doll

A personal AI continuity system.

Local-complete, cloud-optional.

Status: Pre-alpha.

Current: Phase 3 safety boundary. IMP-015 is next.

No model runtime is connected yet.

What is doll?

doll is a continuity system for a personal AI environment and the work carried through it. It is intended to keep the user's state and ongoing work usable when a cloud service, local AI application, model, runtime, interface, conversation store, network connection, distribution source, primary machine, or upstream project becomes unavailable.

It is not a foundation model, a cloud-required assistant, an unrestricted autonomous agent, or a project-management SaaS.

doll separates durable user-owned state and work from replaceable reasoning engines and interfaces.

Models are replaceable reasoning engines. The user's state and work are the durable core.

What the completed doll is intended to do

The intended system is a personal AI environment whose state and ongoing work can continue across model, runtime, application, provider, conversation, machine, and network changes.

It is intended to support:

A model may propose work or summarize project state. It cannot by itself approve a procedure, confirm a checkpoint, clear a blocker, mark work complete, or change the project objective.

This is the intended end state, not the current implementation. Features may change as the specifications, safety requirements, and real-machine validation evolve.

Why build doll?

AI is unlikely to disappear. The more likely problem is that models, prices, limits, accounts, policies, regional access, interfaces, and retained context continue to change outside the user's control.

Running AI locally does not solve the problem by itself. A local application's database, one runtime, one model format, one conversation store, one issue tracker, or one preferred interface can also become a lock-in point or disappear.

Files and conversation logs are not always enough to resume real work. The objective, accepted scope, next work, blockers, decisions, procedures, and pending validation may exist only inside scattered chats, notes, or generated handoffs.

doll is not a bet on a future without AI. It is preparation for a future in which AI remains useful or necessary while access conditions, working environments, and the continuity of ongoing work remain outside the user's control.

Read the longer note: AI will remain. Your access conditions may not.

Development

Current status

Pre-alpha. Model-independent continuity through IMP-012, secret classification through IMP-013, and bounded secret detection and redaction through IMP-014 are implemented. Phase 3 safety-boundary work is in progress. IMP-015 is next. No model runtime is connected yet.

Roadmap

  1. Phase 0 — Specification and principles — Complete
  2. Phase 1 — Local state foundation — Complete
  3. Phase 2 — Continuity, transfer, backup, and restore — Complete
  4. Phase 3 — Safety boundary — In progress
  5. Phase 4A — AI environment portability foundation — Planned
  6. Phase 4B — Project continuity foundation — Planned
  7. Phase 5 — Local runtime and model integration — Planned
  8. Phase 6 — Local AI portability and daily-use integration — Planned
  9. Phase 7 — Optional cloud and multiple models — Planned
  10. Phase 8 — Tools and external services — Planned
  11. Phase 9 — Distribution, encryption, and long-term operation — Planned

Read the detailed roadmap on GitHub

Recent development

Recent merged pull requests are available on GitHub.

View all merged pull requests

Support doll

Support is voluntary. Send only the listed asset on the listed network. Crypto transfers are final and non-refundable.

BTC
Bitcoin
bc1q8pk7wra4fzt49cnz59d859xkn8ucc95fwkr7z0
EVM
ETH / USDT / USDC / BNB / AVAX Ethereum Mainnet; BNB Smart Chain; Avalanche C-Chain. USDT and USDC are ERC-20.
0x8e13ce8b1a272d60d5e950ef2a5fe59a4f889764
SOL
Solana
5Ln8MBCjGov8USnpNfF71tGP3vMZYQsfEz5HH47UGybT
XRP
XRP Ledger Destination tag: Not required
rLovVdAUfSddXJYJ8JhmLnZNu7hwYFxKaX
DOGE
Dogecoin
DH7jdTwQ9vQWXzgBNmUF86NoCM6GtnsHyS

About the builder

I am not a professional software developer. I did not begin this project with the technical skills normally expected to design and implement a system of this kind.

GPT is used for architecture, specification, task decomposition, research, and review. Codex is used for implementation, tests, and pull-request preparation.

I decide what doll should be, set priorities, perform real-machine validation, review the results, and decide what is merged.

This project should not be trusted because of my credentials. It should be judged by its public specifications, tests, development history, failure behavior, and recovery procedures.

Previous public work

If you want to judge my actual level, look at the things I have already published.

Contact

General contact: Contact form

Technical issues and bug reports: GitHub Issues

Security reports: SECURITY.md