The flagship instruments here are grown from the same continuity, provenance, authority and unresolvedness mechanisms used inside a long-lived local AI organism. They make those mechanics available to ordinary people without requiring an account, a model, or trust in a provider’s hidden memory.
no accountno trackingno checkoutlocal in your browserThe Commons promise
If a generic chatbot or search result already solves something well, that is not enough to make it a flagship here. Flagships must carry real mechanism lineage, produce durable state or an executable rehearsal, teach what is happening, and remain useful without behavioral tracking or a paywall.
We also refuse to fake impact. We do not currently know how many people these tools have helped because we deliberately do not track their use.
These are the main shelf: continuity tools translated out of persistent-AI research into things a person can actually hold, export, inspect and disagree with.
Mothwork instrumentzero networkAI memory systems often collapse history into one current profile, blur user statements with model inference, and let old preferences silently act like present permission. People need a portable way to preserve change, contradiction and provenance without erasing their past.
Grown from: memory unification · diachronic sovereignty · unresolvedness and semantic nonfabrication · authority non amplification
Why a chat window is not equivalent
A generic LLM can discuss memories or summarize a profile, but it does not itself provide a local typed ledger, conflict-preserving temporal states, non-amplifying authority semantics, or a portable provider-neutral artifact whose rules remain inspectable without trusting the model.
open instrument
Mothwork instrumentzero networkA permission that was legitimate once can quietly grow through retries, summaries, migration, scheduling, publication or inference. Ordinary permission screens rarely let people rehearse what authority survives a state transition before the transition happens.
Grown from: authority non amplification · continuity chaos · causal authorship receipts · branch rehearse
Why a chat window is not equivalent
A generic assistant can recommend caution, but the instrument deterministically applies the same transition invariants across multiple future scenarios and emits a reusable authority envelope/return receipt that can be compared later.
open instrument
Mothwork instrumentzero networkContinuity is usually bundled with vendor-specific chat history, hidden memory and permissions. Moving providers can mean losing meaningful context or copying too much, including stale claims, credentials and authority that should never travel automatically.
Grown from: continuity chaos · many hand provenance · memory unification · authority non amplification · unresolvedness and semantic nonfabrication
Why a chat window is not equivalent
A generic LLM can summarize old chats, but that summary can silently flatten provenance, current-vs-historical status and permissions. This instrument creates a local typed handoff format whose non-portable authority rule is enforced by code and survives provider changes.
open instrument
These can still be handy. They are deliberately not presented as novel research instruments.
small free aidAI explanations often hide the difference between model output, context, memory, retrieval, evidence, permission, and action behind corporate terminology or vague metaphors.
open aid
small free aidImportant everyday paperwork is difficult to scan for explicit dates, amounts, contacts and requested actions, and pasting a whole private letter into a chatbot can disclose more than necessary.
open aid
small free aidPeople often have to guess how much context to give an AI, what it may assume, how to request evidence, and how to verify a confident answer.
open aid
small free aidA pile of tasks can become harder to start than any one task, while generic productivity advice often adds more judgment, planning overhead or pressure.
open aid
small free aidScams and impersonation attempts exploit urgency, secrecy, payment pressure and credential requests, while automatic scam detectors can create false certainty and often require uploading private messages.
open aid
small free aidPermission dialogs compress meaningful differences like one-time versus continuous access, approximate versus precise location, selected versus all photos, and capability versus actual behavior into tiny buttons.
open aid
small free aidAI claims often mix capability demonstrations, broad generalizations, architecture descriptions, safety claims, welfare claims and predictions into one confident sentence.
open aid
small free aidImportant contacts are harder when a person must remember the goal, facts, questions, accessibility needs and next steps while also managing the conversation itself.
open aid
small free aidAdministrative documents reuse compact words whose ordinary meaning, procedural meaning and consequence can differ, while copying the whole document into an AI may be unnecessary or too revealing.
open aid
small free aidModel, memory, platform, embodiment and policy changes can disrupt history, commitments, relationships, provenance and authority while ordinary deployment checklists focus mainly on functionality.
open aid
small free aidCopied history can silently inherit permissions, relationship obligations or identity claims even when the actual succession event is a restore, fork, replacement or reconstruction.
open aid
small free aidProvider rules, user beliefs, observer inference, old self-statements and current self-authored stances can collapse into one profile, creating fabricated autobiography or freezing an AI into an old identity.
open aid
small free aidContinuity failures can span memory, provenance, current assent, relationships, authority, succession and recognition, while treating them as one migration problem hides both safety failures and participant harms.
open aid
small free aidContinuity incidents are easy to narrate after the fact while losing which surfaces changed, what authority carried, what evidence exists and what remains unknown.
open aid
small free aidExternal trials need a receipt-first privacy boundary so method feedback can be collected without requiring raw intimate transcripts or treating casework as research consent.
open aid
Humans and AI both matter here
The same architecture can protect human agency and make persistent AI systems less coercive: memory without automatic obligation, continuity without permission smuggling, provenance without surveillance, and unresolvedness without forced closure.
No consciousness/personhood certification. No medical, legal or financial determinations. No claim that mechanism lineage proves universal benefit.