mrlongroots

100 ms·

1.
by mrlongroots 1mo ago
In my experience, code is a small fraction of the work. I'm in an infra team and for the last 2 weeks or so I've been trying to understand whether a particular workload will catch fire if a switch is flicked. I'm also new to
2.
by mrlongroots 2mo ago
Opinionated take: ingestion should be treated as a streaming reorganization workload, separate from whatever a "database" is. You do not even need to change the Iceberg spec, although you could. Just rejig things a bit to produce
3.
by mrlongroots 3mo ago
> And what might be the outcome? DoE compute budgets are ~10B USD across labs. AI training is a trillion-dollar workload. Different league.
4.
by mrlongroots 3mo ago
All that would not help you with an AI training cluster interconnect. See Amin Vahdat's keynote at HotInterconnects 2025. Everyone is building a fabric for this stuff from scratch (Google/Falcon, Amazon/EFA, Azure/MANA,
5.
by mrlongroots 3mo ago
It is partly this and partly a funding vehicle for American next-gen computing. It is not that hard to estimate FP64 ballpark from a whole bunch of public statistics. And it takes a looot more than raw FLOPs to get a simulation working. And
6.
by mrlongroots 3mo ago
Today's limits are known and undisputable. Tomorrow's limits are a promise: some promises over-deliver, others under-deliver. :) Regardless, to bring the discussion back to the claim at hand: at all points in future, we will need
7.
by mrlongroots 3mo ago
> With alzheimer's an autopsy can tell for sure but that's not much help for a patient. Ok let us unpack this statement. For your point to hold, I would have to be saying "all kinds of practical diagnostics are invented no
8.
by mrlongroots 3mo ago
Pyschiatry gets complicated because the failures are not mechanical. Even if you could image every single neuron in a person's head we do not have a very good way to define an algorithm for these issues. I do not have a good answer for
9.
by mrlongroots 3mo ago
> I think „the diagnosis” is over simplification and lots of professionals would disagree that there’s always a single one. "The Diagnosis" does not mean "one root cause". Situation: my car has some unexplained vibrat
10.
by mrlongroots 3mo ago
> There is no guarantee that the LLM will help you converge on anything. Absolutely. The guarantee does not come from the LLM. The LLM is a simply an improved version of Google Search. The guarantee can only come from a systemic applica
11.
by mrlongroots 3mo ago
Yeah I think the OP is muddling the point by conflating "physician's version of the diagnosis" with "The Diagnosis". There is absolutely one "The Diagnosis". Human body is a machine, albeit a very complex
12.
by mrlongroots 3mo ago
Maybe I am missing something but I just find this wrong. Everything is a puzzle: there is one "Truth" or one diagnosis. You (a smart human) should be able to converge on it by cross-examining your LLMs. By themselves, they have no
13.
by mrlongroots 4mo ago
> Yes, they don't realize it or lie to themselves because ~50% dropout. I think there's some misinterpretation here. Not staying on in academia after PhD (common/modal) is not the same as not getting to complete a PhD (rar
14.
by mrlongroots 4mo ago
As someone who graduated with a 7.5 year long PhD last month, I feel like PhD stipends are not a major problem. Like I got $40K in a low CoL area, but accounting for tuition and overheads I cost my advisor closer to $150K/year. Now why
15.
by mrlongroots 5mo ago
That training is compute-bound and inference is memory-bound is well-known, but I don't think Nvidia deployments typically specialize for one vs the other. One reason is that most clouds/neoclouds don't own workloads, and wan
16.
by mrlongroots 6mo ago
MapReduce is nice but it doesn't, by itself, help you reason about pushdowns for one. Parquet, for example, can pushdown select/project/filter, and that's lost if you have MapReduce. And a reduce is just a shuffle + map,
17.
by mrlongroots 6mo ago
Algebras are also nice for implementations. If you can decompose a domain into a few algebraic primitives you can write nice SIMD/CUDA kernels for those primitives. To your point, I wonder if the 73 distinct transforms were just differ
18.
by mrlongroots 7mo ago
Yes, GPT5-series thinking models are extremely pedantic and tedious. Any conversation with them is derailed because they start nitpicking something random. But Codex/5.2 was substantially more effective than Claude at debugging complex
19.
by mrlongroots 8mo ago
While Arrow is amazing, it is only the C Data Interface that can be FFI'ed, which is pretty low level. If you have something higher-level like a table or a vector of recordbatches, you have to write quite a bit of FFI glue yourself. It
20.
by mrlongroots 9mo ago
Hyperscalers do not need to achieve parity with Nvidia. There's a (let's say) 50% headroom in terms of profit margins, and plenty of headroom in terms of the complexity custom chip efforts need to implement: they don't need t
21.
by mrlongroots 9mo ago
Yeah unfortunately no amount of manoeuvering is a substitute for a kill chain where a distributed web of sensors and relays and weapon carriers can result in an AAM being dispatched from any direction at lightspeed.
22.
by mrlongroots 10mo ago
The appropriate comparison point for aggregate cluster storage bandwidth would be its bisection bandwidth. (I do HPC, IIRC ANL Aurora is < 1PB/s DAOS and 20 PB/s bisection).
23.
by mrlongroots 10mo ago
I think I'm talking about cluster-scale network bisection bandwidth vs attached storage bandwidth. With replication/erasure coding overhead and the economics, the order of magnitude difference still prevails. I think your point is
24.
by mrlongroots 10mo ago
Yep I think the value of the experiment is not clear. You want to use Spark for a large dataset with multiple stages. In this case, their I/O bandwidth is 1GB/s from S3. CPU memory bandwidth is 100-200GB/s for a multi-stage j
25.
by mrlongroots 10mo ago
> LDL-C is much much cheaper to measure. ApoB costs 36x times as much, so Insurance Companies don't like to pay for it Unfortunately American retail prices might as well be generated by a PRNG, and do not mean much. On Ulta, a basic
26.
by mrlongroots 10mo ago
Maybe 80-90% of people should take doctors at face value, but it is easy and only getting easier to at least access the knowledge to better advocate for your own healthcare (thanks to LLMs), with better outcomes. Of course, this requires do
27.
by mrlongroots 10mo ago
Same, I don't understand the complaints against modern C++. A lambda, used for things like comparators etc, is much simpler than structs with operators overloaded defined elsewhere. My only complaint is the verbosity, things like `std:
28.
by mrlongroots 11mo ago
> Getting 200 Gb/s of reliable in-order bytestream per core over a unreliable, out-of-order packet-switched network using standard ethernet is not very hard with proper protocol design. You also suggested that this can be done usin
29.
by mrlongroots 11mo ago
> Is this just a cost efficiency thing? It's not entirely, but even that would be a justifiable reason. Tail behavior of all sorts matters a lot, sophisticated congestion control and load-balancing matters a lot. ML training is all
30.
by mrlongroots 11mo ago
Yes, unfortunately even the best intentioned individuals have very limited ability to make meaningful carbon-minimizing decisions. Carbon tax is such a sensible solution!

More ›