Stipulating up front: this is out of my area of expertise.
Although the current problems with speculative attacks clearly do not fall into the category of deliberate, manufacturer or government instigated vulnerabilities, I believe AMD could earn a lot of friends (and end-user customers) by formally committing to produce exploit-proof processors and processing environments. The Intel Management Engine clearly demonstrates Intel's design priorities, to the point where I question whether I'll ever buy another Intel based system.
Re: Meltdown & Spectre
Unfortunately, the last time I wrote any assembler code it was for a Z-80, back in '84, so it would be tough for me to do this myself. However, while I believe its essential that manufacturers provide measures to prevent speculative attacks ASAP, given the processing speed penalties, if AMD engineers could come up with a set of benign exploit "tools", capable of mimicking Meltdown and/or Spectre, I believe that approach could enable end-users to run 'attack detector daemons' (ADD's) which would profile their own operations on a clean system. Subsequently, run the ADD's in 'detect mode', which would include comparing the live performance with profiles. Presumably, two attackers running simultaneously might interfere with one another, in a way which could be detected by the ADD. Although I don't fully understand what qualifies as a 'gadget', perhaps the ADD might be accompanied by some 'gadget trojans', able to identify the presence of a real attacker.
Finally, if the above system seems plausible to those who understand the internals better than I do, make it configurable/customizable and 'embedable' (in ordinary system processes), so real attackers would have difficulty determining whether they're being watched. In many cases, I'd much rather have the ability to detect an intruder (even if detection is not 100% guaranteed), than pay the performance penalty.
