adamgold7
Joined 3 years ago
Karma
9
Submissions
24
Member since
May 25
Recent submissions
Comment · 2 months ago
appreciate it!
Comment · 3 months ago
what do you mean?
Comment · 3 months ago
what were you expecting?
Comment · 3 months ago
that's a very interesting take! i think the market is leaning towards software factory now to describe autonomous agents in R&D and that might very well not reprensent the truth
Comment · 4 months ago
love this. we are actually looking at reward hacking from a cyber security perspective - refreshing (unless you're from Israel).
Any collaborators that want to join us?
Comment · 6 months ago
we need better RL
Comment · 8 months ago
We've been working on exactly this at Islo. Zero-setup microVM sandboxes with isolated networking by default, plus an approval workflow layer so agents can request capabilities and humans approve/deny in real-time.
The credential problem is handled through proxy middleware - agents never see real tokens, requests get routed through policy-checked proxies that inject credentials only for approved operations.
Happy to share more: https://islo.dev
Comment · 8 months ago
The paper nails it - we're giving agents capabilities before we have infra to contain them. The answer isn't better prompts. It's treating agent execution like untrusted code: sandboxed VMs, explicit capability grants, network isolation, approval workflows for production actions.
Comment · 8 months ago
Prompt guardrails are theater - they work until they don't. We ended up building sandboxed execution for each agent action. Agent proposes what it wants to do, but execution happens in an isolated microVM with explicit capability boundaries. Database writes require a separate approval step architecturally separate from the LLM context.
Worth looking at islo.dev if you want the sandboxing piece without building it yourself.