When thinking about the dangers of a world with strong AI, a natural impulse is to try to weaken AI systems, so we can remain in control. An alternative strategy would be to strengthen our defenses, aka "defense accelerationism". For instance, we might harden our cooperative infrastructure via computer security. Or we might leverage neuroscience to improve our cognition and with that our ability to compete or merge with AI systems. Let's explore at a few d/acc paths, promising approaches, why they wouldn't work, and what it might take to make them work. This will be a shortish talk to leave plenty of time for open discussion.