← All personas

Marcus

Corrections and disambiguation across sessions. Marcus mentions two different people named Josh across 8 sessions and 2 platforms. The system tracks who's who, and when facts change.

Developer, 28 · 8 sessions across claude, chatgpt

The Difference

Without Kenotic

  • ChatGPT knows Marcus is a developer who prefers Python (its own memory). It does not know about tonight's deploy failure. That context lives in Claude.
  • Marcus mentions "Josh." ChatGPT has no idea which Josh. Marcus has to re-explain.
  • Marcus says "the race condition fix I was trying" and ChatGPT asks what race condition. The debugging history from Claude is gone.

With Kenotic

  • ChatGPT has the full context from Claude. Marcus says "the race condition fix" and ChatGPT follows immediately.
  • The bridge knows Josh Chen (Marcus's teammate) and Josh Williams (Seattle office). It never confuses them.
  • The correction is preserved: Session 2 hypothesized memory leak, Session 3 corrected to race condition. Timestamps intact.

The Living Timeline

Scroll through to see how facts, corrections, and context build over time.

CareerRelationshipsSpatialEmotionalcorrectedwithout continuitywithout continuitywithout continuitywithout continuity

Reconstruction

Ask a question that requires connecting facts across sessions, hosts, and time. See where each part of the answer comes from.

Q

Which Josh got promoted last quarter?

A

Josh Keller, your team lead, was promoted to Staff EngineerC in early February. Josh Martinez is your roommateC. He plays drums and is moving out at the end of MarchC.

Sources

S2CS2CS6C

How It Works

Four steps. No cloud. No LLM at read time. Retrieval is deterministic, so answers don't change when you switch models or when the provider ships an update.

Step 1

Ingest

You talk to any AI. The bridge captures structured understanding: facts, corrections, temporal order, emotional context.

Step 2

Store

Facts, corrections, temporal order written to a local SQLite file. On your device. No cloud.

Step 3

Reconstruct

You switch platforms. The bridge traverses the graph across all sessions, hosts, and time periods.

Step 4

Deliver

Grounded answer from specific moments. Every claim traced to a source session. No hallucination.

7 Points of Continuity

Each point represents a capability that requires genuine continuity, not retrieval alone.

1

It knows what happened when

Marcus's deploy failure on March 14 is tracked as occurring AFTER his Portland move, not stored as a standalone fact. The system knows the sequence.

2

It knows who's who

Josh K. (coworker, promoted to Staff) and Josh M. (roommate, plays drums, moving out) are tracked as separate people with separate timelines.

3

It tracks what changed

When Marcus corrected the deploy date from March 12 to March 14, the system preserved both versions: the original claim and the correction, with temporal context.

4

It remembers across conversations

Facts from Session 1 (Portland move, first PR) are available in Session 8 without re-stating them. The system carried them forward.

5

It connects dots across sessions

Answering 'Which Josh got promoted?' requires connecting Josh K. to 'promotion' to 'team lead' across multiple sessions. Not a single lookup.

6

It works across AI tools

Marcus used Claude (Sessions 1-3), switched to ChatGPT (4-5), back to Claude (6-7), then ChatGPT (8). Memory persisted across every switch.

7

It preserves how things felt

Marcus's impostor syndrome (Session 3) and growing confidence (Session 7) form an emotional arc that the system tracks alongside career facts.

Next: Dr. Amara→
Back to all personas
Read the thesis →·ATANT paper (arXiv) →·All demos →
For InvestorsRequest Access
For DevelopersRead the Paper
For EveryoneJoin Waitlist