dhblumenfeld1

7 ms·

1.
by dhblumenfeld1 6mo ago
Every developer using Claude Code, Codex or Cursor has the same problem: your agent runs your code but it can also read your .env files. API keys, database credentials, anything on disk is visible to the agent. I built keypo-signer, an open
2.
Show HN: Keypo – Secure Enclave encrypted secrets for AI coding agents(github.com)
5 points by dhblumenfeld1 6mo ago|1 comments
3.
by dhblumenfeld1 6mo ago
This is dope. Love local-first tools. Would be dope to have an iOS/macOS version that I can just buy through the app store (and is hence synced across my devices).
4.
by dhblumenfeld1 6mo ago
Increasingly convinced that most AI doomerism is coming from the VC / Startup ecosystem because the industry, or game, they've been playing for the last 20 years is undergoing the most disruption.
5.
by dhblumenfeld1 6mo ago
wouldn't signal fall under this category (same entity control the client and server in between) but they have no way of peaking inside any envelopes?
6.
by dhblumenfeld1 6mo ago
I recently dealt with something similar with a CDN I used for a project 3 years ago. They kept charging me $0.01/month after I got all of my content off of their servers. Took months to resolve, luckily was pennies so it wasn't a
7.
by dhblumenfeld1 6mo ago
Have you found that using a frontier model for planning and small local model for writing code to be a solid workflow? Been wanting to experiment with relying less on Claude Code/Codex and more on local models.
8.
by dhblumenfeld1 6mo ago
this is great. The environment variable control is particularly interesting to me. I've been working on a related problem but different approach: encrypting/decrypting secrets using the Mac's secure enclave ( https:/&#x2
9.
by dhblumenfeld1 6mo ago
V interesting. Does the passport have history of what the agent has done in the past and if so, how are you able to verify that work? Specifically thinking about agents that may work across different systems. Might be worth checking this ou