Skip to content
Evgeny
Rodionov
← All projects

Personal working system / 2026–Present

Knowledge systems · Independent project

AI Working Memory

Continue the work with its context intact.

I built a source-grounded working memory for projects, decisions, agreements and the questions still worth answering.

A closer look at the workExplore the case
AI Working MemorySystem overview
01

Preserve the source

Keep the original material and its context intact.

02

Maintain the knowledge

Connect projects, decisions, people, and source-linked claims.

03

Answer in context

Retrieve the useful account and verify the underlying evidence.

04

Expose the gap

Distinguish known facts, contradictions, and questions worth researching.

Process illustration · Read the build decisions below

01 / The problem

The agreement is somewhere in the history

Useful context accumulates in calls, files and conversations. A later decision may depend on a small qualification or agreement that never made it into the final document. I needed a way to recover that context without reconstructing the history each time.

02 / The build

How I built it

01

The system keeps original sources, maintained knowledge and working drafts in separate layers. Project and company accounts connect the useful facts, decisions and open questions. An answer can return to the original passage when exact wording or attribution matters.

02

The structure developed around the jobs I use it for: recording experience, asking questions, preparing communication and deciding which gaps deserve more investigation. It distinguishes current attention from historical company work, so an old backlog does not silently become today’s priority. Conflicting accounts remain visible until the evidence resolves them.

My role

I defined the purposes and developed the framework with AI, then refined the structure and operating rules through use. My work centers on what the system should remember, how it should retrieve it and how it should avoid treating generated text as established fact.

03 / In practice

From an old conversation to a current answer

01

Question

Example

What did we agree to deliver, and what still needs a decision?

02

Project account

Example

A dated summary with decisions, status, and open questions

03

Evidence

Example

The original passage, speaker, and date

04

Answer

Example

The agreement, the remaining uncertainty, and the next decision

Illustrative retrieval path. No private conversations or actual agreements are shown.

04 / The result

Source-groundedanswers and maintained project context

I use it to recover agreements, understand project state and prepare conversations with the relevant context already available. Its value is being able to continue earlier thinking and investigation. The system can only answer from the material it holds, so missing evidence remains visible.

Keep exploring

Content Machine

I built a content system that connects search research, company knowledge, writing, editorial review and publishing operations.

Next case
Ask my
AI avatar