the newly introduced account-settings that allows to switch between some hotkey presets.
*Ideally with full remapping, but that seems like a dream given how hard the implementation of font-switching was. Legacy apps sure are something...
While I am slowly moving everything from the div-hell from 19XX to sematic elements, there are operations that WILL require shortcuts.
To be precise, I currently have a pop up that users can toggle with `?` and a visisble button on screen that displays (and also reads out) the short cuts.
This seems to refer to websites. I guessed that something like that existed, but thx for the confirmation.
My question was sadly asked with web-applications in mind. I fear that semantic navigation can/will break down with them.
And since I am developing such an application, it looks lile they will.
Gen question: is there a list somewhere that lists/compares the key binds that screen readers (or even other accessibility tools) use?
Would be helpful to either replicate them or avoid using their keys.
The last genuinely new thing for dev that I noticed were the ok color spaces. And I'm a dork that scrolls through all new packages added to nix to see if there is something new.
Tldr; we should read more old books and papers.
And worse, since AI-people rely on the copy&compress machine, they are unaware of all the discussions that have been had already. I've not seen a single tool released in the last 3 years that has not already "existed" before.
All of which we had and studied since the 80s. It has been very rare to read an article that highlights something new. Or look at a tool that has not existed in a different form before hand.
They are not even improving it, but just reinventing.
I want to slightly push back on the pace thing. Things are "fast" if you look that the *number* of things released.
But this illustration falls appart if you categorize things by what they do/claim to do.
- chat interface
- auto completion
- solutions to self created problems (MSβ’οΈ)
If we were to integrate CSS into a C layouting engine and write a big header file; can we trick web devs into developing apps that are offline, multicore, don't need 4gb of ram, and are <20mb including assets?
At this point the only difference between developing for a browser and a normal OS is that the end user doesnt need admin privileges to install your software.
... and that linux is smaller than 500mb including tools while browsers are 2gb+ and still need the OS.
Last year, we launched a web app to assist trans people with legal name changes. Today, weβre removing the login wall and registration requirements and making everything publicly accessible.
This is the biggest embarrassment for multiplayer games imo: when using KLAC you have admitted that you could not design a protocol/server that could prevent this. Same level as needing admin access to launch. Why should I then trust you with my kernel, when you can't even program the easy part!?
Fairphone is pretty nice. I have de-googled it pretty easily and repairability and long term support is great.
Is there for you a point where you say for the purpose of public communication, nuance is destructive instead of constructive?
The output of a build command that shows that it has currently executed 85 out of 54650 steps
Excuse me, how many steps does this build process need!?
So pretty!
If it is so expensive in the US, that it's requiring a loan; might as well take that money and do uni in a country where it is free and you can explore a country you like when you have downtime.
Why not instead use tools like LSPs/tree-sitter/... that can semantically analyze and transform or generate code. They (can) provide better and faster code generation than LLMs, don't burn the world, and are consistent.
Use to me is time invested vs time saved in relation to quality+cost. LLMs produce worse code while requiring more time and money.
It seems obvious that the tool that has problems with logical consistency on the level of "a<b,b<c,a<c" can't produce things at higher levels.
But I trust a compiler. It is reproducible. It does the same thing every time and i can learn how it works and apply it.
I know how LLMs work technologically.
But I can't apply that knowledge to influence results without crafting a book as the system prompt. And even then it is not reliable.
All I've seen Copilot good for in my experience is the alienation from their own work. And no measurable increase in speed. But that's anecdotal.
This is not about speed.
Speed comes from good APIs, decent typing speed and small cycles and other markers of quality.
If you remove the pressure or "having to use the API yourself", you remove the learning opportunity to evaluate your design.
It simply creates bad incentives.
The feeling of competence those tools give is false, as one can notice the moment they are stripped away.
And the anchor bias results in suboptimal solutions.
I'm confused, how you can advocate for the learning and improvement through pairing/mobbing/...
And at the same time say LLM code has a "use"
I am not afraid of AI taking my job. I am afraid of it taking my skill.
I start on paper and end in code.
No reason to involve random text generation somewhere in that process.
I want to know what I'm doing.
That means reading docs and getting taught by other humans that are reliable.
Every java dev should have a quick look at docs.oracle.com/en/java/javase/<version>/docs/specs/man/index.html
Not many know what tools are actually in the JDK and just "use IDE to do everything".
(Yes, this was an IDE rant in disguise. Sorry, not sorry)
If you want to be user friendly, do it the minecraft way.
Package the app to an exe and have extensible parts as jars/dll/...