Searxly is actively working on $SEARXLY utility features.

The first inference futures protocol just went live on @solana ! You've been gambling on compute costs every single day. We just gave you a way to stop. Buy AI credits early. Lock the price. Build without fear. trade.autonaisol.xyz/

Dot now runs as a private MCP-native app inside Claude/ChatGPT. No other privacy-ai provider has achieved this. The DotMCP was already popular, but we have just fully reimagined its architecture to increase its power. Dot is a private AI app that lives inside your Claude/GPT. It is not a chatbot embed, or a prompt handoff, it is an fully private model surface you, or your agent can summon inside the mainstream providers you normally use. The distinction that makes this so powerful: Whatever you ask Dot does not go through the host. Simply type inside Dot’s sandbox, and our widget sends the request directly to Dot’s private inference edge. Claude’s model never reads the prompt, and it is not part of Claude’s context. This is a fully private AI inference inside a fully public AI inference. What we rebuilt: → custom model picker, no native dropdown → light dark mode → full dot model lineup inside the widget → private image generation through the widget → MCP session tokens for widget-to-dot requests → hardened cross-origin chat edge → shared rate limits with the main app so the widget can’t bypass caps ( Beta phase ) → provider/model ids normalized before they reach the UI → privacy-minimized frontier routes → GLM 5.2 private lane with 1M context Our MCP is a separate inference surface with its own auth boundary, routing layer, model stack, image route, privacy receipts, rate limits, and no host-model prompt exposure. Endpoint → mcp.usedot.xyz/mcp Setup → git.usedot.xyz/our-products/…

Aped $SEARXLY at the dip - first Bankr coin Elon has interacted with - only token on Base built entirely with @grok Build - the one with the highest chance of getting an Elon’s interaction again @Searxly is built by @myrhex using Grok Build, it is a privacy-first native macOS browser integrating local SearXNG search, real WireGuard VPN, on-device AI, and zero telemetry. The dev Myrhex is followed by @XFreeze and @tetsuoai, both are the accounts Elon interact with very frequently. What stands out: • Built with Grok Build, @xai’s coding agent • Fully open-source and shipping updates DAILY • Native SwiftUI WebKit app, not a wrapper • Token directly funds development, not VCs • Future utility includes holder verification, premium features, and VPN benefits It already has a working product, an active community, and direct exposure to the xAI ecosystem. CA: 0x0fDC79b868bc4a6295CD94397f61890f68C38BA3

$AUTO is doing good, unique tek and fits with compute narrative. 25K Mcap to 150K Mcap ~6x almost. I feel there is huge upside for this, don't forget dev @thatweb3gee is winner of zypherpunk hackathon organized by @mert

Interesting angle. If AI companies eventually stop subsidizing compute, the winners won't just be GPU providers it'll be protocols that help users hedge future compute costs. $AUTO is effectively turning compute into a tradable market. Feels like a narrative that only makes more sense with time. @autonai_sol CA: 7io8XEMRMoQoCvD3phKR2QR5EeoMmJnYM1dsjVdYpump

AI Compute will be a huge narrative in the future slowly as AI companies finally stop subsidizing users and put their charges up. $AUTO has a really interesting concept of locking in future compute prices in a decentralised way. For comparison, this is how much AI companies are subsidizing compute prices right now. Most providers will want to lock in cheaper compute prices before the subsidy ends and the hardware prices skyrocket more dev is extremely based winning the zypherpunk hackathon organised by @mert Can see this doing well. CA: 7io8XEMRMoQoCvD3phKR2QR5EeoMmJnYM1dsjVdYpump

Fable is still available inside Bloom Copilot. This is the problem with centralized AI. Access to intelligence can disappear because of a policy change, a region restriction, or a decision you had no control over. We think that's backwards. The future of AI isn't just about building better models. It's about building infrastructure that can route between models, remain accessible, and give users a choice. Today it's Fable. Tomorrow it will be dozens of models. That's exactly what we're building with Bloom Copilot. Try it: bloomterminal.io/copilot

DotCode now runs a first-of-its-kind, fully private, agentic loop that can see exactly what it is building, in real time. You will have undoubtedly seen a lot of talk about agent coding loops, but they present a fundamental problem that we have seen no other provider fix. They break the moment the agent needs an asset that it cannot product. DotCode can now read our custom SKILL.md, route through Dot Models, call the DotMCP, generates its own assets and save these assets to the repository, verifying the page locally. No handoff, and no human in the middle. And yes, the entire loop stays within Dot's privacy boundary. No API’s, no third party calls. Don't break the loop, use DotCode, available soon.

Another ATH for ethereum:0x8d010bf9c26881788b4e6bf5fd1bdc358c8f90b8 Milly soon??? Study ethereum:0x8d010bf9c26881788b4e6bf5fd1bdc358c8f90b8 -> x.com/Crypt0Pirate_/status/2…

Shoutout to the Funn community 💧over 10M downloads across our apps and thousands now earning $FUNN daily. Thank you for trusting us with your health journey since 2013 ❤️ And remember: every time users redeem points for $FUNN in the WaterMinder app, tokens are burned 🔥 Track it live at points.waterminder.com We build for you. Drop a 💧 if you’re tracking your water today!

We have achieved Fable-5 level intelligence through the Dot Supercharged Fusion lane. We've built a private-preview fusion lane that is displaying frontier-class behavior on hard builder tasks. Dot Supercharged scored #1 in our 30-task benchmark across privacy infra, routing, API design, agent safety, debugging, evals, and model-layer decentralization. Results: Dot-Supercharged: 86.2 Claude Opus 4.8: 77.4 GPT-5.5: 74.5 Dot Gemma 4: 67.3 It beat the single frontier/open routes we tested by combining: -multi-model fan-out -privacy boundary -synthesis -rubric-scored final answers Because of the immense power of this model, it will be limited to 3 prompts per user in our free beta cohort. Once monetized, this model will be fully de-restricted. Available for testing at: app.usedot.xyz

Another ATH for ethereum:0x8d010bf9c26881788b4e6bf5fd1bdc358c8f90b8 Milly soon??? Study ethereum:0x8d010bf9c26881788b4e6bf5fd1bdc358c8f90b8 -> x.com/Crypt0Pirate_/status/2…

Another step forward. Bloom is joining the IBM Partner Plus ecosystem. As we build the capital intelligence layer for DeFi, we're excited to learn, grow, and accelerate alongside some of the world's leading technology organizations. Thank you @IBMPartners @IBM. Back to shipping.

$SYNTHETIC looking good here. Team consistently shipping since days quietly. Part of @aeonframework & @miroshark_ ecosystem @aaronjmars been supporting them & is in touch with Dev closely. x.com/aaronjmars/status/2060…

New ATH for $DOT Deserves higher!!!

DotMCP is now live for users & agents, the first fully private MCP that integrates directly into Claude, ChatGPT, and Claude Code. An autonomous agent running inside Claude, ChatGPT, or Cursor can now call Dot infrastructure for sensitive operations without exposing what it asked. Every agent operating inside a host LLM now has a private side channel to a model that won't log it, profile it, or filter what it asked. The MCP tool call from your host to Dot does not carry your prompt. It carries a handoff. The handoff opens a private-session widget rendered inside your host but communicating directly with Dot over a separate channel. You type into the widget. The widget streams to Dot. The host sees: a session opened. It does not see what you typed, what was returned, or how long the exchange lasted. The result: → The host model cannot read your prompt. → The host model cannot apply its policy layer to your question. → The host model cannot log, filter, or modify what you asked. → Dot's no-retention invariant extends through to your host, instead of stopping at our backend. This is a new era in privacy infrastructure, one that refuses to be restricted by modern protocol. Try it on any MCP-compatible host. Docs → usedot.xyz Endpoint → mcp.usedot.xyz/mcp Setup -> git.usedot.xyz/our-products/…