I have it on good authority that Murphy didn't actually write Murphy's Law, it was actually another fella with the same name!

After peak the production rate of a field is very sensitive to how its been developed produced and the geology of the field itself. Technical factors come into play to determine the post peak production profile.  Almost all the offshore fields show steep declines post peak because of the way they are produced. For cpu's and Moore's law heat and static discharge have ben the major factors on the real increase in computing power the last few years.

These technical factors certainly sometimes result in higher production rates then would be predicted with a smooth curve generally in exchange for a drop in production later.

In general just looking at the graphs most fields tend to show a initial slow decrease in production rates post peak for several years say 3-10 followed by rapid drops as technical factors come into play. Look at the field production profiles for Texas and the north sea.

Here is a article on the North Sea.

http://www.energybulletin.net/17262.html

The decline rates post peak are all over the map.

I agree I'm guessing but it would sure be nice to have someone come up with a reasonable explanation why the post peak decline rates won't be steep. I can come up with lots of reasons to expect steep declines and so far not single reason to expect them not to be steep. And yes I take  political and engineering issues into consideration because they are relevant.

How can we ensure that we are not headed for serious problems ?

Treat oil as a world resource not a national or resource open the books and do independent analysis. Find out the truth then with this information work out a way to power down in a dignified manner. Transparency regardless of if the news is good or bad is the only answer.

Are we going to do this ?
Probably not.

applications that need a new chip are becoming fewer and fewer.

Not so sure!
This may be a chicken and egg problem.
Some applications which would require orders of magnitude more processing power are put on a back burner and not investigated further.
For instance I know of an on-the-fly compression/decompression algorithm which is limited to a few dozen kb/s with current chips, a 100 or 1000 speed up factor would help.
I suspect there are many, many such hidden nuggets.

Even if the demand for faster chips would slow down more transistors per chip gives more compacty and cheaper systems and flash memory is very practical. And there is a tradeoff between speed and power consumption, if you have the speed you need the development can give you lower power consumption.