Jason Cohen's attempt to commit heresy against the Church of the Lean Startup isn't going to get him excommunicated.  I think that he nailed his thesis to the wrong church door.

Apple doesn't ask customers what they want

Lean Startups don't ask customers what they want.  They validate their own vision by testing it against reality as soon as possible.

You're misinterpreting the 80/20 rule

A startup should never try out-Excel Microsoft.  It should pick one important feature (online collaboration perhaps) where it thinks a new market exists and implement a minimal spreadsheet app that built around this feature.

Mock-ups are faster than code iterations, without some of the drawbacks

I agree in most cases.  This technique is often used by lean startups.

Releasing too early can ruin your reputation

Unless you are already famous, no one will care when you make your first release (as a startup).  If you are famous, then you can do a private beta and won't have any trouble getting testers.

Also note that a release is not a marketing launch.

Ignoring architecture creates waste

Your guesses about optimal architecture are likely wildly off.  If you measure your bottlenecks then you can make just in time scalability changes that address actual issues rather than imagined issues.

The worst thing you can do is built an unnecessary feature.

Lean startups use their best judgement (based on learning from previous iterations) to select features.  They then build these features, test them against real users and remove them if they don't add measurable value.  A properly functioning lean startup allows you to try lots of different features to see which ones work the best.

Customers are notoriously bad at providing feedback

I agree.  That is why you don't ask them what to build.  When interviewing them, you try to get them to envision what their life will be like once they have the feature and try to perceive an emotional response to your vision.  After you build the feature, you measure to see how it changes their behavior.