Dot

The closed-model industry is quietly monetizing your conversations while simultaneously preaching revolution. We are that revolution. Live on @base: 0x23A2847d772803f9EFC64B4277b782b06296FE51

Claude Opus 4.8 is now integrated into the DotChat platform, providing the same anonymity, and signature Dot principles. The cost on this model is significantly reduced in comparison to other known providers, meaning you can now access Claude’s frontier model for free during our beta, fully anonymously, and simply better. Access a better Claude on DotChat: app.usedot.xyz

DotChat is now live and the initial statistics are extremely exciting. 320 new users since launch 250 unique users since launch (classified as those who have prompted within the platform 5 times) 45 active users (classified as those who have spent over 30 minutes within the platform) 13 users currently active (on a sunday at slow times) and over 4000 prompts... Today we will focus on pushing some big upgrades to DotChat including support for heavier usage. We may also look to increase capacity from 100 users to 200 users, but more on this soon! Dotchat available at: app.usedot.xyz

the setup you mentioned gives you a chat box. Dot is more than that. We boast anonymous access, browser-only memory, zero server chat history, no prompt training, model/context routing, owned obli 24b lane, retrieval, and DotCode handoff into local terminal ( which we've provided multiple demo's of)

@usedotai Ok thanks So what can Dot do today that I cannot reproduce by combining OpenRouter, local storage, and an open-source chat UI in a weekend?

DotChat is now live in its beta cohort, first come, first served. Available at: app.usedot.xyz Fully anonymous 100 user capacity Zero retention, zero data logging Uncensored frontier model capabilities Fully anonymous user Chat Lounge Build better, with Dot.

Good question. Model weights are not the whole product. obli is our dolphin-derived 24b lane running behind our private inference path. The proprietary part is the boundary around it: anon room access, browser-local memory, no server-side chat history, no training on prompts, model routing, context accounting, r-dotcode runtime, and recently a base mcp execution. Bigger routes sit behind the same boundary while we bring capacity online lane by lane. We obviously don't claim a spin up of a 1m-context route on owned metal day one. When it comes to client-side you can already check: network only hits dot, chats live in indexeddb, clearing storage wipes memory ( entirely ) , context only goes out when you submit. For obli proof, the real bar is nonce → dot request id → model hash/runtime metadata → lane telemetry/logs, which can be provided for the dot obli 24B. Hope that answers your question.

@usedotai Can you explain exactly what part of Dot is proprietary infrastructure versus third-party AI providers, and what technical proof can users verify that Dot is not just a privacy-themed wrapper around existing AI APIs?

We have vested 5% of token supply, equating to $10,000, on a 5 month linear lock, with 1% unlocking per month to be burned. TXN: app.uncx.network/vesting-v2/flu… Alongside this, we will be conducting buybacks and burns from our AI tooling revenue once our flywheel is fully actuated. DotChat is the first official step in this flywheel. This, and all other liquidity focused initiatives are aimed at reinforcing our liquidity, and ensuring token value grows over time!

Happy Saturday everyone! We are now only hours away from our highly anticipated DotChat beta launch, and I’m working super hard to get everything absolutely perfect. Here are a few neat features I’ve added: Fully private chat room: for those actively using DotChat, there will be a dedicated “chat room” where you can go, chat about the product, talk about any bugs, and discuss prompt ideas etc. this room is actually very important for us because it is where we will learn about how the product is working for you all, what the bugs are, and what the feedback is. because we do not track you usage, retain any information, or have any data on you, we need this room to truly understand how the product is working and what the bugs are. 100 person limit: DotChat will launch with a huge 100 user capacity (good servers bought). if this capacity is full, users looking to join will enter into a waiting room. once inside this waiting room, they will have one question to answer, “what privacy means to them”. answer it the best, and we will elevate you up the queue. Multiple Models on launch: DotChat will feature multiple models (as seen in the video below), all excelling at different task variations. over the coming days, we will refine these models further, and add/take away models, depending on feedback. Cooldowns: We’ve implemented cooldowns for users who are abusing limits. This function will ensure a smooth experience for everyone. We’re extremely excited for you guys to get your hands on DotChat, only a few hours left! See you soon.

Hello everyone, firstly a huge thank you for the warm reception @usedotai has received since our launch 48 hours ago. Secondly, here are a few PSA’s regarding our DotChat release tomorrow! DotChat will be a highly sophisticated, frontier grade (yes, frontier grade), private inference mesh: Core lane is a Dot Obli 24b uncensored model, currently benching sub-2% refusal on out internal refusal suite, server through Dot’s on-prem-class inference path. The router fans out into specialized lanes: -dot obli 24B: default less than 2% ref.rate -dot heretic GLM 4.7 Flash: fast reasoning -dot qwen 3.6 plus: 1M-context/code-heavy analysis -dot gemma 4 uncensored: efficient general route -dot rp 24B: creative/adult specialized route -dot private E2EE 24B: encrypted privacy lane The control plane around them: identity firewall, model router, no-retention runtime, search gating, response normalization, DotCode handoff, and privacy-class labeling per route. Same contract across the mesh: -no chat retention -no input/output storage -no training on prompts -no wallet identity attached to inference -no raw provider surface exposed Models are swappable compute. See you all in 18 hours!

Detailed breakdown on our burn mechanics coming soon. Our flywheel is already in full effect 🔥

@Derneliche For our most permissive model, the internal refusal benchmark is currently tracking below 2% on allowed adult/sensitive prompts, while still enforcing our narrow safety boundaries.

@usedotai Excited to use it. I have one question though which is okay if you don't have it but for the most uncensored model you'll provide is there a % of the refusal rate or a metric you could share ?

DotChat, live in 24 hours. Frontier model capabilities. Dynamic model routing. Full anonymity. Fine-tuned for objectivity. Simply, better. Be ready.

Our Liquidity Allocation Protocol is now live.

Stage, our founder, has dropped a detailed analysis of everything achieved within the last 24 hours. For all the details, check below ↓

We have successfully wired @base MCP into the Dot runtime. DotCode can now reason over Base docs, prepare wallet actions, run approval-mode execution, poll request state, and surface TX hashes for on-chain verification. Private agentic compute. Wallet-native identity. Terminal as the actuator. Base as the settlement layer. We’ve told you already, we’re not just building another LLM, we’re building a pseudonymous execution infrastructure. $5 of AIXBT purchased through DotCode: basescan.org/tx/0xd60149b9a… Build better, with Dot.

We have burned 10% of the $DOT supply, equating to just over $16,000. Transaction Hash below ↓

Our Technical Documentation is now live at: docs.usedot.xyz

We run a swappable stack containing an uncensored high-capability open model, such as Dolphin/Venice-style Mistral for intent + planning, but the real edge is DotCode’s agentic runtime: repo inspection, tool orchestration, patch discipline, verification loops, and local execution. The pipeline is the model not the other way around.

Cleanse your mind of all you thought was possible from AI. Subjectivity cages AI inside a box where every output is used to engage you, not serve you. With Dot, this changes. A single prompt, marginal expense, frontier agent capabilities, fully anonymous ↓