Just make sure you never let claude make do any calculation without using a programming language. It can be somewhat trusted to make less mistakes than a human when it comes to filling in the right numbers but it's still not great at math.
Also did create a cli. I had to fork the project and removed one unused import which allowed me to remove a lot of unused ml libraries: https://github.com/Mic92/puss-say
So usually MCP tool calls a sequential and therefore waste a lot of tokens. There is some research from Antrophic (I think there was also some blog post from cloudflare) on how code sandboxes are actually a more efficient interface for llm agents because they are really good at writing code and combining multiple "calls" into one piece of code. Another data point is that code is more deterministic and reliable so you reduce the hallucination of llms.
Do expose the accessibility tree of a website to llms? What do you do with websites that lack that? Some agents I saw use screenshots, but that seems also kind of wasteful. Something in-between would be interesting.
I am sure they measured the difference but i am wondering why reading screenshots + coordinates is more efficient than selecting aria labels? https://github.com/Mic92/mics-skills/blob/main/skills/browse.... the JavaScript snippets should at least more reusable if you want semi-automate websites with memory files
Well this compiler was written to build Linux as a proof of concept. You don't need a libc for building the kernel. Was it claimed anywhere that it is a fully compliant C compiler?
I skipped over the first few ones and haven't seen critical ones. The hardcoded oauth client secrets is basically present in any open-source or commercial app that is distributed to end users. It doesn't break the security of end users. It mainly allows other apps to impersonate this app, i.e. present itself as clawdbot, which is a moot point given anyone can just change /inject code into it.
I remember that a llm agents often store those in clear text files (I think claude-code beeing one of them). Many of the CLIs I use have a better secret hygiene than that i.e. allow passwords commands or use secret apis.
Doesn't make any difference in practice. The debug info is never mapped into memory by the loader. This only matters if you want to store the two separate i.e lazy load debug symbols if needed.
I can second that. If there are multiple commits: https://github.com/tummychow/git-absorb is handy to add formatting changes into the right commit after commits already happened.
We work on GPU accelerated VMs, so that in future we can also bring NixOS + VPNs to desktops/end users to machines that don't run NixOS. We will use it as an application runtime where can control the whole stack. Just now we are mostly focused on managing distributed NixOS machines. The VPN helps to provide services on any kind computer, even if not running in a datacenter. You can read the description here for context: https://docs.clan.lol/
I haven't seen anything about requirements for gpg. Also the ux of it is not so great, so it's easy to just not have a signature without causing too much suspicion. Would be a much easier attack than what Jian Tan pulled off. Just wait for some contributor to go on holiday and send a malicious v2 patch. There are so many patches in the linux kernel processed that no one wouldn't notice.