The article is a decent summary of the state of SIGINT and the reasoning behind th epressure for "voluntary" backdoors, but this bit isn't really accurate:
Clipper – an encryption chipset with a US-government-accessible backdoor backed by the US National Security Agency (NSA) – foundered on the stubborn resistance of one man in his spare room, Phil Zimmermann, and a modest home-brew application, PGP
Clipper "foundered" at least as much for two other reasons, one technical and one political. Politically, it was simply unpopular; the proponents did a poor job of marketing it, and sentiment was against it from the start, in both government and business.
On the technical side, while Zimmerman showed that competing, non-backdoored encryption could easily be made available, a more devastating blow came from Matt Blaze's crack of the LEAF checksum. That doomed Clipper's key escrow system, which removed the whole justification for the thing. If you want to lay the death of Clipper at the feet of any single person, I think Blaze is the one.
More broadly, you can lump Zimmerman and Blaze in with the surge of interest in computer-assisted crypto that developed in the '70s, '80s, and '90s, after Diffie-Hellman(-Merkle) and RSA were published (1976 and 1978, respectively) and DES was standardized (1977). By the time the Cypherpunks mailing list was established in 1992, the genie was clearly out of the bottle. The publication of ARCFOUR1 in '94 meant that anyone could have a decent stream cipher - RC4 is so simple you could memorize the algorithm (or print it on a t-shirt, etc).2
When Clipper came out in 1993, it was effectively dead on arrival.
1Alleged RC4, the public reverse-engineered implementation of RC4, which was a trade secret. Everyone was pretty sure ARCFOUR was identical to real RC4, and it certainly did an equivalent job.
2RC4 is considered unsafe now, and there are public attacks against correlations in the key stream that mean it can be broken with ordinary resources unless countermeasures are employed. It's not publicly known how much of that NSA knew in '94, or what resources they could devote to breaking RC4. It's likely that widespread use of RC4 would have significantly restricted their ability to do mass real-time or near-time data inspection for several years, though.