Notes on drawing a chip that sleeps

When you sit down to draw something that does not exist yet — a hippocampal-memristive chip that replays its memories while it idles — you have a small problem and a large one.

The small problem is craft. You need a grid, a colour palette, a font system, an icon language. Those are decisions you can make in an afternoon if you have done them before.

The large problem is honesty. The chip in my head is not finished. It has not been fabricated. Its energy numbers are projections, not measurements. The replay dynamics are a hypothesis. You can render a hypothesis as a beautiful poster — neon traces, glossy renders, drop shadows — and it will lie to the viewer. It will feel done.

I did not want it to feel done. I wanted it to feel like a working drawing on the wall of a lab: confident enough to take seriously, honest enough that you can tell which lines are conjecture.

Draft three: the poster trap

By draft three, I had a beautiful poster. Cyan traces on midnight indigo. A precise grid of memristor cells. A radial halo of "energy efficiency" suggesting orders-of-magnitude wins. It looked like something a marketing team would ship.

It was also useless. A reader looking at it could not tell:

1. Which numbers are measured versus projected.
2. Where the architecture is novel versus where it borrows from existing literature.
3. What the limits of the design are — and every honest piece of engineering has limits.

The poster hid the engineering. So I burned it.

Draft seven: the cartouche

The breakthrough came when I borrowed a device from old scientific illustration — the cartouche, a small bordered legend at the corner of a plate that names the piece, the plate number, the date, and the author.

Once I added a cartouche, the rest of the plate had permission to be honest. The plate could say "this is plate I of the MEMPHIS proposal, drawn 2026-04, by T. Singh, projections marked in italic" — and suddenly every italic number on the chart was understood, by the reader, as a claim and not a fact.

The cartouche is the most important typographic element on the plate. It is the thing that lets the rest of the drawing be what it is.

Draft nine: the sleep state

Up to draft nine, the chip was always on. Cells lit up. Spikes rotated. It looked alive, in the cheerful poster-y way that "alive" usually means.

But the whole point of the MEMPHIS architecture is that it sleeps. Replay happens during a low-activity phase. The interesting energy story — the four-orders-of-magnitude reduction — is mostly a sleep story.

So I built a two-phase clock. Awake, the cells flicker sparsely; spikes rotate every 160 milliseconds; the trace looks like a working brain. Sleep, the cells dim, a single radial wave sweeps outward from the centre, and the replay ring overlays the grid. You can watch the chip transition.

That is the moment the plate stopped feeling like a poster. The viewer is now watching the architecture do its thing, not just be its thing.

Draft eleven: what got dropped

The final draft is significantly less ornate than draft seven. Things that went:

• The glossy chip housing. (Distracting. The chip is conceptual; it does not need a render.)
• The energy halo. (Replaced with a log-scale bar chart on a separate plate, where the comparison can be read precisely.)
• Three of the five colour accents. (The plate now uses indigo, cream, and copper. Period.)
• The "TM" on the title. (Not yet. Possibly never.)

Each of those removals made the plate more honest. The fewer ornaments a working drawing has, the more its claims can carry weight.

The lesson, such as it is

If you are drawing something that does not exist yet, your job as the illustrator is not to make it look real. Your job is to make the shape of the claim legible.

A poster makes the claim invisible. A working drawing makes the claim visible — and lets the viewer judge it.

That is the whole craft. Eleven drafts to get there.