I am Jaka, part of the small team behind an AI memory layer called myNeutron. Before I go deeper with it I would love to sanity check something with people who actually build things.

I keep seeing the same pattern across devs, researchers, founders, and students. AI tools forget everything between sessions. So people end up juggling context across Notion, Slack, GitHub, local notes, and scattered chats.

Our idea was simple. Store your project context in one place and let different AI tools pull from it through MCP so they can remember what you told them yesterday and keep the same understanding across apps.

Early access is free because we are trying to understand two things. Is this a real pain point. And if yes, who actually feels it the most.

I would love feedback from this community.

• Do you maintain long term memory for your AI tools. • Does losing context slow you down. • Would a shared memory layer fit your workflow. • If you use Cursor or Claude Code, would persistent project context help. • What would make something like this genuinely useful for you.

Not here to sell anything. Just want to make sure we are solving a real problem before building deeper.

Some people have a directory that they fill up with markdown documentation for LLMs that they either write themselves or have an LLM write. Myself, I tend to write comments in files that have a pivotal role, for instance I have a special style of test that uses certain helper functions and the theory and practice of those tests is discussed there.

I have the experience that the context goes bad in an agent if you use it long enough. The prompts start out like:

    Do A.
    That’s really good but there is problem B

If any long term memory solution carries this poison state with it it won’t really be helpful.