Replacing the layers of patchwork technology that have accreted onto each other, with clean, direct architecture (reminiscent of earlier days) and more sensible abstractions. Reliability and security baked in from the start. A focus on clarity of intent - not just through documentation (for which the quality bar would be set exceptionally high) but also how the code is crafted and structured. Tending toward mental models that are friendly for humans to grasp and reason about. Basically, designed according to the same opinionated, high standards that I bring to my own development work.
Then one of the things I'd use that stack for would be to introduce a legitimate competitor to Apple and Android smartphones.
That disposes with the harmful games the incumbents play when it comes to user control, privacy, advertising, and the many dark patterns engineered (by them and publishers on their platform) to drive their revenue ecosystem. There are more humane and decent means to carry on the business of building a successful and modestly profitable platform (and one of the great constraints of unlimited resources is I wouldn't feel compelled to optimize along the $ axis). Fostering a culture that genuinely cares about its users, the prime guiding principle would be to remain a worthy custodian of their trust, by making decisions that consistently put user interests first. Maybe reminiscent of how early Google wanted to be when they still tried to practice the 'dont be evil' mantra.
Next I'd work on the problem of identity on the web (and in the digital world), by pivoting it into one of reputation. It wouldn't rely on government issued ID, age checks, KYC partners, etc. Rather you'd spin up a new identity (with a choice whether or not it remains anonymous) and work to build up reputation for it in a manner that builds value along a metric you care about - eg. social media engagement, forum posts, GitHub contributions, Pokeman collected, whatever. The key game theory behind this is it's something the corresponding user cares about. That way they are motivated to preserve their reputation and not undertake offenses that would 'burn' it. The solution here isn't as clear a vision as my other initiatives, but the idea is to devise a primitive that can then be used to solve all sorts of other problems when strangers need to interact digitally (eg. I can trust this vendor because they have a history of delivering what's promised and honouring returns, I can trust the information provided by this human as they're recognized as an expert in their field of Pokemon hunting, etc). Maybe tack on some kind of zero-knowledge, six-degrees vouching mechanic as well.
Finally, I'd work to improve how civilization practices politics.
I was lucky to be born to a time where I got to witness the birth of the Internet and see how it impacted and evolved nearly every industry. How we communicate. How we collaborate. How we trade/shop/bank. Yet democracy is still run in a similar fashion as it was when first invented in the era of the printing press. Oh what the founding fathers might have envisioned in a world where every inhabitant has a rich, two-way channel directly to their institutions, their representive politicians, and each other.
I'd start by creating a space on the web committed to gathering solid background knowledge and opinions on issues (from all sides) and surfacing the best and most important facets in an easy to digest fashion. Something like a cross between Wikipedia and StackOverflow, but for political/civil issues. We'd take on all issues - from the mundane, to the most divisive. With a means to promote discourse that relies on transparently sourced facts, and a moderate emphasis given to subject matter experts (based on their reputation in a specific domain - see above). The aim is for the most informed or insightful arguments to float to the top, and the vitriol to sift to the bottom. So average Joe can pull up an issue, digest the top opposing viewpoints in bite sized chunks in a matter of minutes (after all, these are modern day attention spans we need to cope with!), and dive deeper into the rationale, facts and supporting data behind each should they wish to become more informed.
Then have a mechanism where you can "tune in" to matters you care about and even cast votes or otherwise influence the outcome. Eg. Get notified when local matters come up, or matters you've marked an interest in (for me, that would be topics like right to repair, copyright, privacy, etc).
Provide tooling to decision-makers, eg. to run an ad-hoc poll, pull in data from outside sources, etc. A simple example use case would be collecting citizen pothole reports, visualizing them on a map (with color-coded severity, photos, etc) and prioritizing which ones to fix - with an incentive for me to vote not just based on which roads I use personally, but also with empathy for sensible arguments like "that one poor Jim reported is in really bad shape" or "this is next to a school bus stop and could result in a kid getting injured". Don't care about the minutia of roadwork? That's ok, leave it delegated to your civil servant and those fellow citizens who are passionate about that topic.
Some of this has materialized already in various forms - eg. Twitter can be a powerful tool to call out and correct corporate bad behavior - but we need to figure out incentive mechanisms that put more emphasis on empathy in interactions and contribution to effective outcomes, than to superficial metrics like likes and fake popularity.
I don't have all the answers, but would love to gather like-minded people (in terms of ethics and intrinsic motivation, not political views!) with unlimited resources to work on these challenges.
Oh, and one last bonus - if those unlimited resources included a greatly extended lifetime in which to operate - I'd study advanced physics and work on cracking warp drive to spread our civilization to new frontiers. That'd be a childhood dream come true.