I'm working on https://draftin.com/. If I can be helpful at all with anything, please hit me up on Twitter (@natekontny) or email me anytime (nate.kontny+hn@gmail.com).
So ideally, these environments are just in the cloud. But Claude Code web, is so "environment lite" that it really isn't a substitute. Or are you all having Claude Code web install a bunch of dependancies and even API keys to do things there?
I've got a system of VMs that get deployed now that I'm encouraging folks use, because Claude can just use those cloud VMs directly to work on code, run tests and things without a full slow CI workflow. But it seems so fragile.
What are you all using to power up employees at your company with dev environments something like Claude can use without those employees actually having local dev environments?
I have a folder of static HTML and CSS. Think: docs, reports, AI prototype, whatever. It’s just a tiny html microsite. Nothing dynamic.
And I want to share it with some people. Easily.
But now the hard stuff: SSO gets involved. Github Pages is public unless you are using an Enterprise account and just with engineers? Google Sites? I guess the old days that would work like you would hope: and let you upload sites. Today, it's more like a website builder? Can't upload raw HTML sites there.
My wife works at 1pass :) So my brain is in 1pass a lot. Would you use some kind of additional app that is basically “Vault Pages”? I don’t think you can have any kind of tree inside a vault. So things have to be pretty flat or deflate something like a zip file with the folder structure in it. You’d be able to share urls like (as long as you all agree to use this additional .app, and I’m assuming you're a Mac shop):
`vaultpages://engineering-handbook/deploy-checklist`
Would you us this? Or you already have a great private microsite solution that isn’t messing with additional SSO, VPNs, IAM policies, etc.
fintuneing seems to be out of fashion (if it were really ever in fashion), but I still see folks like Karpathy mention reaching for it as a tool.
But is anyone in any business capacity on here doing that? Are you finetuning any remote LLM or something self-hosted? What for?
I’m just curious where the line is of “oh this is better encoded in the models weights rather than in RAG/thinking over context stuff it needs to figure out.