I work on four or five projects in parallel, in fragmented days. Without a memory system, I lose the thread within half an hour.

Let me tell you about a real day. Not a use case built at a desk: what's happening today, as I write.

A quick technical note. When I say "Obsidian" and "vault", I mean the folder of markdown notes on my computer. Timo indexes that folder and makes it readable to AIs through a standard protocol. Nothing esoteric: text files in a folder, plus a bridge to the AIs.

08:30 — Rebuilding context

I sit down, open the computer, start with the project I'm furthest behind on — a corporate portal where I do external architecture. I haven't touched it in four days.

I open a chat with Claude. I write:

"Let's pick up the portal. What were the open todos last week, and what was I trying to resolve at the end?"

Claude searches through my notes via Timo, the mgportal section. Finds the recent work notes. Replies with three open todos, and flags that the last session closed on an unresolved decision about the names of the translation modules (i18n).

I didn't do the search. I didn't open Obsidian, I didn't look for files, I didn't reread notes. The context arrived ready, in thirty seconds.

Without Timo, I'd have spent twenty minutes realigning myself.

11:00 — The decisional déjà vu

I run into an architectural doubt. I was about to implement an endpoint a certain way, but I have a déjà vu — I feel like I've already discussed this.

I ask Claude:

"Did we decide something, in the past, about how to handle pagination of long lists in this project?"

Claude finds two notes: one from three weeks ago with the initial decision (cursor-based, not offset-based, for consistency reasons), and one from a week later with a documented exception (offset-based for the report module, because tied to a legacy component).

I proceed with cursor-based, remembering the exception too, which otherwise I'd have redone from scratch.

This is what makes me choose to keep decision notes instead of deciding from memory: from memory, you're sure until you're not anymore.

15:00 — The thought dump

Afternoon dedicated to a code review on another project. As I read, I take quick notes in my markdown files. I'm not writing a document; I'm dumping the flow of thoughts: "this pattern is repeated here, worth extracting a helper", "silent side effect here that whoever reads later won't notice", "this function behaves well but the name is misleading".

Five pages of raw notes. I save, move on.

Timo indexes them in the background, automatically. I don't have to do anything. Tomorrow, if I ask "what was I supposed to flag to the team about that module after the review", the answer is already there.

22:00 — The recurring errors archive

A note I've kept for months is recurring-errors.md. Every time I run into a stupid error — one of those where you spend an hour before noticing — I add it there.

Tonight I add one: I lost forty minutes because I confused two very similar environment variables in a config file. I write: date, error, symptom, cause, solution, lesson (rename the variables to make confusion impossible).

Tomorrow morning, when I start a debug session on a random project, I can ask Claude:

"Have I seen a similar error before? Check recurring-errors.md."

If yes, Claude finds it and cites it, with the old solution. If not, it tells me it's new, and when I solve it I'll add it.

What would happen without

Without Timo, the alternative isn't "I remember by heart". It's one of these three things, all lived before:

  • I remember wrong. I decide as if it were the first time, I redo errors already solved.
  • I search the vault by hand. Fifteen minutes to find an information I need in thirty seconds.
  • I copy-paste context into every chat. Open a file, copy the relevant block, paste it in the prompt, start over. Every time. On every project.

Three ways to waste time, all real.

The difference isn't speed

The difference, in the end, isn't that I work faster. The difference is that I work on one hour of work at a time, I don't start from zero every time.

The project's thread stays. Past decisions stay. Already-solved errors stay solved. What changes is a continuity that, before, I couldn't keep without enormous cognitive cost.

It's not a revolution. It's just that now my working memory doesn't disappear every time I close a chat.

Try Timo free for 15 days — timoai.xyz. No credit card required.