@jbm_1992 I agree. I actually wrote about it today: https://t.co/1cpEtiLyOe
I think they might realize the DTC offering offers no moat, and may “refound” into an agent centered company à la Harvey for health
“Just add padding and _mm_pause” sounds great until you actually measure it.
In this experiment, one pause per retry made the queue slower. Bounded backoff helped.
Next in my C++ concurrency series, with runnable code and results:
https://t.co/sHbNfwR4ED
This is a good filter for people who I don’t want to work with. Please understand how computers work. This isn’t even about understanding Docker. Any software person should know how the layers work at a high level and this is a VERYYYYY high level.
Several producers, one consumer. One shared queue or a queue per producer?
I compared SPSC fan-in and MPSC with runnable C++20 examples and initial benchmarks. AWS bare-metal testing comes later in the series.
https://t.co/OVyxrL2v0x
#cpp#concurrency#lowlatency
@aslambeg84 Time is lost everywhere, in the actual searching for records, the fetching of the indexes from memory assuming they are not found in memory caches, the actual data that the indexes / keys point to being in main memory and having to be pulled into caches, scheduling, NUMA traffic
@GohilHardy Honestly lukewarm take. The people I see doing this are overwhelmingly “founders” who are non technical, or people saying they just created acrobat for free and they are going to release their “suite of products” to the world, like Libre Office doesn’t exist
“Just replace the mutex with atomics” can turn a performance problem into a correctness problem you won’t notice until production.
I wrote about why, with C++ examples you can run and assembly you can inspect.
https://t.co/WFbJ470iyV
Im sorry im not buying this ‘not reviewing code’ thing. Maybe later, but certainly not now, even with frontier models with max thinking
Every single day I’m catching nonsense. If you’re merging this shit without looking, the systems your working on are either
- not important and/or
- have zero consequences for being wrong
@SyntheticBeef How would Claude know hthat an upstream dependency is giving you bad data and causing a crash? How would Claude know the data is actually wrong? I’ve had instances where Claude proposed that the issue was in the Linux kernel and not because data was mis configured
@SyntheticBeef I mean this is such a stupid comment. Apparently you are a dev since 1996, so you should know that production can break because of logic errors or data errors, but maybe you haven’t worked on anything complex.
@DanKulkov You’re missing nothing. @theo has the arrogance of a principal engineer and the résumé of a junior. An internship, 4 years bouncing across 9 teams at Twitch, then startups and reaction videos. He’s remarkably condescending for someone with so little demonstrated engineering depth
The thing that scares me about AI isn’t hallucinations. It’s watching experienced engineers lose their shit over “optimizations” that are either obvious, complete bullshit, or outright regressions.
https://t.co/mB5JD4FCwR
#AI#SoftwareEngineering#Programming#LLMs