Most tools assume the hard part is storing a thought. It’s keeping it faithful.

Note-taking has spent two decades optimising the wrong half of the problem. Storage is solved — capture is fast, sync is free, search is instant. Fidelity is not solved, and almost nobody tests it.

A note you wrote at 2am still reads as present tense a year later. It carries no signal about whether it caught what actually mattered, whether the version you now remember is the version you wrote, or whether it has quietly stopped being true. It was accurate when written, it is still sitting there, and it is read as current. That is not a memory system — it is an unexamined archive you have been trusting without evidence.

The interesting question isn’t how much can you keep. It’s how much of what you kept survives contact with the person you became. HighGround exists to measure that, on real captures, in the only way that isn’t self-reported: ask for the thought back before showing it to you.

1. 01

   Capture

   One field. No categories, no tags, no mood picker, no required anything. Friction at intake is the thing that kills a thought before it becomes useful.

2. 02

   Structure, invisibly

   A server-side model classifies and tags the entry after the fact. The person who wrote it never sees a taxonomy and never makes a filing decision.

3. 03

   Resurface

   The system chooses a past entry and pushes it. You never have to remember to go back and look, because remembering to look is exactly the capability under test.

4. 04

   Cold recall

   You get a category and a rough date. Nothing else. You reproduce what you think you wrote — and only then does the original appear.

5. 05

   The two questions

   Does this still hold up? And did it convey the important part, or did something get cropped? Two answers, a few seconds, logged against the entry.

Steps one through four are plumbing. Step five is the instrument — and it is the part no other capture tool ships, because it is the part that can return an unflattering answer.

In Dota 2, the high ground is the elevated terrain holding the last line of towers before the ancient. You cannot see onto it from below. Fights lost beneath it cost you nothing permanent. Lose the high ground and you lose the game.

A thought behaves the same way. One that never reaches defensible ground doesn’t become anything — it was had, and then it was gone, and nothing was built on it. HighGround is the terrain a thought has to reach to survive long enough to be built on. Not a trophy: a vantage.

These are documented build constraints, not stylistic preference. Each one has survived a specific temptation to break it.

• Formless at the point of entry

  Structure is the system’s job, applied afterward, never surfaced as a choice the person has to make.

• No gamification, anywhere

  No streaks, points, badges, levels or progress bars. A product that is only useful because it is gamified is not useful.

• A missed recall is never a failure

  It is diagnostic. No score, no red, no corrective tone. The moment this reads as a test you can fail, people start logging for the grade.

• The system initiates

  Resurfacing is pushed on a schedule the person doesn’t manage. Anything requiring them to remember has already failed its own premise.

HighGround owns mechanics — capture, storage, resurfacing. It deliberately does not own meaning. That separation is the architecture, not an accident of scope.

Each consumes HighGround’s output and supplies its own why. Only HighGround is in build — the others are deliberately parked until the substrate is proven, because a capture layer with nothing real flowing through it cannot tell you whether any of them are worth building.

Stated plainly, because a research instrument that overstates its own readings is not an instrument.

One user: the person building it. No waitlist, no beta list, no usage claims. The instrument has to produce a reading on its own author before it is worth pointing at anyone else.