--- Roadmap Milestone 1: Native Token Conversion. Network Security: We received a preliminary security audit report from Halborn, and now working together on a final report. Once the final report is ready, we'll be able to hand it over to the exchanges to integrate the TARA native token into their systems. Continued aggressive stress-testing of the network and fixing the current protocol issues, pointed out by Halborn. We were able to address most of them including potential DDoS attacks. Community Site, Explorer, and Documentation: Completed on-chain economics integration into the Community Site and deploying the tests on the development network, and then on to the mainnet candidate. This will eventually remove the KYC requirements for staking yields. Improved the stability of the Explorer, which is a crucial graphic user interface to keep track of transactions on the network. We're very close to completing the GraphQL integration into Explorer to improve the its stability. Completed the exchange integration technical documentation. Taraxa's RPC layer is fully backwards-compatible with Ethereum, so we don't expect any technical difficulties, as the exchanges can just port over the code from Ethereum and change the RPC node they connect to. So, the exchange's conversion queue-time shouldn't be driven by technical difficulty, but there might be a backlog from other projects. --- Roadmap Milestone 2: Social Listening Platform. Analytics Layer: While we're moving forward on all aspects of the application data collection, and data analytics for a decentralized data collection client, we'd like to highlight the progress made on the Hype app's analytics layer. To those of you who's new to Taraxa: our social-listening app makes publicly available social data highly trusted. Hype is the first application in the world that allows people to hype up a specific project, or a topic, by automatically identifying and rewarding the right types of community-building activities. This is a big problem familiar to anyone who has ever run a community bounty: bots and bounty hunters would immediately overwhelm the conversation and drain all the bounty funds. Analytics Layer Highlight: Spam Detection Echo and Hype are currently integrated with Telegram, simultaneously listening in on a few thousand public Telegram groups' conversations to figure out who's making quality contributions. A critical part of the Analytics Layer is filtering out spam and bot-like behaviors. But how do we identify spam? The very first thing we look at is the frequency of posts: if a specific user has posts a message at a far higher frequency than what a normal user has in a chat, then we automatically filter it out as spam. But what if someone simply changes a few words or characters message, how could we tell they're the same message repeated over & over again? To tackle that, we're using a highly-scalable LSH (locality-sensitive hashing) algorithm to detect near-duplicates, i.e. messages that are not exactly the same but look the same to a human. The second part to figure out if the contribution is relevant to a specific social campaign (e.g., a social bounty). Here we use something called an embedding, which is a vector representation of a message. We use this to compare against known (provided by the project) embeddings representations of the project's general descriptions. If they are extremely dissimilar, then the message is not relevant to the project and is not rewarded. The third part that hasn't been implemented but is in our roadmap in terms of bot detection is creating a reputation for individual accounts. Instead of looking at interactions with a specific message, tweet, or campaign in isolation, we'll go back and look at the account and the history of posting. For example, an account can post the same comment to a lot of places, while that account doesn't look like a bot-spammer in a single chat / channel / tweet, they are indeed a bot when looked at the full posting record in aggregate. Coincidently, one of Taraxa's tweet (below) was hit by hundreds of bots today (see below), you can see a lot of the bot / spam-like behaviors in the replies. This tweet earlier was hit by hundreds of obvious bot repliesAlthough we're currently focused on Telegram, the same analytics tools that we use to filter out bots will directly apply to the way we're going to detect bots on Twitter. So this is definitely something we're looking to do in the future. Our initial spam-filtering results are extremely good, this is very exciting! We built an analytics pipeline and ran a 3 million message data set through it. We now move to test our analytics pipeline with a much larger data set, and we'll share the results with the community.  We're super excited about this!