tacoooooooo

8 ms·

31.
by tacoooooooo 9mo ago
This is an interesting read, and while I support being nice to every_thing_ in principle. Most of the research into this actually shows that being mean yeilds better results
32.
by tacoooooooo 9mo ago
This looks pretty cool. I keep seeing people (an am myself) using claude code for more an more _non-dev_ work. Managing different aspects of life, work, etc. Anthropic has built the best harness right now. Building out the UI makes sense t
33.
Traditional NLP is not dead(alex-jacobs.com)
1 points by tacoooooooo 9mo ago|0 comments
34.
by tacoooooooo 9mo ago
fizsh sounds really cool, but the last commit was 7+ years ago. do you run into any issues? https://github.com/zsh-users/fizsh
35.
by tacoooooooo 9mo ago
not sure there are any models yet that you can get the quality out you need to do this and run on your mbp
36.
by tacoooooooo 9mo ago
This is a wildly out of touch thing to say
37.
Can't Beat BERT–comparing small LLMs and fine-tuned encoders on classification(alex-jacobs.com)
3 points by tacoooooooo 9mo ago|1 comments
38.
by tacoooooooo 9mo ago
correct afaik :( https://github.com/timescale/pgvectorscale/issues/113
39.
by tacoooooooo 9mo ago
the main issue with pgvectorscale is that it's not available in RDS :(
40.
by tacoooooooo 9mo ago
This probably doesn't count as an "app" in terms of what you're looking for, but was a fun little project https://alexjacobs08.github.io/lobsters-graph/ (i built this in search of a lobste.rs invite
41.
by tacoooooooo 9mo ago
This looks awesome. ai-sdk is an excellent library. excited to see it proliferating
42.
Show HN: The Lobste.rs invitation tree, visualized(alexjacobs08.github.io)
5 points by tacoooooooo 9mo ago|1 comments
43.
by tacoooooooo 10mo ago
Standard benchmarks (like BEIR/MS MARCO) are great, but they are likely already in distribution for foundation models training sets, and crucially, they lack the complex, structured metadata needed to test real-world filtering scenario
44.
Show HN: Generate a 1M-document RAG eval dataset from a single prompt(alexjacobs08.github.io)
2 points by tacoooooooo 10mo ago|1 comments
45.
Show HN: Dataset Factory – Generate RAG evaluation datasets from a text prompt(alexjacobs08.github.io)
1 points by tacoooooooo 11mo ago|0 comments
46.
Generate RAG evaluation datasets from a single prompt (1K to docs)(alexjacobs08.github.io)
2 points by tacoooooooo 11mo ago|0 comments
47.
by tacoooooooo 11mo ago
it's an odd choice. I'd be curious why they picked that. it's not the cheapest, most expensive, best, or worst. It does have a relatively large context window, and ime is very good at format adherence
48.
by tacoooooooo 11mo ago
well said! we demo'd milvus (or zilliz i should say,) and while we didn't ultimately go with it--it seems like a great option
49.
by tacoooooooo 11mo ago
Author is a human :). Performance and semantic accuracy are both important. The point about pre-filtering _youre still searching millions of vectors_ is important because once you apply a filter you can no longer use your vector index. And
50.
by tacoooooooo 11mo ago
pgvectorscale is not available in RDS so this wasnt a great solution for us! but it does likely solve many of the problems with vanilla pgvector (what this post was about)
51.
by tacoooooooo 11mo ago
this is at the expense of precision/recall though isn't it?
52.
by tacoooooooo 11mo ago
i discuss that specifically! > The problem is that index builds are memory-intensive operations, and Postgres doesn’t have a great way to throttle them. You’re essentially asking your production database to allocate multiple (possibly do
53.
by tacoooooooo 11mo ago
guess it depends on your scale? for some, 10+ GB of RAM being consumed on an index build is > 25% of the DB's RAM. apply that same proportion to your setup and maybe it'll make more sense
54.
by tacoooooooo 11mo ago
for sure people are running pgvector in prd! i was more pointing at every tutorial iterative scans are more of a bandaid for filtering than a solution. you will still run into issues with highly restrictive filters. you still need to unders
55.
by tacoooooooo 11mo ago
> Anyways, all this vector stuff is going to fade away as context windows get larger (already started over the past 8 months or so). We're searching across millions of documents, so i doubt it
56.
by tacoooooooo 11mo ago
I mention this towards the end of the post. it looks like a good solution, but it's not available on RDS
57.
by tacoooooooo 11mo ago
how are all of the mentioned issues resolved?
58.
by tacoooooooo 11mo ago
some fair points points on the specifics. > maintenance_work_mem sure, but the knob existing doesn't solve the operational challenge of safely allocating GBs of RAM on prod for hours-long index builds. > REINDEX CONCURRENTLY this
59.
by tacoooooooo 11mo ago
We actually looked into vectorchord--it looks really cool, but it's not supported by RDS so it is an additional service for us to add anyways.
60.
The Case Against PGVector(alex-jacobs.com)
381 points by tacoooooooo 11mo ago|137 comments

More ›