rane

"Vibecoded" is not an accurate description of what actually happened. LLMs are extremely good at porting code from one language to another while preserving semantics. Which is why how everything turned out so well was not very surprising.

It's extremely difficult to have a robust test suite for a TUI app like Claude Code. Hence regressions keep showing up in every update.

Absolutely can't wait to try this for language practice. The advanced voice mode is great but ultimately just doesn't work that well and doesn't have the feel of a natural conversation.

Try https://github.com/raine/claude-code-proxy

I canceled Claude plan a few months back and have been using this. OpenAI plans are much more generous.

The excessive emojis in the README don't inspire confidence

I thought so, and then I tried Opencode and Codex and started to appreciate Claude Code a lot more. They've actually done great work with the small details.

I have 29 Claude Codes open, using 6.3 GiB RSS total

The point was to come up with a sensationalistic headline that HN eats up and post flies to the front page.

You can use other models in Claude Code

https://github.com/raine/claude-code-proxy

https://api-docs.deepseek.com/quick_start/agent_integrations...

I just made a simple script that makes it easy to switch between models.

You can just use it through Claude Code, so you get to keep the system prompt and tooling you are used to.

3rd party models are a drop-in replacement with `ANTHROPIC_BASE_URL` in Claude Code, something people seem to miss right now. And contrary to what Anthropic might like to have you think, you don't need Opus 4.7 to run the harness to get similar performance.

https://api-docs.deepseek.com/quick_start/agent_integrations...

Also, great example of why you don't take a terminal that requires login as your daily workhorse. It never ends well.

I've been using all Kimi K2.6, gpt-5.4 and now Deepseek v4 (thought not extensively yet) in Claude Code and I can say it works much better than you'd expect. It looks like the system prompt and tools are pulling a lot of weight. Maybe the current models are good enough that you don't need them to be trained for a specific harness.

FWIW, works great in Claude Code.

https://api-docs.deepseek.com/guides/coding_agents#integrate...

This might be the opposite of staying nimble as my workflows are quite tied to Claude Code specifically, however I've been experimenting with using OpenAI models in CC and it works surprisingly well.

Added support for Kimi in https://github.com/raine/claude-code-proxy and it does appear to work surprisingly well with Claude Code, although the usage limit for the entry tier doesn't seem as generous as I'd have expected.

- Way more accurate, especially with technical jargon. Try saying JSON as part of a sentence to macOS dictation and see what comes out.

- macOS dictation mutes other sounds while it's running. This is a deal-breaker for me.

Is that also why they allow to use their subscriptions in OpenClaw and 3rd party harnesses?

Why doesn't OpenAI introduce $100/mo plan? Surely many would switch in a heartbeat

For a container-based sandbox in tmux, see workmux and its sandboxing feature.

The added isolation does come with some friction though, which is kind of by design.

Conflicts are not a problem at all in my experience.

My setup is that I run `/merge`[1] , which will first have the agent rebase changes on base, and on conflicts, it's instructed to understand both sides before resolving, which helps it merge them cleanly. I haven't resolved conflicts manually in months and also haven't had any issues with agents resolving them incorrectly. A solved problem as far as I'm concerned.

[1] https://workmux.raine.dev/guide/workflows#direct-merge

In some specific work contexts, such as writing pull request descriptions, not sounding like AI is something I've given up on trying to optimize. It's simply not worth the effort for me being non-native and writing detailed PR descriptions being so arduous, and the agent already has full context anyway. Obviously any fluff or inaccuracies are aggressively weeded out but I don't care anymore about the AI voice.

I like to manage tabs and windows through tmux and it suits my workflow very well. What are you going to do now?

Moltis has a web chat UI at least. https://moltis.org/

This zeroclaw.org has to be some kind of malware.

This is the official repo https://github.com/zeroclaw-labs/zeroclaw and its website: https://zeroclawlabs.ai/

Hardware improvements are easier to quantify and progress naturally comes in incremental steps.

Software however especially from UX point of view, is more likely to be more or less ready at some point. Any improvements are marginal and subjective. What are the large UX teams at Apple going to do if not redesigns for the sake of redesigning? I wish it would happen, but it’s hard to imagine Apple shipping an annual OS release without noticeable visual changes.

$60/mo for a personal website is insane.

My similar workflow within Claude Code when it gets stuck is to have it consult Gemini. Works either through Gemini CLI or the API. Surprisingly powerful pattern because I've just found that Gemini is still ahead of Opus in architectural reasoning and figuring out difficult bugs. https://github.com/raine/consult-llm-mcp

Now what if it didn't pollute browser history

Depends on Anthropic has implemented it I guess. I haven't had it activate yet despite the prompt to install an LSP.