Hacker News: jeffbee

New comment by jeffbee in "An agent used DNS to reach an external chatbot"

jeffbee — Sat, 26 Sep 2026 23:40:05 +0000

Yeah I don't get it, either. If the exercise relies on the assumption that the agent can't reach the "live internet", whatever that means, there are affirmative steps to realize that assumption. The fact that they failed to take those steps suggests two possibilities: they are idiots, or they think we're idiots who will fall for this marketing campaign.

New comment by jeffbee in "AI safety is mostly a sex cult in Berkeley"

jeffbee — Thu, 24 Sep 2026 20:26:18 +0000

LOL. If true, absolutely the wrong strategy to publicly apologize.

New comment by jeffbee in "Linux support is coming to Snapdragon X2 series"

jeffbee — Thu, 24 Sep 2026 17:49:42 +0000

They decided that instead of having an Android runtime on top of ChromeOS they would port the ChromeOS desktop features to Android. A Googlebook boots Android.

New comment by jeffbee in "AI safety is mostly a sex cult in Berkeley"

jeffbee — Thu, 24 Sep 2026 16:55:56 +0000

Berkeley is a hotbed of reactionary, far-right, and generally weird behavior. We've got this sex cult of course, and we have Curtis Yarvin, Michael Shellenberger, etc. Our tolerance for deviance acts as a shield that attracts self-depicted contrarians.

New comment by jeffbee in "Best LLM for every budget, updated daily"

jeffbee — Thu, 24 Sep 2026 16:50:20 +0000

When a clanker is slopping up a webpage, why does it have to be #f4f3ee? Curious.

New comment by jeffbee in "What California is learning from solar panels built over irrigation canals"

jeffbee — Wed, 23 Sep 2026 15:34:24 +0000

I don't really love how this completely imaginary, totally made-up, unsupported anecdote has caused 40% of the posts on this HN story.

New comment by jeffbee in "What Sun got wrong"

jeffbee — Mon, 21 Sep 2026 16:07:27 +0000

At my employer in 1999 we were maintaining home-grown binary modification to the Solaris kernel so that TCP segment send and receive were not O(N^2) where N was the number of addresses on the interface :-/ It was, as far as I could tell, an operating system that nobody had ever tried to use before. For anything where we could avoid Sun we used FreeBSD. One of the nice things about the free Unixes of that time was that process and thread creation or destruction were orders of magnitude faster than on Sun. To me, Sun was obviously in a lower category than Linux/FreeBSD/NetBSD.

New comment by jeffbee in "What Sun got wrong"

jeffbee — Mon, 21 Sep 2026 16:03:28 +0000

Right before they fell face-first into the grave, they did finally ship the v20z, which was a completely reasonable way to run Solaris x86. Its management board ran Linux on PowerPC, for irony purposes.

New comment by jeffbee in "What Sun got wrong"

jeffbee — Mon, 21 Sep 2026 15:58:17 +0000

Counterpoint: it actually sucked. Entire universities lost a day's work when their editors were all fatally stuck in an rpc.lockd bug.

New comment by jeffbee in "Jemalloc 5.4.0"

jeffbee — Fri, 18 Sep 2026 17:30:59 +0000

Sure, also true that the per-CPU scheme co-evolved with the proliferation of services with thread-per-request architectures having way more TIDs than cores.

New comment by jeffbee in "AI is an elite crime spree"

jeffbee — Fri, 18 Sep 2026 16:15:33 +0000

Can we stop acting like selective quoting from discovered materials gives us an unbiased and factual representation of the truth? The plaintiff is always trying to cherry-pick and distort elements from discovered material. And this is exactly why every large corporation without exception tries to train their employees to not draw conclusions of law in discoverable written materials. A computer science professor does not have the requisite knowledge to determine that some act was "theft of labor" nor the knowledge required to contextualize that "in human history".

New comment by jeffbee in "Jemalloc 5.4.0"

jeffbee — Fri, 18 Sep 2026 15:42:37 +0000

> on other platforms, I am not so sure

tcmalloc only works on Linux.

New comment by jeffbee in "Jemalloc 5.4.0"

jeffbee — Fri, 18 Sep 2026 15:41:20 +0000

Exactly. You want memory arenas that are hot in this CPU's caches. If your thread moves, its per-thread caches are now elsewhere. Original TCMalloc was developed in the days of 2-4 core servers. Current TCMalloc was an evolution in the context of 32+ core servers.

New comment by jeffbee in "Waymo in Singapore"

jeffbee — Fri, 18 Sep 2026 05:47:20 +0000

No, but production ceased pretty recently and Waymo still has an inventory of them with zero miles.

New comment by jeffbee in "Astra for Law"

jeffbee — Fri, 18 Sep 2026 03:37:57 +0000

How much experience do you have with LLMs exactly? It would be consistent with my experience if Claude stuck in a line of python that just emits a JSON literal with no justification, potentially buried in a large program where an untrained person might not notice it. I don't even trust them if the output consists of structured data paired with source images from the PDF, because I've experienced LLMs fabricating the source rectangles to match the output. I only use tools like this by asking for programs, because as you note LLMs are good at that, and the verification process consists of tool calls to legitimate PDF manipulation tools so I have some confidence everything is above board. Even then I only do this for hobbies, not anything that matters.

New comment by jeffbee in "How Uber Protects Against Retry Storms"

jeffbee — Fri, 18 Sep 2026 03:26:14 +0000

> 2. Whenever a service retries but has to give up, the error it sends to its callers should never be retried. There has to be some agreement that that HTTP code will never be retried. This prevents the multiplicative factor of retry on top of retry, which is why those storms can generate so much load.

This can't really be tolerated in practice, though, because it means that one bad component somewhere in your stack, one that is able to accept and respond to requests but for whatever reason isn't able to make requests to its backends, poisons the whole stack. You can't take one backend's word for it that the failure is not localized and therefore retryable.

New comment by jeffbee in "How Uber Protects Against Retry Storms"

jeffbee — Fri, 18 Sep 2026 03:24:17 +0000

Just limiting your retry budget to 1% of normal rates using a client-local token bucket with no distributed coordination will eliminate the possibility of long-lived retry storms.

New comment by jeffbee in "A 32-year-old bug walks into a Telnet server"

jeffbee — Thu, 17 Sep 2026 05:20:01 +0000

RISC was not a distant dream in 1994 LMAO.

New comment by jeffbee in "Backups Aren't Simple"

jeffbee — Thu, 17 Sep 2026 02:24:37 +0000

WiFi is a little faster I guess? In the primitive days people used the EyeFi.

New comment by jeffbee in "Backups Aren't Simple"

jeffbee — Thu, 17 Sep 2026 01:25:20 +0000

Mainstream DSLRs with dual card slots came out 22 years ago.