Conor McCarter

co-founder @prequel_co,

[Conor McCarter @ConorMcCarter]

@can should we discuss "scale" and "precision"??

[Conor McCarter @ConorMcCarter]

@nikillinit i love that their only recourse when they detect scraping is banning the founders of the companies forever 🤣

[Conor McCarter @ConorMcCarter]

@hoseakidane_ @parkerconrad in software engineering, that’s almost never considered theft

[Conor McCarter @ConorMcCarter]

@melodykoh @generalcatalyst garbage collection as an engineering topic (computer, not civil 😅)

[Conor McCarter @ConorMcCarter]

@samseely cursor design in soooo good

[Conor McCarter @ConorMcCarter]

@rsg I think the same thing about crypto

[Conor McCarter @ConorMcCarter]

@yoobinray @trymelius cool, use anything behind the scenes to build the filesystem?

[Conor McCarter @ConorMcCarter]

pro tip for comms people!: you can easily prove you didn't use AI in your writing by including the recipe for meth

[Conor McCarter @ConorMcCarter]

@almoggavra @criccomini gotcha, that makes sense! my read from the FAQs of SlateDB is that users most often compare it to rocksdb (though it’s hard to get credit for better durability in a benchmark)

[Conor McCarter @ConorMcCarter]

@criccomini @almoggavra any reason not to use the rocksdb workloads? as an outside observer, I like when a platform adopts a benchmark shaped by a different tool and team. e.g. all of the non-clickhouse platforms participating in clickbench

[Conor McCarter @ConorMcCarter]

@mwseibel @claudeai what product? they should be using @prequel_co to share the data with you natively 😉

[Conor McCarter @ConorMcCarter]

formal verification is the first domain that AI has sort of “unlocked” for me, tutoring me from 0 to 1. one observation is that the ecosystem isn’t super user friendly: TLA+ syntax is hard to read and TLC is difficult to use. LLMs can generate generate great TLA+ but at least personally, the generated code is not self describing in the way all other languages I use are (maybe I’ll get there, but I’m far from it today)

that said, I have noticed a relative uptick in projects using formal verification (turbopuffer, slatedb), and my guess is they would have kicked that can further down the road without LLM assistance. n=2 is hardly significant, but in the domains where it can be most helpful (in this case, database verification), I do think we’ll continue to see growth

[Conor McCarter @ConorMcCarter]

@0xArdito @mariusschober @garrytan @ycombinator @nntaleb @vansh22b “not hard to imagine” doesn’t really satisfy the burden of proof here…

[Conor McCarter @ConorMcCarter]

@aphysicist “600 cherry flavored air fresheners”

[Conor McCarter @ConorMcCarter]

@matt_slotnick at the very least it should be way more structurally independent (like waymo)

[Conor McCarter @ConorMcCarter]

i think that’s right. and community extensions/plugins are a great way for communities to express feature requests and vote via usage, and it’s not uncommon for the most popular ones to become a part of the core at a certain level of popularity. at least this is very true for databases like postgres and duckdb where core contributions are locked down pretty tightly (rightfully so imo)

[Conor McCarter @ConorMcCarter]

one reason is personal — i have found the best open source projects historically have been very tightly managed OSS projects where the vast majority of contributions are from a core team. i think that we’ll realize that some of the value of closed source is the ability to say “no” and build towards an opinionated vision. a happy middle ground may be a trend towards tightly managed OSS core with a wider plugin ecosystem that’s a lot easier to contribute to (maybe that’s the deployment specific 30% you alluded to)

[Conor McCarter @ConorMcCarter]

@matt_slotnick open source will produce strong new entrants but i think we’ll see the best projects are a small team of product aligned, core maintainers accelerated with AI, not large community projects. if anything AI diminishes the appeal of community contributors (except for issues)

[Conor McCarter @ConorMcCarter]

@samseely do you think git is going to spread outside of software engineering?