John Hilliard (@praetorian) 's Twitter Profile
John Hilliard

@praetorian

Making blockchains fast. Focused on performance, reliability, and tools at @0xPolygon || Boston || @MIT

ID: 983781

linkhttps://john.dev calendar_today12-03-2007 04:27:46

8,8K Tweet

1,1K Followers

956 Following

John Hilliard (@praetorian) 's Twitter Profile Photo

When you're unfamiliar, LLMs are a productivity boost. When you're deeply familiar, they're a bottleneck. Translating fast, abstract thought into step-by-step instructions slows you down.

Tib3rius (@0xtib3rius) 's Twitter Profile Photo

Looks like the Google Cloud incident report is out: status.cloud.google.com/incidents/ow5i… Summary: - On May 29, 2025, a new Service Control feature was added for quota policy checks. - This feature did not have appropriate error handling, nor was it feature flag protected. - On June 12,

John Hilliard (@praetorian) 's Twitter Profile Photo

CRUZ: no no no, that's not what I meant... TUCKER: you said Squidward is a squid. he's an octopus, Ted. he has a bulbous head and it’s confirmed in the DVD commentary. that's a fact. CRUZ [visibly sweating]: I - I mean symbolically, he... TUCKER: name one time he used ink. just

John Hilliard (@praetorian) 's Twitter Profile Photo

PebbleDB calls posix_fadvise(..., POSIX_FADV_RANDOM) to dissuade readahead. Perfect for some hardware; maybe wasteful when most of the SSTs could fit in the page cache? Anyone benchmarked a NORMAL or custom readahead build?

John Hilliard (@praetorian) 's Twitter Profile Photo

There were 3 different Semafor "vibe coding" articles in 3 days? Hard to tell if this is trend spotting or trend making to make this seem inevitable

There were 3 different Semafor "vibe coding" articles in 3 days? Hard to tell if this is trend spotting or trend making to make this seem inevitable
John Hilliard (@praetorian) 's Twitter Profile Photo

Coding is still open to everyone: download a compiler and hack away. But the LLM era brings rent... tokens for prompts, credits for builds, "pro" tiers for context. Productivity rises, but so does the cost of entry. Will future young devs get priced out?

John Hilliard (@praetorian) 's Twitter Profile Photo

Flame graphs are great for performance profiling. They're also underrated as architectural maps. I've been using them to (try to) trace the hierarchy of caches behind a typical SLOAD.

Flame graphs are great for performance profiling. They're also underrated as architectural maps. I've been using them to (try to) trace the hierarchy of caches behind a typical SLOAD.
John Hilliard (@praetorian) 's Twitter Profile Photo

TIL RLPx has a hard ceiling: the 3-byte length field limits every frame to 16,777,215 bytes (16 MB). That's a protocol limit, not just an arbitrary setting.

TIL RLPx has a hard ceiling: the 3-byte length field limits every frame to 16,777,215 bytes (16 MB).
That's a protocol limit, not just an arbitrary setting.
John Hilliard (@praetorian) 's Twitter Profile Photo

I think this is only the second time I've heard of the parliamentarian stepping in... this time it's to block Mike Lee's plan to sell millions of acres of federal land. He tried to tuck it into a tax bill apnews.com/article/public…

John Hilliard (@praetorian) 's Twitter Profile Photo

Planning to stress-test a new KV layout: a memory-mapped, copy-on-write B+-tree index that points to an append-only value log (key -> (offset, len)). Hopefully we can trade a bit of disk for better page cache utilization and lower tail latency on random reads.