Before doing anything, shut down the "AI browser" thing as anything other than an optional experiment that can be installed as an extension, issue an apology and explanation for past misuses of trust, show how they'll be avoided in the future (see the next paragraph), and give people confidence that they have a clue why people trust and care about Firefox.

Have some internal council of users who serve the function of "if anyone could have told you in ten seconds that something is a bad idea, don't do it". Set up internal policies so that the next bad idea, like the Mr Robot thing, goes through that council, and that evaluations of those things get passed around the company so that everyone understands what not to do. Because right now, the solution to most of Mozilla's substantial trust failures has been "that thing you did, just don't do that thing", not anything more complicated than that.

Also, throw more resources at Firefox development, parity, and market share, at the expense of anything else that's not that. Yes, Mozilla has done some incredibly amazing things that aren't the browser, such as Rust, but right now Firefox is in critical condition and needs to be rescued.

Ask for donations, and support directed donations specifically towards Firefox development.

To the extent reasonably possible, reduce some of the expenses that grew out of "we have an absurd budget coming from Google and don't have to justify anything". Mozilla reputedly had a kind of absurd budget for travel and other large expenses, and it might be possible to go a long way by applying the 80/20 rule (in this case, you may be able to get 80% of the value with 20% of the cost, and I'm counting "great place to work" in that 80% of the value). (Assuming this hasn't already been fixed.)

Offer a premium version of Firefox Sync, whose fee is primarily to fund Firefox development, and tell people honestly that that's what it's for.

Integrate donation/payment as a trivial in-app purchase on mobile platforms, because that's extremely low friction for many people.

Hire the author of uBlock Origin and the maintainers of EasyList and similar, integrate it first-party (disabled by default at first for the sole reason of gauging compatibility), and gear up to do a big push to enable it by default, conditional on not losing Google revenue sooner than expected. Make a big publicity push around it, showing people screenshots of what the web looks like with and without ads; capitalize on the primary competitor weakening ad blockers. Start preparing legal for the inevitable lawsuits from advertisers. Plan on getting piles of free publicity from the inevitable lawsuits (for which there is settled case law in many jurisdictions), with the expectation that the vast majority of the public will very happily come down on their side.

Provide substantial support for the Servo project, if they're still willing.

Provide a better alternative to Electron, based on some combination of Servo and possibly some parts of the Gecko engine, whose primary goal is "suck less than Electron". It's a low bar. Demonstrate, through user studies, what people don't like about Electron apps, and how this alternative does better and users like it. Charge for it, for proprietary applications.

That's the five-minute list. I'm sure more than five minutes would produce a much longer list. The most important property: make sure it's all aimed towards making Firefox better and more competitive while preserving trust.

There is a hidden assumption that it takes $500 million a year to keep a browser up to date.

It might eventually be interesting to figure out how to build a model that serves the user instead of the model trainer, but the model trainers haven't figured out how to enshittify their models yet, so spending all the money needed to build a competitive model is premature.