Misconceptions About Clawbot Clawbot has been blowing up on Twitter recently. It's positioned as a personal assistant agent. After doing some research, here are a few misconceptions you need to be aware of: 1. Clawbot has long-term memory This isn't entirely accurate. Clawbot ultimately relies on large models like Claude, and these models have strict limitations on context length. If the context exceeds the limit, the model won't be able to retain it. It might be using skills or RAG (retrieval-augmented generation) techniques to manage context. 2. Clawbot doesn't upload local data This is incorrect. When Clawbot remotely calls the LLM (large language model), it transmits tokens to the remote model, and those tokens are essentially local data. If you want complete privacy, you'll need to use a local model. 3. Clawbot is open-source, so it's safe This is also wrong. Although the code is open-source, Clawbot has very high permissions and can operate your personal computer. It may also have publicly known vulnerabilities. Similar to open-source blockchain contract code, open-source doesn't guarantee safety—it can even introduce vulnerabilities that hackers exploit. Web3 users are strongly advised not to install Clawbot on machines with assets. In summary, Clawbot is an open-source, free personal assistant bot. Community feedback suggests the user experience is pretty good, but you should also be mindful of its risks.