BUILD LOG
Roadmap
A private AI service on hardware we operate: what works today, what earns trust next, and what comes later.
shipped in progress planned
Phase 0
Foundation
DONEThe gateway, billing and private chat that the rest of the service builds on.
- OpenAI-compatible gateway · billing · credits · subscriptionsshipped
- No-prompt-logging architecture · usage-only ledgershipped
- Local-first chat history · erase it instantlyshipped
- Web chat and API on hardware we operateshipped
Phase 1
Private AI platform
LIVELong-context AI, useful chat tools and private payment rails are live.
- OpenAI-compatible API · SillyTavern ready · 1M contextshipped
- Vision: image understanding in the chat and APIshipped
- Web search in the chat, on our own metasearchshipped
- Calculator · PDF, DOCX and XLSX reading · pictures in resultsshipped
- BTC · ETH · USDT · XMR payments, with an address of your ownshipped
- Tor .onion access · XMR subaddress per invoiceshipped
- 2FA · recovery codes · email verification · IP-restricted API keysshipped
Phase 2
Trust & developer platform
NOWMake privacy verifiable, spending legible and the API dependable in real applications.
- Usage dashboard: requests · tokens · cost · modelsshipped
- Machine-readable model capabilities in /v1/modelsshipped
- Public live status · fleet uptime · model availabilityshipped
- Warrant canary that reports its own expiryshipped
- Referral revenue share, paid in creditsin progress
- Privacy architecture: what is stored, where, and for how longplanned
- File and search retention controlsplanned
- Per-key budgets · expiry · revoke · model permissionsplanned
- Per-key usage · latency and error breakdownplanned
- OpenAI compatibility suite · API docs and client examplesplanned
- Configurable moderated API for applicationsplanned
- Incident history · maintenance noticesplanned
Phase 3
Distribution & reliability
NEXTScale when demand requires it, with measurable availability and reproducible operations.
- Smarter routing · automatic failover · capacity managementplanned
- 99.5%+ availability backed by monitoring and redundancyplanned
- Automatic node provisioning · reproducible configuration · backupsplanned
- Abuse and DoS protection without prompt loggingplanned
- Encrypted export/import; multi-device sync if demand proves itplanned
- OpenRouter as an additional discovery channelplanned
Phase 4
Optional modalities
OPTIONALSeparate product lines to pursue after the core platform is ready.
- Image studio — FLUX / SDXL / Ponyplanned
- Whisper STT · TTS · voice chatplanned
Phase 5
Compute-backed token economy
LATERThe token track stays separate and follows the service, trust and reliability work.
- Base testnet: $GM/$CMP contracts (OZ ERC4626 staking, non-transferable $CMP)planned
- $GM TGE + community points airdropplanned
- Fleet capacity scaled to back $CMP demandplanned
- Confidential-compute privacy tier (H100/Blackwell)planned
- Fine-tune / LoRA marketplaceplanned
Want in early?
Start with free trial credits in the web chat, or connect any OpenAI-compatible client.
Open the terminal