I'm sure that's true, but how important is that really for a set of crappy plastic speakers?

To get around this, Google put in the TI TAS5825M smart audio amp. By measuring the speaker parameters through V/I measurement and a model, it drives the speaker in a closed loop way with far more power than it would actually be able to handle nornally to compensate for the resistance from the enclosure air pressure, and throttles to maintain the coil at a safe temperature. The chip also does DSP to compress the audio signal, cutting the peaks off the bass as needed when the volume is turned up so volume is maintained at the cost of bass.

One way to explore could be to just feed I2S audio from an I2S ADC i.e. PCM1808 to the digital input of the amplifier. The processing is internal to the amp so theoretically you won't lose the tuning. However this may turn out to be a relatively annoying reverse engineering project with fine magnet wire involved.

Note: I2S is different from I2C - the amp will likely have both. You will likely need to keep the original system around to program the amp over I2C (or capture the transaction and replay it) - otherwise you will likely get no audio.

The "raw" audio performance of this device (just an amplifier connected directly to the internal speaker and dsp on the computer) is impressive, kicking out bass down to 40Hz. It will, however, not last long like that. Reports online are that these blow speakers easily even when used with the default amplifier.

I would recommend that if 3.5mm input is desired, to replace them altogether with the IK Multimedia iLoud Micro Monitors. These have sound quality just as good as the Google at similar size, with the same DSP tricks, but have regular inputs and no smart features.

You're right that iLoud Micros sound similar, they're 3x the price (The Nest Audios were sold at $50/each on sale). Definitely worth it, I just like tinkering with things.

You don't even need a lot of "lab equipment dollars", measuring the basics can be done with ~100€ calibrated USB mic. As a rule of thumb, you cannot develop a good speaker without measuring, unless you have advanced modelling tools and experience to use them correctly, in which case the measurements will mostly match the simulations.

Says who? You should really acquaint yourself with current research, starting with Toole's infamous "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms".

Anyone who has read about or experienced instruments and music halls that were made before the commodification of electricity?

My point was more along the lines of „a layman reading this might draw the conclusion that this is the proper way to solve this problem, because the text is written with an engineering mindset and demonstrates a certain degree of sophistication“, and I wanted to caution people that this is not the case.

Turns out you can, they just did and are happy with the results.

Which means most people can’t tell the difference. This is mildly upsetting to the people who can.