mrlongroots

6 ms·

31.
by mrlongroots 1y ago
The other funny bit is that one-way PCIe latency is 250ns-ish (don't quote me on the exact numbers), which imposes a hard 1us constraint on latency between two hosts.
32.
by mrlongroots 1y ago
I think standard relational databases/schemas are underrated for when you need richness. OTel or anything in that domain is fine when you have a distributed callgraph, which inference with tool calls does. I think the fallback layer if
33.
by mrlongroots 1y ago
Our social contract bestows individuals with freedom and governments with the monopoly of violence to police the limits of individual freedom. If the acts of law abiding individuals (or groups) are a net negative for society, that is not an
34.
by mrlongroots 1y ago
If a selection mechanism is orthogonal to a property, it seems weird to argue that the selected subset is distributed differently along that axis than the broader population.
35.
by mrlongroots 1y ago
> "oh it's just database joins" is about like me ripping your arms off and describing it as "chemical reactions" This argument is both inconsistent and counterproductive. Inconsistent as in, the harm to me from h
36.
by mrlongroots 1y ago
> No doubt Deloitte or any other contractor shop would be able to do the same thing - but they don’t choose to. I'm sorry but I absolutely disagree that the reason say Deloitte is leaving a few hundred billion dollars on the table i
37.
by mrlongroots 1y ago
Kitchen knives murder people. Toyota Hiluxes have powered more jihad than modern battle tanks. Our tastes, beliefs, and opinions as a society are shaped by recommendation algorithms run by facebook/instagram/twitter, to our profou
38.
by mrlongroots 1y ago
I think Palantir is highly misunderstood. As a technology, it is just database joins. It is just that they are able to pull in data from everything from S3 to SAP to ArcGIS, and provide analytics, visualization etc. on top to provide global
39.
by mrlongroots 1y ago
> The CPU can load from these virtual addresses directly into registers This requires that device to bring meaningful amounts of its own memory. GPUs do that with VRAM. A storage device does not come with its own RAM, but interesting po
40.
by mrlongroots 1y ago
From a response elsewhere in this thread (my current understanding, could be wrong): > You can temporarily stream directly to the L3 cache with DDIO, but as it fills out the cache will get flushed back to the main memory anyway and you w
41.
by mrlongroots 1y ago
> This was a comparison of two methods of moving data from the VFS to application memory. Depending on cache status this would run the whole gambit of mapping existing memory pages, kernel to userspace memory copies, and actual disk acce
42.
by mrlongroots 1y ago
Agree: flat/relational structure is highly efficient to store, query planning/optimization/scaling become a lot easier, the language is clunky but relational algebra is beautiful, but... I think pushing multiple joins to what
43.
by mrlongroots 1y ago
You don't need hugepages for basic 5GB/s sequential scans. I don't know the exact circumstances that would cause TLB pressure, but this is not it. You can maybe reduce the number of page faults, but you can do that by walking
44.
by mrlongroots 1y ago
Yes, it doesn't take a benchmark to find out that storage can not be faster than memory. Even if you had a million SSDs and somehow were able to connect them to a single machine somehow, you would not outperform memory, because the dat
45.
by mrlongroots 1y ago
Another thought: this is not meant as criticism or anything, just an aspect of performance that I thought was interesting but not covered by the blog. A test that would show PG's strengths over ClickHouse for OLTP would be a stress tes
46.
by mrlongroots 1y ago
Yep! A short summary of my n=1 findings (currently very speculative, but also sound I think, am meeting with my cardiologist next week to see what he thinks). The type of ADHD I have seems to have an "autonomic nervous system impairmen
47.
by mrlongroots 1y ago
Ok I think the confusion arises because of the probabilistic nature of LLM responses that blurs the line between "intelligent vs not". Let's take driving a car as an example, and a random decision generator as a lower bound o
48.
by mrlongroots 1y ago
A rebuttal is just an alias for "counterargument", it does not define the structure of the counterargument. However flawed, what I said did have a structure (please refer to my other response in this thread for why).
49.
by mrlongroots 1y ago
> you have stated an assumption I don't think I stated an assumption, this is an assertion, worded rhetorically. You are welcome to disagree with it and refute it, but its structural role is not that of an assumption. "Can an L
50.
by mrlongroots 1y ago
I have diagnosed ADHD and I agree largely with the person you are responding to. The claim is not that ADHD is not a set of people with real psychiatric disorders, but it is a loose umbrella for what are actually disparate problems. I recen
51.
by mrlongroots 1y ago
They could've just used `clock_gettime(CLOCK_MONOTONIC)`
52.
by mrlongroots 1y ago
> they will struggle just as a model does surly A human will struggle, but they will recognize the things they need to know, and seek out people who may have the relevant information. If asked "how are things going" they will r
53.
by mrlongroots 1y ago
I very much disagree. To attempt a proof by contradiction: Let us assume that the author's premise is correct, and LLMs are plenty powerful given the right context. Can an LLM recognize the context deficit and frame the right questions
54.
by mrlongroots 1y ago
It seems that most approaches treat neutron diffusion as memoryless (or at least stochastic): https://en.wikipedia.org/wiki/Neutron_transport
55.
by mrlongroots 1y ago
> Consider the nuclear reaction metaphor. It's clearly not memoryless. Eventually you'll run out of fissile material. I think no real process is memoryless: time passes/machines degrade/human behaviors evolve. It is a
56.
by mrlongroots 1y ago
It could, but important to keep in mind that the filesystem architecture there is also very different with a parallel filesystem with disaggregated data and metadata. When you run `ls -l` you could potentially be enumerating a directory wit
57.
by mrlongroots 1y ago
I think "making an unreliable system emulate a reliable one = bad" is too simplistic a heuristic. We do this all the time with things like ECC and retransmissions and packet recovery. This intrinsically is not bad at all, the ques
58.
by mrlongroots 1y ago
Even as a matter of safety/investigation hygiene, "pilot error" should be the conclusion of last resort, arrived at after months of poring over data, and because nothing else seems viable. If we decided to pin all aviation in
59.
by mrlongroots 1y ago
This is very clearly EAFR data, so the logical/electrical switch state. Nothing about the mechanical state of the switches has been mentioned, except a picture that shows their final state to be in the RUN position (which makes sense g
60.
by mrlongroots 1y ago
At the token layer an LLM can make things up, but not as part of a structured pipeline that validates an invariant that all suggestions are valid entities in the database. Can google search hallucinate webpages?

More ›