Really excited about the 2 projects that popped up. Claw Code is porting from Python to Rust which will be interesting. And OpenClaude will use an existing codex login to operate on Gpt5.4 which I’m looking forward to playing with during the long weekend.

Using codex on Gpt5.4 with x-high reasoning got me similar results to ClaudeCode on Opus. So I’ll be interested to see the results

1. Fixed credential-exfiltration risk in /api/proxy/image: Previously the endpoint could:

• accept arbitrary auth_id
• load stored API keys
• forward them to attacker-controlled URLs

2. Enforced outbound host allowlist globally Previously:

• allowlist existed
• but outgoingFetch() didn’t enforce it
• plugins/engines could bypass it

3. Fixed extension store path traversal Previously a malicious store manifest could:

• inject … paths
• escape install directories
• reference arbitrary files

4. Hardened proxy IP trust Previously:

• rate limiting trusted any X-Forwarded-For header
• clients could spoof their IP

5. Fixed inconsistent settings authentication Previously:

• settings UI stored an auth token
• but the settings modal didn’t send it when saving

6. Implemented Improved proxy deployment support

• Added proxy-aware behavior:
• DEGOOG_PUBLIC_BASE_URL for canonical URLs
• secure cookie handling when X-Forwarded-Proto=https

Additional Improvements:

• suggestion fetching hardened
• DuckDuckGo suggestion parsing fixed
• unified outbound request handling
• install state guard properly cleaned up

Made some other changes for my specific deployment. Very happy with your work so far. Thanks so much

ZFS will use memory to cache. Right now my unraid server is using 16gigs, most of that is being used by ZFS to cache files. 4gigs is all ~20 containers are using.

It’s not, the second I cloned it and gave codex access it found a whole whack of privacy issues. This was 100% human coded

Well I’ve done most of those except for using an immutable distro, and quite honestly have zero interest in running one.

Just this past year I’ve setup a media server using a bunch of servers I nabbed from work, did unraid because I’ve been toying with the unraid idea for years. Learned and fell in love with Docker. I avoided it like the plague for the better part of a decade. Now I run damn near everything in docker containers. Started using home assistant quite recently actually, my wife got annoyed with the heavily sanitized results she was getting with Gemini for Google Home. So I setup M5Stack ATOM Lite ESP32’s around the house acting as microphones that I then use with home assistant to ferry the query to the Anthropic API and the result to the nearest Google Home device. It’s a work in progress, but happy with the results

It’s called a thorn, and as for why: I’d really like to know as well, considering it fell out of use in the 1500’s

I, as a sane person enjoyed watching the game being 100%'d by someone else.

That explains alot. I have both PyCharm and RustRover open as I steal convert stuff from a project I found. Anywho I was typing in discord and I was typing faster than it rendered and I thought that was strange