This looks very cool! My concern is just that it's a lot of things bundled in one and I kind of have to trust you for all of them. I would prefer something that puts together a stack of better known components, like a docker compose of agent-sandbox and tailscale or something equivalent, etc ... ideally with each of those swappable.
Thanks. Just noting that those docs say the cache duration is 5 min and not 1 hour as stated in sibling comment:
> By default, the cache has a 5-minute lifetime. The cache is refreshed for no additional cost each time the cached content is used.
>
> If you find that 5 minutes is too short, Anthropic also offers a 1-hour cache duration at additional cost.
Apparently Anthropic downgraded cache TTL to 5 min without telling anyone. My biggest issue with the recent issues with Claude Code is the lack transparency, although it looks like even Boris doesn't know about one:
https://news.ycombinator.com/item?id=47736476
Does anyone ever worry about prompt injection attacks against yourself in these?
When I was into hypnosis and NLP between one and two decades ago, I used to worry about what the instructions were once I was hypnotized. I lacked the terminology then but there days we would call these prompt injections, just against the human brain.
I guess social engineering is another form, although that's probably more akin to a CSRF or flawed auth logic exploit.
It is still orders of magnitude away from breaking RSA 2048 even under the most optimistic assumptions. And qubits double waaay slower than transistors so far.
Still requires thousands of logical qubits, which would correspond to millions of physical qubits. And this machine isn't even fully there for the physical qubit part. It's like the first step to physical qubits.
Upload queues are better than cooldowns
I almost didn't read it because I wasn't interested in a rant. This is a genuinely good idea though so I'm glad I did.
Alas, I did click through so perhaps the title is more effective than my sentiments.
reply