Joseph Garvin
@josephhgarvin.bsky.social
voicecoding latency and throughput hacker. How did I get here and what am I doing in this hand basket? he/him
"Why are there so many instantiations of this template? I only use one set of types with it."
*several minutes of reading ridiculously long types*
"Oh mothe-"
*several minutes of reading ridiculously long types*
"Oh mothe-"
November 10, 2025 at 11:06 PM
"Why are there so many instantiations of this template? I only use one set of types with it."
*several minutes of reading ridiculously long types*
"Oh mothe-"
*several minutes of reading ridiculously long types*
"Oh mothe-"
Finding goto is still relevant for code size optimization. Several early exit paths that all need to construct some return object? Better to construct and return in one place and change all the early returns to gotos. Clang/GCC don't optimize it reliably (look for -1640531527).
November 7, 2025 at 5:54 PM
Finding goto is still relevant for code size optimization. Several early exit paths that all need to construct some return object? Better to construct and return in one place and change all the early returns to gotos. Clang/GCC don't optimize it reliably (look for -1640531527).
I'm sure it will increase the quality of discourse to prevent people publicly quoting what people publicly said. Definitely won't have the unintended effect of making things easier for grifters.
November 7, 2025 at 5:09 AM
I'm sure it will increase the quality of discourse to prevent people publicly quoting what people publicly said. Definitely won't have the unintended effect of making things easier for grifters.
Chat, should I write a linker?
November 2, 2025 at 9:40 PM
Chat, should I write a linker?
Pop quiz: Can you guess what this prints and why?
October 24, 2025 at 6:56 PM
Pop quiz: Can you guess what this prints and why?
Can you spot the bug?
October 20, 2025 at 5:29 PM
Can you spot the bug?
Lay people won't appreciate why this is impressive but this is one of the craziest things I've seen GPT5 do. I gave it this graph from FT and asked it to add Facebook adoption for comparison, and rather than generate a new graph it seemingly correctly superimposed a new line!
October 19, 2025 at 5:40 PM
Lay people won't appreciate why this is impressive but this is one of the craziest things I've seen GPT5 do. I gave it this graph from FT and asked it to add Facebook adoption for comparison, and rather than generate a new graph it seemingly correctly superimposed a new line!
So is it just the same old farts talking to each other now?
October 16, 2025 at 8:37 PM
So is it just the same old farts talking to each other now?
No but really, compilers are not that smart (at least GCC isn't, Clang gets it).
October 8, 2025 at 8:34 PM
No but really, compilers are not that smart (at least GCC isn't, Clang gets it).
Don't flatter me Claude
October 7, 2025 at 9:08 PM
Don't flatter me Claude
Compilers are not *that* smart
October 5, 2025 at 4:01 PM
Compilers are not *that* smart
Pop quiz: you look at your disassembly and you can see something is weird about these call instructions. What's going on?
September 18, 2025 at 8:32 PM
Pop quiz: you look at your disassembly and you can see something is weird about these call instructions. What's going on?
Still not worried about AGI anytime soon. Any basic math slightly off the beaten path is still wrong.
September 16, 2025 at 6:24 PM
Still not worried about AGI anytime soon. Any basic math slightly off the beaten path is still wrong.
Interesting retrospective 👇 on why V8 team abandoned Sea of Nodes. Have to say SoN seems obviously correct, is Javascript just not designed to leverage it. AFAICT they treated memory as one giant register re: dependencies; langs like Rust w/ better aliasing info should do better
August 27, 2025 at 4:45 PM
Interesting retrospective 👇 on why V8 team abandoned Sea of Nodes. Have to say SoN seems obviously correct, is Javascript just not designed to leverage it. AFAICT they treated memory as one giant register re: dependencies; langs like Rust w/ better aliasing info should do better
Are you kidding me Intel
August 20, 2025 at 4:43 PM
Are you kidding me Intel
What drugs were the Intel engineers on? "Oh you want to encode this instruction? Look up the opcode row and column by nibble, then inside the box for each operand lookup in 2 other tables what the uppercase and lowercase letter mean. Then you will know 1% of what you need."
August 18, 2025 at 2:26 AM
What drugs were the Intel engineers on? "Oh you want to encode this instruction? Look up the opcode row and column by nibble, then inside the box for each operand lookup in 2 other tables what the uppercase and lowercase letter mean. Then you will know 1% of what you need."
C++ horror quiz: Can you guess what I'm doing here?
August 17, 2025 at 8:11 PM
C++ horror quiz: Can you guess what I'm doing here?
Is this true? Never considered it.
August 10, 2025 at 5:59 PM
Is this true? Never considered it.
Linus is mad at make_u32_from_two_u16 for the same reason loop combinators are annoying. When you create new verbs for trivial ops you replace needing to know how to do the op with needing to know the name of the verb and how to call it and there is little to negative savings.
August 10, 2025 at 5:36 PM
Linus is mad at make_u32_from_two_u16 for the same reason loop combinators are annoying. When you create new verbs for trivial ops you replace needing to know how to do the op with needing to know the name of the verb and how to call it and there is little to negative savings.
TIL C++ UDL can't handle extreme negative numbers properly because the type you are passed by the compiler is `unsigned long long` and you can't tell whether the literal is attached to an unary minus operator (negative sign).
August 6, 2025 at 6:21 PM
TIL C++ UDL can't handle extreme negative numbers properly because the type you are passed by the compiler is `unsigned long long` and you can't tell whether the literal is attached to an unary minus operator (negative sign).
I went to the Internet archive and the official Congress Constitution page to verify this myself and it is real.
August 6, 2025 at 3:46 PM
I went to the Internet archive and the official Congress Constitution page to verify this myself and it is real.
I think Opus 4.1 is the first model to play Connect4 to completion with no hallucinations and correctly recognizing that I won. Still failed to recognize it could have beaten me by dropping a piece in the last column (previous models made me careless!).
August 5, 2025 at 10:17 PM
I think Opus 4.1 is the first model to play Connect4 to completion with no hallucinations and correctly recognizing that I won. Still failed to recognize it could have beaten me by dropping a piece in the last column (previous models made me careless!).
Are you kidding me
August 4, 2025 at 8:02 PM
Are you kidding me
Today in "reasons to always actually check the data"
July 24, 2025 at 5:21 PM
Today in "reasons to always actually check the data"
Calling all x86 wizards, earn some magic internet points (link below)
July 17, 2025 at 5:36 PM
Calling all x86 wizards, earn some magic internet points (link below)