A profile stores a preference. Useful memory belongs in the conversation. That’s the design detail I’d watch in @gakkuai’s Cabin demo. Raku recalls the learner’s interest in an author, then asks about an “aha” moment in the book they’ve made the most progress on. The learner brings an idea about categorization and analogy into the exchange. Raku responds to that thought, and a Journal entry captures it afterward. Personal context gives the conversation a direction; the learner still has something to explain. See remembered interests meet a new idea: gakku.ai

Great share

There’s a big difference between releasing weights and releasing something people can use. The first gets attention. The second builds an ecosystem. That’s what stands out about Tencent Hy4 preview. The model is enormous—770B total, 49B active—but it arrives with Apache 2.0 licensing, a native 1M context window and immediate support for vLLM and SGLang. The ~214GiB mixed GGUF is especially notable. It’s down from ~1.5TB, while Tencent’s reported evals remain close to BF16. The model is unfinished by design. That might be the best part: builders are being invited into the feedback loop. @TencentHunyuan @TencentAI_News

5/ Hosting isn't the product. The live link is. Your agent acts as the designer and developer. Showly.ai simply takes that finished work and makes it online, accessible, and ready to share 🚀 Complete with saved version history. You can roll back to an earlier state anytime you need. Preview → Approve → Publish → Restore.

4/ Discover what teams are publishing with Showly.ai. 📊 Weekly Performance Report 📋 Product Spec 👀 Client Review 🚀 Prototype Launch 👥 Team Onboarding 🔬 Research Synthesis The agent builds it, Showly.ai ships it.

3/ Stop sending screenshots of localhost. Outline the page. Your agent does the building. Check your private Preview. Publish when it's perfect. Distribute a single link. 📊 Weekly reports. Product specs. Client reviews. Portfolio demos. Research summaries. Same simple workflow. One clean link.

2/ Built for the agents you already use. Seamlessly connect with Claude Code, Codex, Cursor, OpenClaw, Hermes, or any other compatible agent in your stack. 🔌 Forget about learning another clunky editor. Just detail your task, tell your agent to deliver the final work as a web page, and pop open the Preview it hands back. ✨ Showly.ai slots perfectly into the exact AI workflow you are already running today. 🚀

1/ 🏆 Meet Showly.ai — The Publishing Layer for AI Agents. Too often, brilliant AI output gets trapped inside endless chat windows, raw .md files, or stuck on localhost. With Showly.ai, your agent delivers that hard work as an actual web page: perfectly structured, highly readable, entirely presentable, and instantly ready to share 🚀 Your agent handles the creation. Showly.ai provides the live page that anyone can easily open.

Your AI agent is already doing the heavy lifting—writing detailed reports, running market research, drafting proposals, and building out websites. 🤖 But please stop letting all those amazing results get lost and buried inside endless chat logs. 🛑 Ask your agent to deliver that work as actual web pages that your team can easily read, present, and share with others. 📄 Your agent handles all the content creation. Showly.ai handles the preview and generates the shareable links. 🔗

4/ Change Two Lines. Keep Your Entire Stack. DIT.ai integrates seamlessly with the OpenAI SDK, LangChain, LlamaIndex, Vercel AI SDK—and whatever framework you already use. To switch: Change the base URL Replace the API key That’s it. Minimal adjustments. Keep your workflow.