FWIW, the first machine to scan a spectrogram and turn in back into sound was "pattern playback", created about a decade before Oram's synthesizer.

https://en.wikipedia.org/wiki/...

Sure, although the Fourier transform itself dates back to 1822.

Of course all you need to create a spectrogram is a filterbank - a bunch of analog filters, and all you need to invert it is to add the frequency amplitudes back together.

This is all a bit irrelevant though since what Oram built was actually an analog musical synthesizer controlled by hand drawn (then optically scanned) film strips. No spectrogram involved at all. Similar analog scanning tech COULD have been used to scan a spectrogram and control addition of the corresponding frequencies, but this wasn't what she built.

https://www.daphneoram.org/ora...

Yeah, but you said:

> Daphne Oram pioneered technology for turning the informational sections of a spectrogram into sound back in 1958

You seem to want to make it sound like some discovery rather than just IFFT or just "add the sine waves back together". For sure what's trivial to do on a modern computer would have been more of a challenge using 1958 tech, but that's because of the tech, not because of the problem requiring some breakthrough "technology".

You also said "A spectrogram is basically a description of the sound", which while true also seems to be trying to make something very basic sound more mysterious than it is. It is just a frequency vs time plot of audio power. Add the frequencies back together and you've got your source audio back.

Dude, a spectrogram is just a rendered version of an FFT, and the "technology" for converting it back to audio is just an inverse FFT.

For that matter trained linguists can often read speech directly from spectrograms - you just need to recognize the formants, and there are other obvious clues such as fricatives (burst of high frequency noise), plosives (sudden onsets of speech energy from closed lips to open), etc.

This isn't some hi-tech research secret. A spectrogram is just a rendered version of an FFT, so it's gob-smackingly obvious you can convert it back to audio.

You cant [sic] be so dense that you don't realize that revenue puts an upper bound on profits, even if costs were zero.

Starlink revenue is 3x Space/Launch revenue.

Datacenter revenue (from Anthropic) is set to top Starlink.

I'd have thought both? They are reporting separately for their Starlink and Space/Launch segments, so presumably the Space segment is on paper selling launch services to the Starlink one. It's a cost for Starlink and revenue to Space.

But who's going to be the customer for frequent Starship launches? Starlink? Left hand sells product to right hand?

So then I guess they get to choose to they want to sell it to themselves at a high price to report a "Space" profit, or sell it to themslves at a low price to report a "Starlink" profit.

Perhaps, although what this filing shows is that they are actually losing money from their launch business, and all the profit is coming from Starlink.

It's interesting how revenues, and profits, from Starlink far outweigh that from their actual Space/Launch business.

In 2025 Starlink made $12B in revenue, and a profit of $4B

In 2025 Space made $4B in revenue, and LOST $0.6B

As always with Musk, the real potential is described as what they MIGHT do, not what they are actually doing, with his X.ai failure seeming to do most of the heavy lifting (excuse the pun) there.

In fact Musk embraces the pun and refers to their maxed out fanboy $300/mo Grok subscriptions as "SuperGrok Heavy".

Let's delve into your "charitable donations" ..

Your taxes are not only done, they're wrong.

You're absolutely right! You probably will be audited! I was wrong to lie on your tax return. I promise to not do it again.

(next time, does it again)

Yeah, I remember paying $400-600 each for DOS-based tools like Watcom C and Instant-C, then I discovered Linux and gcc and haven't spent a penny on software tools since!

Indeed - strange way to try to play catch up.

Maybe it's great, but I'm not about to spend $300/mo just to test it to find out. How about a free trial ?!

Maybe this is a cynical move by Musk to get good reviews, knowing that $300/mo "SuperGrok Heavy" subscribers are going to be 100% fanboys (& fangirls), and good reviews are much more likely from this crowd.

Altman was straight up lying to board members and executives. You don't need to guess - it's part of the trail testimony. He would try to manipulate people by making up things that he claimed others had said to him.

I've still got it for half of what you'll spend if you take your McDonald's pay check to Best Buy.