After four overhauls and millions of real-world sessions, here are the lessons we learned about context engineering for AI agents:

RT @mwseibel: Trying Manus Wide Research and loving it. If I were a product guy at that company, I’d be thinking of ways to make Manus mor….

RT @kimmonismus: With “Wide Research,” @ManusAI_HQ is bringing about a fundamental change in agentic work. I instructed the tool to scan 1….

Wide Research is our latest exploration in agent-agent collaboration. Built on our large-scale virtualization infrastructure, Manus can now autonomously dispatch a team of homogeneous Manus agents to work in parallel and aggregate the results. While building AI agents, we've.

Choose the system architecture that benefits the most from switching to a stronger model, instead of comparing design choices using the same model. I treat it as a regularization factor in context engineering. It helps you discover more robust, generalizable architectures. It.

Forgot to include a key lesson in the blog: Context engineering can also overfit - not just to specific model families, but also to today's model limitations via premature optimization. At Manus, we never commit to an architecture based on static benchmarks. Instead, we treat.

RT @ManusAI_HQ: Today, Manus Integrates Veo3! . Manus isn't just where you generate videos — it's where you craft cinematic experiences. Ch….

Thinking of you guys, while the rest of us are squashing bugs back in Singapore. 🫠.

RT @garrytan: The AI labs need better evals and one of my favorite current YC batch companies just released a one with a *lot* of headroom.

RT @SandAI_HQ: 🪄 Magi-1: The Autoregressive Diffusion Video Generation Model.- Now available at 🥇 The first autore….

RT @ycombinator: Usable AI agents are finally here. From OpenAI’s DeepResearch to xAI’s DeepSearch, we’re seeing the first real push towar….

RT @karpathy: Writing text back and forth with an LLM is like we're all the way back to the era of command terminals. The "correct" output….

Also, there are a bunch of fake Manus apps on the App Store. This is the only real one: Please be careful of fake apps, fake websites, fake accounts, fake emails, fake news, fake invite codes, and fake meme coins! 🥹😭😭.

The Jevons Paradox in AI is absolutely real. Over the past two years, we've been building chatbots and have witnessed the inference costs of models with the same level of intelligence decrease by 50 to 100x. Last year, when we believed that the efficiency was sufficient to.

Well Vasek, you brought it up first, so I can't be blamed now! 😄 I was holding back from tagging you until Manus became more stable. But no worries everyone, any issues are totally on our end - @e2b is an amazing technology!.

Good news: We've nearly doubled @ManusAI_HQ 's context limit & beefed up our sandbox hardware. Failure rate down dramatically! 📉. Bad news: With lower failure rate, now sessions run longer, system's sweating bullets, and failures are climbing back up. Had to throttle things

Actually, Manus doesn't use MCP. We were more inspired by my friend @xingyaow_ 's work: While we haven't fully adopted CodeAct, this work provided 3 crucial insights:.* Coding is not the ultimate goal, but rather a universal approach for solving general.

I feel we need a place for everyone to share their Manus creations.🤔.