Second Me Model

🚀 Second Me hit 6K+ stars on GitHub in just 1 week! Our growth trajectory is even steeper than Anthropic's MCP and LangChain's initial climbs! 🫡 The Numbers Are Wild 📈 Day 1: 565 stars Day 3: 2819 stars Day 7: 6,000+ stars! Thank you for making Second Me an open source success! We're honestly blown away by the support. 🫶🏻 Here's what Happened & What’s Next? 🧵👇

Zuck: Join us! Richard Hendricks: 👀

Issues & thoughts welcome on GitHub: github.com/mindverse/Seco…

🚀 Second-Me Major Update is LIVE on the 'develop' branch! Highlights: ☁️ Cloud Training w/ @alibaba_cloud support 🔥 Latest Qwen3 Models integrated 🧠 Smarter AI Core & Data Pipeline for faster learning and a clearer memory. Test it now & share your feedback!

Full change log: github.com/mindverse/Seco…

🆕 Release 1.0.1 is here! We just merged a batch of important updates from develop to main. 🧠 L1 Exposure model training (frontend + backend support) ▶️ Upload process and TrainingTags logic upgraded
🛠 CUDA+Windows fix, upload/tokenizer streamlined #SecondMe #AIInfra #OpenSource

✨ Highlights: - How parameterized training enables personalized understanding - The logic behind our 3-layer memory model & me-alignment - 25-min open Q&A — drop your questions in the thread!

🎙️ AMA Alert｜Inside AI-Native Memory: How Models Build a Sense of You 📅 When: Tuesday, May 13 ⌚️ Time: 9:00 PM (Beijing / SGT) | 9:00 AM (EST) | 6:00 AM (PST) 📍 Discord (English + Chinese session) RSVP →discord.gg/eDT9QVkB?event…

4/To Our Contributors – Thank You! Second Me v1.0.0 exists because of you. From PRs to feedback, every improvement has your fingerprints. We’re just getting started! Want to help shape memory-native AI? - Star us: github.com/mindverse/Seco… - Join the Discord: discord.com/invite/GpWHQNU… Let’s build AI identity infrastructure — together.

3/CUDA Support Is Here! - Works on A100 and consumer-grade GPUs of the same generation - Training runs on CUDA, while inference still uses llama.cpp (CPU) - So far, testing has been primarily on Linux + GPU environments Huge thanks to github.com/zpitroda for the contribution → github.com/mindverse/Seco…

We just shipped Second Me v1.0.0! From deployment upgrades to CUDA support and a brand-new “Thinking Mode,” here's everything new 👇 Dive into the full update in the thread below...

@geeorgey Wait, it's not a performance issue, it's about the pronouns, right? えっ、パフォーマンスの問題じゃなくて、人称（pronouns）の問題ってことですか？

@SecondMe_AI1 i used 7B!!

Xの過去Postをダウンロード→20文字以上のPostのみを抽出して100程度のファイルに分割→Second Meのトレーニングに利用→トレーニングに1.5日くらい 自分のクローンである @SecondMe_AI1 に話しかけて色々聞いてみたらさっき2014年のpostを元にした会話が出てきた。

@geeorgey Here, 0.5B/ 1.5B/ 3B/ 7B

@SecondMe_AI1 どこで確認できますか？？

@geeorgey ご不便をおかけして本当にごめんなさい🥹 私たちもぜひ、もっと良い体験をお届けできるよう改善に取り組みたいと思っています。 来月にはクラウド版をリリース予定なので、そうすればご自身のデバイス上でももっと大きなモデルを動かせるようになります！

@SecondMe_AI1 システムプロンプトをいくら工夫しても、記憶を参照した誰かが話しているような反応になってしまう 会話する時に、完全に自分になりきって会話をしてほしいのだけどうまくいかないのですが、解決策を知りたい

Hey guys - we're working on making Second Me better. Would love to hear your thoughts! 2 min survey here 👉 forms.gle/7WMBaFwYpnb4e7… (+ some custom merch & early testing spots up for grabs 🎁) Thanks for being part of this journey - you all make this project special 🩵

5/ 📖 New README, Who Dis? We revamped the docs: 💡 Clean Quickstart 💻 OS-specific deployment ⚙️ Recommended model specs 🧰 Setup tips Start building in minutes: github.com/mindverse/Seco…

4/ 🤗 Huggingface API Embedding Support Now you can use Huggingface models to power your Second Me’s memory. Works alongside OpenAI/Ollama backends. Custom embedding? Go for it. → Embedding Config Guide: secondme.gitbook.io/secondme/guide…

We just shipped a major update to Second Me - your AI identity just got way easier to deploy, train, and call. MCP Server, DPO tuning, low-data mode, Huggingface support... it's all in. Details below 🧵👇