I've had people make some weird inferences about my skills before (judging from the mystifying endorsements people have given me on LinkedIn), but assassination planning is still at the "I'm interested, but waiting for Coursera to offer a course before I actually get into it" stage.

Here's what we need to do to carry out this killing using the car plan. We need to hack the car to accept some kind of remote control. This will require hacking at least the steering, the engine, and the brakes. We need all three because to ensure killing. If he can control one of them, he has a good chance of making the accident non-fatal.

After we've got the car hacked and can remotely control it, we need to actually cause the accident. There are a couple of approaches here. One is to follow him around in our control vehicle into an opportunity arises. The other is to wait at some place that we know he is likely to go and that is a good place to make a fatal accident happen.

After we've cause an accident, our problem with him is not yet solved. We still have to worry that someone will decide his death was not an accident, examine the car, find that it was tampered with, and maybe tie that to us. Even if it only leads to people we hired and we were competent enough to leave nothing that links them to us, we lose those contractors for future work.

So, what can go wrong with this plan?

First, we've got to do the car hacking. Even if steering, engine, and braking are all "fly by wire" on his car, they are likely to be separate systems. We are going to have to rig all three of them for remote control. We'll probably need to install some hardware for the remote control.

Any suspicious contact with the target before killing him is risky, and right off the bat this plan involves complicated fucking around inside his car.

If we do get his car rigged, then we have to actually take over and cause the accident. The "follow him around hoping for an opportunity" approach is not good. He could go weeks where he only drives in high traffic areas, where we won't get an opportunity to get him up to a speed that ensures a high chance of fatality. The longer we follow him, the more chance we'll be spotted, or that he'll take his car in for service and someone will find our remote control rig.

Our chances are better if we know some place he regularly drives where conditions are better for an accident. Let's assume that is what actually happened--they were waiting for him at the place the "accident" actually happened, and consider how risky that was.

Accelerating up to high speed and nailing a tree seems pretty straightforward. However, it's not as easy is it seems. If you check the photos of the crash scene, on some it looks like the trees are tightly packed, so it would be easy to nail one. However, those photos are looking down the street and mae the trees look more densly packed then they actually are. Find a photo looking from the side, and you can see the trees are much farther apart.

If there is any noticeable lag in the remote control, it will actually be pretty hard to aim at a particular tree. Best you'll be able to do is veer toward the trees, and hope that when the car crosses the curb that doesn't turn it enough to send it between the trees.

Even if you do hit a tree, it will be hard to ensure it is a good solid fatal hit, as opposed to some kind of off center hit that doesn't stop the car but just sends it spinning.

But suppose we get lucky, and actually get a good fatal hit. We've arranged the accident in a place where it will be immediately noticied. We'll have civilians on the scene quickly--before the fire has died down enough to let us get in and make sure any evidence that we want destroyed was actually destroyed. Police and fire and ambulance services will arrive quickly.

This is a Wile E. Coyote plan.

In general, the more steps a plan has that must succeed for the plan to succeed, the lower the reliability of the plan.

Shooting him at an anonymous meeting has two steps that must succeed. (1) Call him from a burner phone with a fake anonymous whistleblower story and arrange a meeting. (2) Shoot him when he shows up for the meeting. This either works, or it fails at step 1 leaving nothing behind to tie it to us and nothing to raise suspicion.

I asked because it seems to me that the only way to know how reliable particular assassination means are is to have tried them out, with the full resources of a wealthy country's intelligence organization at one's disposal.

There are no doubt trade-offs with either approach. A car accident produces a strong explanatory narrative that can be used to squelch "conspiracy theories". Merely disappearing (as the persistence of POW/MIA organizations proves) provides no such benefit.

So, from a consequences perspective, a successful "apparent accident" hit is better than an obviously foul-play one, or even a disappearance.

The trade-off, as you have thoughtfully articulated, is one of practicalities. It's not particularly difficult to imagine that a wealthy espionage org would have developed reliable methods for making hits that look like accidents (for narrative squelching reasons). We know, now, from the declassified histories of these organizations how elaborate and sophisticated their techniques and tools can be. Might they have some "turn-key", field-refined method of taking remote control of a vehicle? If you were running such an organization, would you want such a capability? Given what we already know about the attack surface of vehicle control systems, if you had a large budget and a team of smart and motivated engineers, do you think you could develop this capability in a way that would work well enough for most modern cars?

In the Hastings case, even serious injury might have been an adequate outcome (perhaps this was an time-sensitive situation?).

I agree: certainly, for a one-off offing, the practicalities would be a major execution obstacle, making the scenario highly implausible. But killing people is one of the things these organizations do for a living -- the practicalities might well have been dealt with years ago. If they have done so (as they should if they're half-competent) then the highest expected value method changes considerably.