Kelly Sommers

You asked for longer rants, so here are longer rants! After almost 10 years I'm back to blogging. Thanks for the encouragement. Link below in the thread 🔻

@dave_cramer 😂 they totally ruined their team in 1 calendar year haha

@kellabyte They sell out every night why would they spend money trying to win

Let me get this straight. The Toronto Maple Leafs hired a GM who had 6 computer screens up with statistics and dashboards. IT was called and they discover 5 of the computers are fake with fake images displayed. This is who the Toronto Maple Leafs hired as top candidate? 😂

Most database benchmarks are bad. The problems are often subtle. Mismatched instances, cold caches, and coordinated omission all produce misleading results. Today, we’re publishing guidance on how to do it right, and updating our acceptable use policy to allow benchmarking.

@BenjDicken Stop doing graph theory in a GC and just nuke the block of memory when the request is complete.

ZIG IS SO COOL I've been learning about Zig's memory allocators, specifically ArenaAllocator are amazing. Also love the simplicity of `defer` for guaranteeing code execution on function /scope end. Going to stream on Zig + Philosophy of Software Design tomorrow 10am.

@BenjDicken For a long time I’ve felt like for backend services the generalized GC we see in Java/Go/C# is just super generic but backend services are mostly ephemeral. Everything a request processes typically can be released except maybe a few things. Arena allocators make a lot of sense.

@sunbains Is this going to be open source? I’m curious to check it out!

When you are on a roll, don't stop. I think I'm mostly done with this. The overhead is actually from the interface not the server code. Single thread performance is now at: 813,181 rows/s Scales linearly up to 32 threads and then a smaller jump from 21m rows/s to 27 rows/s.

@myfear @thsottiaux I use farfield on GitHub for this today but it’s janky

@kellabyte @thsottiaux THIS 👆🏻👆🏻👆🏻🫶🏻

What are we obviously not getting right with Codex?

@hibaymj Think of it like waterfall vs agile. I’m not an agile person but one of the key findings was that too much up front design and too little ability to adjust = bad software and blown delivery dates. Same thing here. Accept iteration and agility. Great specs and agents help!

@kellabyte I guess the assumption here is you haven’t already pushed down those guard rails into skills or agents? Probably completely foolish to one shot and never read again, but I’ve seen some wild (and good) output from a single (final) prompt.

"One shotting" an entire implementation is the dumbest thing I've ever heard (for now). For sure you can get a lot done in a batch but removing yourself from the architectural decision steps along the way that you discover as you measure is foolish.

@hibaymj I think designing an entire system before beginning removes a lot of important iteration discovered along the way. Yes great structure and specifications can raise quality and chance of success. But it doesn’t mean you’re going to get the best end result.

@reubenbond Thanks for all the kind words! It’s been really fun. I struggle to figure out how mature some parts of the system are when you’re building 5 years of DB engineering in 30 days. I am trying to learn how to measure performance and correctness. It’s still a toy but exciting!

@kellabyte It's fun to see how far you can push vibe coding for a highly sophisticated distributed database - and all the guard rails you've set up to make it work. I only started looking into Accord after you posted about it.

I'm so sorry about all the HoloStore tweets lately. I've been struggling w/ burnouts & all the things I wanna fix, try, build have lived rent free in my head. I just don't have enough hrs in the day. I'm excited for the first time in years cuz I've become less of a bottleneck

Jensen is one the smartest and most far seeing folks the world. "If an AI scientist warns people that AI is going to permeate across radiology and radiologists are going to get wiped out, it might seem helpful but it's hurtful. If we convince everybody not to be radiologists and we now need radiologists, that actually is hurtful to society. "It is hurtful to convince all the young college graduates not to study software engineering because we are going to need more software engineers than ever. That's hurtful." "Scaring people with nonsensical things, which are not going to happen, that this is an existential threat, there's a 20% chance that is is existential, that's ridiculous. "That it's going to wipe out 50% of college level jobs. "That is it going to completely destroy democracy. "These kinds of comments are not helpful. They are made by...CEOS. And you become a CEO, maybe you adopt a God complex and somehow you know everything." Brutal. And right.

@dave_cramer 😂

@kellabyte did it once, had no idea what it built. Never again.

If you’re hacking on something cool crank this tune to MAX volume! 🔻

Too excited to sleep apparently. Gotta keep bumping those agents forward.

@kellabyte porcupine saved me so much pain

Porcupine linearizability tests & Antithesis are my heroes right now It's REALLY easy to build a naive distributed system that you _think_ works. Support correcting data manually on support calls Whether a DB or multi-API interaction. It's never been easier to test anomalies