TESTER BUILD · FREE WHILE TESTING

Moshi Moshi!

Coffee While You Chat ☕

You've been invited to test Moshi Moshi-san, a friendly AI that runs on a real computer in Saint Marys, Georgia — not a giant data center. Ask it questions, get help drafting posts and emails, brainstorm ideas, or just talk. While it's in testing, chatting is free — your feedback shapes what it becomes.

Powered by Cush Coffee — grab a cup, then say hello.

Moshi Moshi-san, a cheerful red telephone handset character waving hello while holding a steaming cup of coffee
Moshi Moshi-san, your host — always happy to pick up.
Moshi Moshi-san Qwen3 8B · running on Cush's laptop
Not connected
Moshi moshi! 👋 I'm Moshi Moshi-san. I'm connecting you automatically — once the line shows Connected up top, you can ask me anything — drafting, ideas, explanations, a little code help. What's on your mind?

Enter to send · Shift + Enter for a new line

Good things to try

Tap one to drop it in the chat box — a fair workout for an 8B brain:

Connect the line

Moshi Moshi-san lives on Cush's laptop. The current tunnel is already filled in — this page connects on its own when it loads. If the address ever changes, update it below and reconnect.

Changes every time the tunnel restarts — always use the freshest one Cush sent.

Preset to the LM Studio model on Cush's laptop. Ollama: qwen3:8b. LM Studio: the exact model ID shown in its model list.

Good to know

  • It's a laptop, not a cloud. If Cush's laptop is asleep, off, or the tunnel dropped, Moshi Moshi-san can't pick up. That's the #1 thing testing will shake out.
  • No live internet. It doesn't know today's news, prices, or scores — anything time-sensitive, double-check elsewhere.
  • It can be confidently wrong. An 8B model is bright but small. Verify anything that matters.
  • Free for testers. No accounts, no payment — just chat and report back.
  • This page keeps nothing. Your conversation lives only in this tab, only while it's open.

Host checklist

For Cush — getting the line live before testers arrive:

  1. Use the Qwen3-8B you already have. In LM Studio, pick the Qwen3-8B Q4_K_M already downloaded (~5 GB) — for Qwen3, plain Qwen3-8B is the instruction-tuned model (Qwen3-8B-Base is the raw base one). If replies come back as gibberish or repeated characters, first try the official Qwen/Qwen3-8B-GGUF Q4_K_M instead of the Aldaris conversion, or cross-check the same prompt in Ollama (qwen3:8b) to tell whether it's the file or LM Studio. Or in Ollama: ollama pull qwen3:8b.

  2. Start the local server. LM Studio: Developer tab → Start Server on port 1235, CORS on. Ollama: set OLLAMA_ORIGINS=*, restart Ollama, confirm it's listening on 11434.

  3. Open the tunnel. cloudflared tunnel --url http://localhost:1235 (LM Studio) or :11434 (Ollama). Copy the https://…trycloudflare.com address it prints.

  4. Share the address. This page is prefilled with the current tunnel, LM Studio, and lmstudio-community/qwen3-8b and connects automatically on load. If the tunnel restarts, send testers the fresh address to paste into “Connect the line” as a fallback.

  5. Stay plugged in. Laptop on, plugged in, and awake for the whole testing session — if it sleeps, the line goes dead.

The road ahead

After tester feedback, Moshi Moshi opens to everyone with pay-as-you-go chat in crypto. Planned coins:

BTCBitcoin LTCLitecoin ADACardano SOLSolana FETFetch.ai / ASI DOGEDogecoin SHIBShiba Inu HBARunder consideration VETunder consideration VARAunder consideration

Accepting many coins helps the coins, helps the site, and puts Moshi Moshi in front of every one of those communities.