The household operating layer
A child’s hospital appointment, historical blood results, and the Spanish needed for the visit. One person, history and calendar supports the whole chain.
The investment thesis, November 2026
Other AI apps connect apps to their assistant. Akashic connects a person’s whole life into a record they own, then lets any assistant work with it.
Akashic must earn its place by making that foundation work every day. This is an early household pilot, not evidence of product-market fit or scalable economics.
Akashic is permanent storage plus an organising system. The model is the processor working on it.
The foundation captures permitted signals, preserves evidence, and connects people, life areas, time, geography and matters. Compatible models and applications use the relevant context to analyse, create and act within their authority.
Changing the model should not mean rebuilding your history. Local inference may be part of that future where the model, hardware and task suit it; a fully local replacement for every workflow is not a current promise.
Inputs
Communications, documents, photos, health, location, calendars and external information.
Lasting asset
Identity, relationships, history, preferences, provenance and permissions.
Replaceable capability
Better reasoning, new workflows and personal tools built on existing context.
Booking a child’s dentist needs the insurance policy, the clinic history, two calendars and the family’s places. Today those sit in separate places.
A policy is in an inbox, a payment in a statement, a renewal in a calendar. Useful work means reconciling them and keeping them current.
An assistant needs to know whose document it is, when it applied, what replaced it and which source supports a claim.
An appointment or a renewal means tracking commitments, replies, permissions and outcomes across days and channels.
AI providers already offer memory, connectors and agents. Our thesis is that a user-owned foundation can support these across providers and applications, with continuity beyond any one interface.
These describe the direction. They are not claims that the complete workflows are delivered today.
The user’s accumulated history is valuable to the user. It is not a data moat Akashic should hold hostage.
Our potential advantage is the work of making that record reliable and useful: integrations that survive change, accurate identity and time handling, traceable evidence, permission boundaries, dependable workflows and a product people trust with daily life.
Reusable integrations and workflow knowledge can improve delivery for future users without pooling their private records.
Competitors can copy features and pursue the same architecture. Defensibility has to be shown through execution, retention, distribution and service quality. It is not established by declaring incumbents unable to compete.
| Area | Current basis | Next proof |
|---|---|---|
| Daily use | Founder use and an early household pilot; browser workbench and Android capture. | Repeatable onboarding and sustained usefulness beyond the founder. |
| Connected record | Document and message ingestion, structured knowledge, search, agenda and source-specific capture. | Reliable coverage, freshness and identity across more sources. Capture gaps remain. |
| Replaceable AI | Model and harness integrations working against a shared record. | Measured task quality and context continuity across providers; qualify local tasks. |
| Useful workflows | Assistant, calendar and tracked work surfaces; explicit authorisation for external actions. | Repeatable completion of the scenarios above, with bounded delegation and recovery from failure. |
| Exit | A versioned export mechanism with a manifest and integrity hashes. | Independent completeness, restoration and useful migration tests. An archive alone does not prove portability. |
Our direction is a documented, testable export profile that preserves originals, identities, relationships, timestamps, provenance and permissions, so another implementation can make useful sense of it.
Relevant building blocks include JSON-LD for linked data, W3C PROV for provenance and RO-Crate for packaging. Solid and the Linked Web Storage draft are relevant to independently controlled storage and access.
These are standards to evaluate, not a claim of implemented conformance or certification. Where a personal-record profile is missing, we can publish one with validators and migration examples and invite other implementations to use it. Adoption would make it a standard in practice; a name alone would not.
People coordinating a busy household: messages, appointments, documents, recurring administration and shared plans.
Reliable capture and organisation, a managed service, and useful work completed. Household subscriptions and specialised workflow offerings are hypotheses to test, not proven economics.
From individual to household, then to professional and specialist uses where the same context already helps. Shared access must preserve each person’s ownership and boundaries.
Security assurance, permission design, connector reliability, phone capture and operational recovery remain substantial work. The pilot shows development is real; it is not yet evidence of product-market fit or scalable unit economics.
Akashic is founded by Ehsan Ahmed, who goes by East. It began with his own need for continuity across family, work, health and administration. We are looking for conversations with investors who understand patient infrastructure work and the value of user ownership.