joatmon-snoo

6 ms·

31.
by joatmon-snoo 2y ago
bidi streaming screws with a whole bunch of assumptions you rely on in usual fault-tolerant software: - there are multiple ways to retry - you can retry establishing the connection (e.g. say DNS resolution fails for a 30s window) _or_ you c
32.
by joatmon-snoo 2y ago
...how is _this_ the insight that you come away from this post with? This post is a commentary on product quality issues, the underlying cost models (both goods and services), and the interplay with American culture. There's like 20+ c
33.
by joatmon-snoo 2y ago
> This seems like exactly the sort of question the market will quickly decide over the next couple of years and not worth arguing over. Discussions like this are _how_ the market decides whether or not this achievement is real or not.
34.
by joatmon-snoo 2y ago
If Google adopts an existing OSS technology, it usually takes the form of Google contributors joining a core team for the OSS in question. The OSS community generally isn't a fan of single companies taking _over_ a project and generall
35.
by joatmon-snoo 2y ago
Different DBs implement locks differently. Postgres allows obtaining advisory locks at either the session _or_ transaction level. If it's session-level, then you have, ergo, a connection-level lock. https://www.postgresql.or
36.
by joatmon-snoo 2y ago
This is what lockfiles are for.
37.
by joatmon-snoo 2y ago
Good file watching that provides flexible primitives absolutely requires: - ok, a single ext4 file inode changes, and its filename matches my hardcoded string - oh, you don’t want to match against just changes to “package.json” but you want
38.
ts-blank-space is a fast type-stripping compiler(bloomberg.github.io)
48 points by joatmon-snoo 2y ago|46 comments
39.
by joatmon-snoo 2y ago
It's much easier to say "I'm going to make it impossible for us to have to worry about the Australian government filing a lawsuit against $my-state-agency, because legal said so" than "Well, if we allow Australian I
40.
by joatmon-snoo 2y ago
Assuming you're the founder, this is the type of BS comment that makes the rest of us hate AI founders. It's vacuous, makes vague claims that don't leave room for proof/disproof, and doesn't offer any reason that it
41.
by joatmon-snoo 2y ago
Yep, it’s Chrome’s incognito mode. The lawsuit describes it as “private browsing”.
42.
by joatmon-snoo 2y ago
Seriously! I actually can’t think of any openly documented security incident with such impressive remediation timelines. There’s a lot that has to go into fixing things on such a tight timeline too: - oncall-level alerting for your security
43.
by joatmon-snoo 2y ago
Wait, so you’re blocking on a Raft round-trip to make forward progress? That’s the correct decision wrt durability, but… I’m now completely lost as to why you believe this was a good idea over using something like MySQL/Postgres/A
44.
by joatmon-snoo 2y ago
This is cool! I’m always excited by people trying simpler things, as a big fan of using Boring Technology. But I have some bad news: you haven’t built a system without a database, you’ve just built your own database without transactions and
45.
by joatmon-snoo 2y ago
Post hoc ergo propter hoc. Sure, Apple offshoring their semiconductor mfg compounded the US' lack of domestic semiconductor mfg capacity/expertise. Apple also did that because it is a _business_, and has not always been the behemo
46.
by joatmon-snoo 2y ago
Were you at the company before they brought things in-house, or were you brought on for that job explicitly? I haven't been in this situation before, but I imagine one of the trickier challenges in a role like what OP is describing is
47.
by joatmon-snoo 2y ago
CodeMirror's been fantastic for us: as folks building a programming language, we wanted a way to create a playground for people to try it out (much like what Goland and Rust did), and we found CodeMirror super helpful for putting toget
48.
by joatmon-snoo 2y ago
I read a lot of Stonebraker's work as an undergrad and spent 3+ years at Google working on databases. I like to think that I know some things about databases, even if I wasn't around when XML databases were a thing :)
49.
by joatmon-snoo 2y ago
Er- kinda? Maybe? Not really? constexpr does not mean that you can evaluate arbitrary C++ code at compile time. It allows you to evaluate a _very specific subset_ of C++ at compile time that is not at all easy to wrap your head around: look
50.
by joatmon-snoo 2y ago
It's really easy to read this as "shitty interviewer runs off good candidate". It's also easy to read this as "interviewer hand-held a candidate through a problem".
51.
by joatmon-snoo 2y ago
I don't know how I feel about this paper: on the one hand, I agree with the sentiment that the relational data model is the natural end state if you keep adding features to a data system (and it perfectly captures my sentiment about ve
52.
Building a new programming language in 2024, pt. 1(boundaryml.com)
2 points by joatmon-snoo 2y ago|0 comments
53.
by joatmon-snoo 2y ago
Hmm, that's an interesting way of thinking about it. The way I see it, I trust XML less, because the sparser representation gives it more room to make a mistake: if you think of every token as an opportunity to be correct or wrong, the
54.
by joatmon-snoo 2y ago
Author here! I very deliberately avoided making that claim; the table is actually very unsorted right now, in no small part because all the solutions in the space satisfy a very different set of usage criteria - some folks use Python, other
55.
by joatmon-snoo 2y ago
Author here- yes, reprompting can work well enough if the latency hit is acceptable to you. If you’re driving user-facing interactions with LLMs, though, and you’re already dealing with >1min latency on the first call (as many of our cur
56.
by joatmon-snoo 2y ago
Author here- besides hellovai’s point about the performance bottleneck, it’s a really tricky semantic problem! LLMs today are really good at producing output that satisfies the very vague metric of “this looks good to a human” but aren’t ne
57.
Building RAG in Ruby, using BAML, with streaming(boundaryml.com)
1 points by joatmon-snoo 2y ago|0 comments
58.
by joatmon-snoo 2y ago
Not at all! Out of curiosity- do those orgs not find the loss of generality that comes from custom models to be an issue? e.g. vs using Llama or Mistral or some other open model?
59.
by joatmon-snoo 2y ago
Constraining output means you don’t get to use ChatGPT or Claude though, and now you have to run your own stuff. Maybe for some folks that’s OK, but really annoying for others.
60.
by joatmon-snoo 2y ago
We actually built an error-tolerant JSON parser to handle this. Our customers were reporting exactly the same issue- trying a bunch of different techniques to get more usefully structured data out. You can check it out over at https:/

More ›