Reckoning

“It doesn’t work” and “it doesn’t work right now” are different findings, and the second is only worth saying if you can name which constraints are dated and show your working. Otherwise it is the same wishful thinking as the first, wearing better clothes.

Reckoning is that working, done properly.

Why it’s called Reckoning

Dead reckoning is navigation by extrapolating forward from a known fix, using a known speed and heading. That is exactly the method: take a measured present, apply measured rates, project.

It also carries the right warning. Dead reckoning accumulates error, which is why a navigator checks it against a real fix whenever one is available. Here that fix is the past — the method is run backwards to a year we can verify before it is trusted forwards. The same design in 1996 comes out at 27.9 kg with a battery that would have cost about $22,000, and four of its eight subsystems could not be bought at all. The method retrodicts, so it has earned a little trust going the other way.

The four kinds of constraint

Sorting a design into these is most of the work, because in a mass table they all look identical:

Physics floor — never moves. Water’s heat capacity, conservation of energy, what a human can carry all day.

Technology-limited — moves with time, on curves you can cite.

Market-limited — moves with volume, not time. Nobody makes the part at your scale because nobody has needed one. This is the class that had no name, and it is the one that matters: it is moved by somebody selling enough units to justify tooling, which makes a product launch the mechanism that fixes it rather than something incidental to it.

Cost-limited — falls fastest of all, and affects price rather than mass. Conflating cost and mass is the usual error; a design can become affordable long before it becomes light.

Two forward paths, never blended

Iterated is today’s technology on its own curve — extrapolation. Purpose-built is a market bet, contingent on one commercial decision nobody has made yet.

Blending them is how extrapolation becomes marketing. Averaged into a single “the future will be better” line, the two answers become something that looks reachable by waiting. Usually it isn’t: in the worked example, thirty more years of everything quietly improving still lands above the comfortable carry band. The curves don’t finish the design. Volume does.

The honesty note that cuts against the optimism

Lithium-ion specific energy rose 4.3%/yr from 1991 to 2018, and about 1.2%/yr across mainstream cells since. Extrapolating the long-run rate to today overshoots the best available cell by 27%.

Curves flatten, and a forecast that doesn’t allow for that is a wish.

What it isn’t

Not a forecast. Every rate is cited, every projection states its conditions, and anything without a measured history — solid-state cells, lithium-sulfur — is deliberately left off the charts. Drawing a line through a single announced point is illustration, not extrapolation.

Free, MIT licensed, one HTML file, zero network requests. Works alone, or reads a session exported from Prebuild and recomputes against your own design.