devstein

Hey HN, which embedding models are people using? There has been so much development around foundational LLMs, but haven't seen much news about embedding models.

Hey HN - it's time for our annual pen test.

We typically work with a 3rd party security firm each year, but now I'm wondering if there are more efficient and more AI-native options that people are using? It seems like a clear use-case for agents. Does anyone have experience working with a product/vendor in the space?

Since June of this year, we've noticed a spike in latency for DB queries routed through Supabase's Supavisor.

Direct DB queries will take a tens of milliseconds while queries through Supavisor will take a minimum of 200ms and up to 5s.

This is phenomena independent of the underlying query even `SELECT 1` through Supavisor can take up to 5s.

Has anyone else experience this before? Is this a known failure mode of Supabase's Supavisor?

I personally find writing internal documentation very monotonous even though I know it's important.

I'm curious if you were told you could have decent internal documentation automatically generated for you that were good enough, but not competitive w/ human maintained docs.

Would you still take the effort to write your own documentation?

Hi HN, we use Jinja templates to manage our prompts at Dosu (https://dosu.dev/).

We use MyPy to ensure type-safety across our entire codebase, but we do not have coverage on Jinja template variables.

We want to know if someone either

(a) forgets to pass a required template variable to a Jinja template (b) passes a name of a template variable that doesn't exist

We already get these errors at runtime with Jinja, but want to know if there is a way to statically check for this.

I'm curious if anyone has run into this before or has suggestions for type-safe templating in Python.

Thanks!