Jake Lucas

Programmer. Videos on

[Jake Lucas @polyscone]

Finished enough of the resolver in my assembler now, so next I'll be moving onto actually trying to lower parsed instructions down to machine code through my x64 encoder https://t.co/qtP4fwhMKd

[Jake Lucas @polyscone]

@ChShersh I've only just managed to get used to using `git switch` for changing existing branches but still type `git checkout -b` for creating a new one...

[Jake Lucas @polyscone]

AI writes worse code than humans can. Quickly producing bad code that works in specific cases isn't the same thing as writing good code. AI code works, but it's still bad. Focused AI usage gives better results, but then I may as well just be writing the code myself anyway...

[Jake Lucas @polyscone]

@EliasDaler @karpathy I've consistently found the more it writes the more time I have to spend checking and correcting it, to the point it may as well have not even been used...

[Jake Lucas @polyscone]

If you know software isn't "solved" and you know that writing code by hand is still a good thing to do, then you might enjoy following along as I tap out a compiler toolchain from scratch from the bottom up; currently up to the assembler https://t.co/pzEUirHFuK

[Jake Lucas @polyscone]

@yutongwu111140 We need more Ada_Case

[Jake Lucas @polyscone]

@signulll If this is what people thing then things aren't looking good for software in general...

[Jake Lucas @polyscone]

I've been writing a compiler toolchain from scratch, starting with an x64 encoder, ELF/COFF object file writers, and now onto an assembler https://t.co/pzEUirHFuK

[Jake Lucas @polyscone]

Started the implementation of symbol resolution and const expression evaluation

[Jake Lucas @polyscone]

@bojang I've still only seen marginal speed increase from people around me, and not even consistently, pretty much always at the expense of quality. Speed isn't everything, so I don't get why so many are willing to make the trade-off.

[Jake Lucas @polyscone]

Added unary operators and align/attributes to object declarations

[Jake Lucas @polyscone]

Still a couple of missing pieces, but the parser is mostly done now. Time to move onto semantic analysis and I'll come back to finish off the remaining bits of the parser later.

[Jake Lucas @polyscone]

I don't get the idea that languages don't matter anymore. Even if you use AI to do your programming, each language still provides different structure and guarantees over whatever the AI feels like doing with something like assembly.

[Jake Lucas @polyscone]

Now successfully parsing section and object declarations

[Jake Lucas @polyscone]

@valigo I wonder how much overlap there is between those people and the people who pretend no one knows what they're doing to encourage beginners...

[Jake Lucas @polyscone]

@TheGingerBill @graphitemaster I don't see how the idea of technical excellence is misaligned or higher now because of LLMs? All it means for me is that there's a lot more unimpressive code out there in the same way that copy pasting someone else's code is unimpressive.

[Jake Lucas @polyscone]

Recursive descent expression parsing done

[Jake Lucas @polyscone]

Setting up parser tests that are just in-source s-expression comments is actually pretty easy if you have the ability to lookup trivia for source byte offsets https://t.co/WPBM9Q3VMw

[Jake Lucas @polyscone]

It's true https://t.co/IMkkxC1tso

[Tim McNamara @timClicks]

Assembly language is too high level. Too many macros and lots of unpredictable codegen. Real programmers write in hex codes & drop down to microcode for performance.

[Jake Lucas @polyscone]

I've started work on my assembler today, since I've already got an x64 instruction encoder and ELF/COFF writers done