As with all security, threat models are key. Is your threat model a state actor trying to reveal something about you? Is it a prosecutor or investigator that is particularly vicious? Is it someone you're in a personal relationship with? Is it an unknown adversary with unknown capabilities? And how long do you need to have your strong OPSEC? And so on. With the right threat model and timeline laid out, having strong OPSEC is not impossible. But it's still quite hard.
All of that being said, I would not classify any aspect of Levandowski's behaviour as being passable as OPSEC.
Google will never find out what you do on your home PC, especially using TOR /some other engine (Google displays the captcha often on TOR). Even if they did, it would be illegal and /or super secret and they'd never reveal it in court. Now, if shady means that NSA is after you, not so sure.
However, if you want to steal from Google and use their computer to search their search engine from their office...I guess you deserve to get caught. Let's not forget that Google has a right to safeguard their trade secrets and when you accept the job, you agree to their terms.
yes, one of many ways to get 'anonymous' internet, another is cash buying old cable modem, even unprovisioned they still log into cmts and give you dns access(hint hint)
>and download a Tor browser
no, you go swap meet/yard sale/goodwill/craigslist(using library computer) and cash buy old used($100) working laptop with preinstalled windows
Thankfully that’s not how law works. Also, I use Tor everyday, for totally legal things I want to keep private.
Once an attacker knows or has guessed at your identity, they can focus their investigation and make it much easier to find admissible evidence against you. It's best to keep them from ever thinking of you in the first place, if you can. That means not doing anything visibly suspicious.
Although ironically I think in this case your point actually still stands.