rosslazer

Author here. I've been struggling to characterize what has made me so uneasy about the Hugging Face incident. It finally clicked for me this AM: our threat modeling priors are all broken. You should assume every threat actor has nation state level capabilities, and move to an assumed breach model.

Love the idea of using page rank. Right now, there is so little adoption of vouch that I opted for the most basic thing.

Repo: https://github.com/rosslazer/vouch-book

Are there examples of the prompt or transcripts for the human testing?

Dumb question - What actually needs to be done to close the gap?

I don't know why so many people here are shitting on you guys. It is an impressive demo and launch and a wide-open market. I'm rooting for you, guys. There's an easy opening even just after the "inbox" for the security alert use case.

A candid reflection of one tech startup's journey so far.

Nice! I did something similar with GPT 3.5 and slack https://rosslazer.com/posts/fine-tuning/

One of our founders wrote a blog series on this https://www.observeinc.com/blog/why-observe-part-3-the-platf.... It all boils down to this snippet:

– Ability to ingest 100s of terabytes to petabytes of data per day, and store those for months to years.

– Ability to provide a delightfully interactive user experience, despite managing terabytes of data.

– Ability to work efficiently with semi-structured data (i.e. the messy slop you find in logs).

– Ability to efficiently execute the kinds of band-joins and window analytic functions needed for us to realize efficient queries on Resources.

– Unit economics that make sense for the Observability use-case.