🇺🇦 Ingvar Stepanyan
banner
rreverser.com
🇺🇦 Ingvar Stepanyan
@rreverser.com
Sr. Principal Engineer at Cloudflare by day, WebAssembly consultant by night.

You might also know me from my work on OSS tools and libraries (JS / Wasm / Rust) or Wasm DevRel at Google Chrome

📝 https://rreverser.com/
📷 https://instagram.com/rreverser
WebAuthn works with on-device biometrics too which most modern devices have and users are already familiar with.
November 15, 2025 at 1:39 PM
Dammit I came to write this but looked at the existing replies.
November 15, 2025 at 1:33 AM
Not enough traits. What if some of your users need to parameterise it, huh? Huh??
November 14, 2025 at 11:14 PM
Github Copilot already has regular automatic "summarising conversation" where it generates concise instructions for itself when getting close to the context window so that it doesn't forget any important high-level points.

I assume same can be reused for git history.
November 12, 2025 at 2:18 PM
[nods knowingly] that's why they call it diff3
November 12, 2025 at 2:13 PM
Huh. Didn't hear about that, but I'm hard pressed for cases where I wouldn't value correctness over agent doing seemingly random changes in a loop (which still happens with Sonnet, and especially infuriating when I see it trying exactly same things it tried before, and just burning time and money).
November 10, 2025 at 4:26 PM
It does seem to integrate a bit worse as a standalone agent than Sonnet, but I guess it's a matter of time.

When I need a surgical improvement / implementation, it works a lot better.
November 10, 2025 at 4:22 PM
Yuppp, it's pretty unhinged one. Like most weird instructions, apparently motivated by cryptography.
November 10, 2025 at 3:47 PM
This is also one of the reasons I strongly prefer "massaging" data to make it autovectorizer-friendly and only dropping down to raw SIMD when necessary as suggested by the profiler & generated asm.

Autovectorizer is pretty powerful nowadays and has lots of hidden goodies I wouldn't think of.
November 10, 2025 at 1:52 PM
Reduced example of my code similar to where I found it - reversing bits in 8 u64s at the same time: rust.godbolt.org/z/PYjYoxeYz
November 10, 2025 at 1:45 PM
I tried so many prompts, and eg ChatGPT outright refuses to stop asking "Do you want me to blah blah?" at the end of each response. Best case, it apologises, and in the next answer uses "If you want, I can blah blah." instead.

I don't need your random suggestions that ruin my train of thought!
November 9, 2025 at 6:50 PM
And yes, I know about personalisation prompts, yet way too often models ignore them either because of the bias in the training data, the weight of the built in prompts, or because context window ran out and it forgot all about my personalisation.

As a user I shouldn't have to even know why.
November 9, 2025 at 6:48 PM