vice that had been F.D.A.-approved.
“He dropped a lot of reality on every-
body about the level of engineering we’d
have to do,” Michelle Easter, a mecha-
tronics engineer who usually works on
actuators for spacecraft, said. The doc-
tor explained what a ventilator was: he
told them that, if a patient initiates an
inhalation, the ventilator must notice;
that ventilated air must be delivered at
body temperature; that it should be hu-
midified; that it has to provide high
concentrations of oxygen. The engi-
neers, working from first principles,
peppered him with questions about
pressures, volumes, and rates of change.
They decided on a new goal: build the
simplest possible easy-to-manufacture
ventilator, made from readily supplied
parts, that was capable of treating all
but the most complex cases of COVID-19,
and get it approved by the F.D.A.
The project unfolded in typical NASA
fashion—the gathering of “functional
requirements,” the building of increas-
ingly sophisticated prototypes, the hold-
ing of team-wide engineering reviews—
except over a period of weeks, not years.
The team had to adjust to the fact that
its design would be produced at vol-
ume. “We’re used to having one of the
thing, not thousands of the thing,” Eas-
ter said. The project had attracted more
than a hundred participants, many of
whom had never collaborated before.
For almost forty days straight, they
worked from sunup to sundown—bru-
tal days that were a relief, in their way,
from the ennui of lockdown. J.P.L.’s
mountain campus, in Pasadena, was
mostly empty; on breaks, at picnic
tables outside the lab, the engineers
watched families of deer graze among
wildflowers.
J.P.L. worked up two designs in par-
allel. One is more portable, and uses an
air compressor; it’s in the final stages
of testing. The other, like the Vermonti-
lator, accepts air from a wall outlet.
VITAL, as the device is known, operates
in a unique “volume targeted, pressure
limited, time limited” mode, invented
at J.P.L. A clinician can set tidal vol-
ume, inspiration-expiration ratio, PEEP,
and breaths per minute. Although it
doesn’t have a CO 2 sensor or a touch
screen, it replicates many of the features
of more sophisticated, customizable
ventilators; a team of artists and illus-
trators helped design its faceplate. It is
expected to cost between one and two
thousand dollars, and its approximately
three hundred parts have been carefully
chosen to avoid siphoning supplies from
medical-grade ventilator manufactur-
ers. VITAL was tested at the Icahn School
of Medicine at Mount Sinai, in New
York, on a specialized lung simulator,
and has received an Emergency Use
Authorization from the F.D.A.; J.P.L.
has begun looking for manufacturers.I
n software, “feature creep” is the
process by which an initially simple
program, through the accretion of en-
hancements, becomes gargantuan, slow,
and hard to use. Microsoft Word can
take a long time to load in 2020, even
though today’s computers are incredi-
bly powerful. The program’s core func-
tions haven’t changed since 1983, but
there are many new ones, and manyfeatures add complexity beyond them-
selves. If you want your program to have
both an Equations Editor and Track
Changes, then you must teach it to track
changes to equations. Even the Undo
function is a complex subprogram wor-
thy of several full-time engineers: it
must be able to undo not just typos but
table reformattings, image resizings, and
comments that have been added during
a review session.
One wouldn’t want to accuse a ven-
tilator of feature creep, since each new
feature has the potential to save lives.
But a ventilator’s complexity also ex-
pands nonlinearly as the number of
parts, sensors, and functions grows. The
problem is especially acute because
medical devices must clear a high set
of regulatory hurdles. As devices grow
more capable, and more complex and
expensive, they require more careful
regulation. This dynamic, which is“ You can just leave it in that trash can.”