Entries for May 2025

@onusoz · 2025-05-29

/2025/05/29
Headless makes running these things in a sandbox much easier. Sandbox means you can give all permissions and just let it run until completion. See my efforts to do so here: https://t.co/9UUFOTwk2A

@onusoz · 2025-05-29

/2025/05/29
I was trying to figure out why @AnthropicAI Claude Code feels better than @cursor_ai with Opus + Max mode. I can’t put my finger onto it, but one of the reasons might be that it’s faster, because it doesn’t use another model to apply the diffs, which you have to wait for

@onusoz · 2025-05-27

/2025/05/27
Just an update, building this now The repo is claude-code-sandbox under TextCortex GitHub. The Proof-of-Concept is there, check the TODOs and current PRs to watch the current progress https://t.co/9UUFOTwk2A

@onusoz · 2025-05-25

/2025/05/25
I've been using Claude Code extensively since last week What I'm wondering is, since you can run Claude Code locally, why isn't there any tooling to let you run it in a sandboxed mode in local Docker containers yet? Or did I miss it? cc @AnthropicAI

@onusoz · 2025-05-20

/2025/05/20
The more I compare coding agents, Cursor, Claude Code, Codex, it becomes more apparent to me that those running locally will win over those that are running remote. The UX is just superior

@onusoz · 2025-05-16

/2025/05/16
ty is already very fast for a Python type checker. It checked around 800 files in our backend repo in around 2-3 seconds uvx ty check > /tmp/ty_log.txt 3.46s user 0.79s system 208% cpu 2.038 total