e1g

9 ms·

31.
by e1g 7mo ago
Creator here - didn't expect this to go public so soon. A few notes: 1. I built this because I like my agents to be local. Not in a container, not in a remote server, but running on my finely-tuned machine. This helps me run all agents
32.
by e1g 7mo ago
Claude Code recently downgraded the default thinking level to “medium”, so it’s worth checking your settings.
33.
by e1g 7mo ago
FWIW, both OpenAI and Anthropic have a toggle to do a “Temporary/Incognito Chat” that does not use or update memory. I too wish this was the default, and then you could opt in at the end of the chat to save some long term aspects into
34.
by e1g 7mo ago
+1 for the idea. Enforcing hard boundaries between is surprisingly helpful for AIs to reason about how to structure their changes. We recently rolled out our own static analysis using oxc-parser and oxc-resolver, and it runs surprisingly fa
35.
by e1g 7mo ago
I test a bunch of these every day too, so I made a local sandbox to jail all TUI clunkers to $CWD and run all of them in —-yolo mode https://agent-safehouse.dev/
36.
by e1g 8mo ago
I like this! I built something similar for sandboxing CLI agents, and in the repo have a collection of minimal profiles for sandbox-exec to use - https://agent-safehouse.dev/
37.
by e1g 8mo ago
I haven’t found an easy way, but I have a working theory - sandbox-exec cannot filter based on domain names, but it can restrict outbound network connections to a specific IP/port (and drop the rest). If I can run a proxy on localhost:
38.
by e1g 8mo ago
Claude Code sandboxing uses the same basic OS primitive but grants read access to the entire filesystem and includes escape hatches (some commands bypass sandboxing). Also, I wanted something solid I can use to limit every agent (OpenCode
39.
by e1g 8mo ago
On a Mac, I use built-in sandboxing to jail Claude (and every other agent) to $CWD so it doesn’t read/write anything it shouldn’t, doesn’t leak env, etc. This is done by dynamically generating access policies and I open sourced this at
40.
by e1g 8mo ago
Thank you - you inspired me to open-source this work properly -> https://eugene1g.github.io/agent-safehouse/
41.
by e1g 8mo ago
Here's my ruleset https://gist.github.com/eugene1g/ad3ff9783396e2cf35354689cc6... My goal is to prevent Claude from blowing up my computer by erasing things it shouldn't touch. So the philosophy of my sanboxi
42.
by e1g 8mo ago
You can use an alias, which takes priority over $PATH. e.g. I have this in .zhsrc to override the "claude" executable to run it in the OS sandbox: alias claude="sandbox-exec -f ~/agents-jail.sb ~/.local/b
43.
by e1g 8mo ago
I have both Codex Monitor and this new Codex app open side by side right now; aside from the theme, I struggle to tell them apart. Antigravity's Agent Manager is obviously different, but these two are twins.
44.
by e1g 8mo ago
Wow, this is nearly an exact copy of Codex Monitor[1]: voice mode, project + threads/agents, git panel, PR button, terminal drawer, IDE integrations, local/worktree/cloud edits, archiving threads, etc. [1] https://
45.
by e1g 9mo ago
Hmmm, no, I don't know this for sure. In my testing, the /compact endpoint seems to work almost too well for large/complex conversations, and it feels like it cannot contain the entire latent space, so I assumed it keeps po
46.
by e1g 9mo ago
"encrypted_content" is just a poorly worded variable name that indicates the content of that "item" should be treated as an opaque foreign key. No actual encryption (in the cryptographic sense) is involved.
47.
by e1g 9mo ago
Yes, Codex compaction is in the latent space (as confirmed in the article): > the Responses API has evolved to support a special /responses/compact endpoint [...] it returns an opaque encrypted_content item that preserves the
48.
by e1g 1y ago
+1 from another happy Whispr Flow power user. I tried 4-5 similar apps and even built one with Assembly AI, but Whispr is a significant upgrade above the rest for correctly recognizing my accent and jargon. Having the custom vocabulary help
49.
by e1g 1y ago
Do you happen to have a link with a more nuanced technical analysis of that (emergent) behavior? I’ve read only the pop-news version of that “escaping” story.
50.
by e1g 1y ago
Knowing who is behind this campaign, 90% chance the extra white space, the graffiti, and this article were all commissioned by them intentionally. Luckily this did not translate to sales, or we’d have another wave of Cluely BS copycats.
51.
by e1g 1y ago
It takes no imagination or insight to see reasons why something wouldn’t work. It’s the default mental pathway for every risk-averse beast. Skepticism is not born out of contentment and abundance but out of self-preservation. It’s not cor
52.
by e1g 1y ago
The cynic always sounds smart and remains poor.
53.
by e1g 1y ago
To remove resulting notifications, see instructions here https://github.com/orgs/community/discussions/174283#discuss... These spam repositories have been deleted, but I still had lingering notifications stuc
54.
by e1g 1y ago
Extremely slow for me - takes minutes to get anything done. Regular GPT5 was much faster. Hoping it’s mostly due to the launch day.
55.
by e1g 1y ago
> clean/erase/undo/soft-delete/hard-delete mistakes[...] make the change tracking capable of time travel itself [...] Transitioning to an EAV I just finished building out all of that + more (e.g., data lineage, multi-
56.
by e1g 1y ago
For “Organizational Use”: the people who designed and own Myers-Briggs say this is a bad idea - > The MBTI® assessment is not intended for use in selection of job candidates, nor for making internal decisions regarding job placement, sel
57.
by e1g 1y ago
Excellent example. And at that scale, you are generating 100TB/s in UUIDs so if you need to store them, you have much bigger problems than collisions.
58.
by e1g 1y ago
You need to be generating >100M of them within the same millisecond before even remembering that collisions can theoretically happen.
59.
by e1g 1y ago
Another strategy is to pass your values as an array param (e.g., text[] or int[] etc) - PG is perfectly happy to handle those. Using ANY() is marginally slower than IN(), but you have a single param with many IDs inside it. Maybe their ORM
60.
by e1g 1y ago
At least in Manhattan, no one will give you a contract for more than a year - they will want to renegotiate in a year and demand is high enough to not worry about having no tenants. In Europe multi-year contracts are somewhat typical, but n

More ›