Who is the intended crowd for this? I couldn’t make out If it’s used to centralize control over mcp servers on multiple user machines, or just locally for one
Unless a new architecture for LLMs emerge that has an inherit way of separating context from safe user data and external unsafe data every interaction is susceptible to PI.
My question here is why would the bank agent need to look at the transaction data that is exposed to the outside? Apart from guardrails etc. high risk scenarios where agents are involved should aim to exclude external untrusted data whenever possible
It's getting harder to review the plans with Fable. So do we plan with Opus and let Fable implement or just start trusting blindly. Feels to me that this is another shift in how we operate these systems.
The security layer needs to parse the full agent activity with the context. It watches everything, but only interrupts the human when it matters.
Commands that can run arbitrary code need to be treated differently and can't get escalated in this opaque way.
A large part of the solution should be to drastically reduce the amount of permission approval prompts a user gets. This ensures the ones he does get are evaluated with the same concentration a manager gives a new hire's most consequential decisions.
Most importantly, because we ask him rarely, when we do he feels the accountability. The yes is his.
Most of the internet is clickbait for a long while now.
No one would read a title like "MCP and CLI can both be usefull in certain scenarios. Ask your AI and he'll tell ya" :)
Still early days, but if it's better than 4.6 and doesn't have 4.7's issues then I'm a happy camper. When 4.7 dropped I switched back to 4.6 the same day. I have a good feeling about this one though. You know what they say "Honesty is the best policy"
Cyber security is always a cat and mouse game. Always was and always will, just with AI now as another tool in the arsenal of defenders and attackers. I think that we might see a shift though where the winner will be the one with most compute advantage
The common definition of zero day vulnerability is:
A zero-day vulnerability is a software security flaw that is unknown to the vendor or developers responsible for fixing it. Because they have had "zero days" to address it, no patch or fix exists at the time the flaw becomes known to attackers.
I also feel like we still need to steer Claude. It doesn't always help to have stuff in the CLAUDE.md (even when it's lean). I have a lot of cases where I still need to remind the agent to do something even if it's routine.
To me I think that connects with working longer on the planning and specs. It requires reading and re-reading, but when that's done, implementation is usually much cleaner and adheres to your standards
I think that in general, people need to understand that they need to invest most of their time in the planning phase. High level plan, then spec are the baseline imo