I (along with a few others) started a project about 20 years ago, inspired by things like the STEPS project at VPRI, HyperCard, and similar things, to build a new kind of desktop environment. We wanted to move away from applications as an abstraction and provide documents and tools for manipulating documents. We wanted something that respected Raskin’s First Law, so there would be no save button. Documents (including their editing history) were automatically persisted, the user abstractions were not save (losing data should be explicit, retaining it should not), but ‘tag’ and ‘export’. The former associates some metadata with a revision so that you can find it later. The latter converts the document into some form for external interoperability (e.g. in a video editing app, the internal representation would be raw footage and transforms but export might give you an MPEG-4 video file).

This required a uniform object model that different components could interoperate with it and end-user programming tools that made it easy to write custom tools that operated over the data. This is where I started down a rabbit hole that ended with my building hardware. Current MMU-based isolation is inadequate for this. I wanted users to be able to write code in safe (and easy!) languages that could use existing libraries. Unlike STEPS, I have no interest in rewriting libavcodec in a DSL so that it fits in my 20KLoC budget for the system, I want to just use it without worrying about whether it’s memory safe and, if I use it so wrongly that it crashes, get a recoverable error (restarting the library instance is fine). I joined a project that has the skeleton of the things I wanted and evolved it to the point where it did everything I needed: I could write managed-language code with object-granularity interop with C/C++ and have strong guarantees that the C/C++ could not violate type safety for my code, now all I needed was to get it into mass production.

That’s basically what I’ve been working on since 2017 (though also on some other relevant bits). $1bn would be sufficient to stop waiting for Arm to find a customer willing to say that they’ll buy a large quantity of chips if CHERI makes it into the architecture and just tape out my own moderately competent 64-bit RISC-V core with a licensed GPU and other bits and still have a couple of hundred million left over for the software work.

I helped to start a non-profit ISP. A billion dollars would mean we could replicate the model 1,000-1,500 times over.

Fund research into curing or mitigating retinitis pigmentosa, the macular degeneration disease afflicting my wife and her kids.

Pretty much building servers, OSes and languages from first principles to incorporate the knowledge we’ve gained, to shed the legacy we’ve accrued, and to bring hardware and software people closer again.

At some point in the not-so-distant past I would probably be keen on self-driving cars, as road traffic injuries is one of the top 5 causes of death for people aged 5 to 29 years. After a few years of disappointment with the progress I’m not confident that a billion dollars would will suffice, so I’d probably retire and start a YouTube channel called “rich idiot builds a wooden yacht with no experience”.

Tools for direct democracy. Most people (in the US) are unengaged in politics at all level, partially because their input is limited to a small selection of votes. I’d like to build systems where citizens can provide direct input to politicians, and possibly even get a vote of their own to supplement representative bodies.

I’d spend a billion dollars on creating a competitor to the Mozilla foundation that would fork the Firefox codebase and pay a bunch of smart developers to maintain it, without doing stupid and user-hostile things like this thing in the news that Mozilla has been known to do.

My other idea would be “buy out Discord, hire @arathorn and a bunch of other Matrix core devs, and make Discord become the open-source Matrix client with by far the best UX (and sell server hosting on the side)”, but I fear that a billion dollars wouldn’t actually be enough to do this.

I’d create a universal standard for electrical mesh networking in order to create a fully distributed power grid. Buy a solar panel, plug it into the grid. I wouldn’t need centralized crypto ledgers for the payment. Each vertex in the graph would know how much it had sent/received. Since we’re talking about physical connections here, you could just meter it and bill using normal wire transfers using traditional banking. If a node ended up with an unpaid balance, it would just be automatically disconnected or de-prioritized.

I really want a world where rather than power outlets we have Universal Power Transfer sockets. You buy a solar panel and just plug it in. You buy a battery and just plug it in. You have a laptop? Just plug it in. No thought to it what-so-ever, no electrician needed. Just consumer distributed grid power.

If I wasn’t so obsessed with consumer distributed grid power, I’d create a stenotype system with electromagnets which would allow you to read using the keyboard (like that the keys would move). I’ve already developed a low power prototype. I don’t use electromagnets to actuate the keys (that takes too much electricity and is too bulky), instead I use electromagnets to hold the keys down. You can press down all the keys and then the computer can allow some to pop up, showing you the word. Pressing the word down again causes the next word to pop up. Stenotype reading is going to be super weird, I think that it is fundamentally strange that you would read with the same exact neurological pathways that you used to write. It might also be useful for blind people, but I’m mostly curious about how it would feel myself.

I would like to research technology that does not require attention. I would also research technology that reuses old hardware instead of throwing it away. I would also research technology that makes running the software in your own home (as opposed as in a SaaS) feasible again, possibly piggy-backing on the “reusing old hardware” research.

I would love for my mom to be able to (relatively) securely run a piece of software at home without having to understand the intricacies of setting it up. Something like how my smart tv runs, at most. I know that phones home, but like. You get the idea? Kinda? Idk.

More local shit that doesn’t require your attention, more software that you own on hardware that you own, that will still run when my company goes under. Peaceful, community-building things, that help you be in the real world, not suck you into a virtual hellscape.

A screen reader that uses machine learning, computer vision, etc. to literally read the screen, i.e. interpreting the pixels on the screen to reconstruct the text and the UI structure. I’ve concluded that this needs to exist on all desktop and mobile platforms, as soon as it’s technologically feasible. Getting semantic information from applications through accessibility APIs, semantic markup, ARIA, etc. is a useful stopgap, and will continue to be for a while longer, which is why I’m not giving up on my AccessKit project. But getting the whole world to accommodate disabled users by implementing this kind of accessibility is a never-ending battle, and fighting against most people’s natural preference to pretend that we don’t exist gets discouraging. I want to solve the problem once and for all.

Apple has already implemented a feature called Screen Recognition in their VoiceOver screen reader for iOS. Here’s their paper on Screen Recognition. They had to train it on ~77,000 screens from ~4,000 apps, using a total of 50 workers to gather and annotate the data. And as they admit, the results are far from perfect. I wonder how much better could be done with $1 billion, or how soon we’d reach diminishing returns.

Edit to add: No, that’s not ambitious enough for the posed question. Let’s go for a pair of smart glasses with stereoscopic cameras, built-in bone conduction headphones, and an on-device AI, that can not only read a computer or phone screen, but describe whatever it’s seeing in real time (something akin to the audio description tracks available for many movies and TV shows), translate text to Braille and send it wirelessly to a Braille display (and maybe we should build a larger, higher-resolution tactile display while we’re at it), and use spatial audio to help with independent navigation.

I don’t think planning $1 billion up front is a good idea, but I could use $1 million to pay people to develop a shell GUI, which I mentioned here yesterday:

https://lobste.rs/s/uxy6ge/even_more_hindsight_on_vim_helix_kakoune#c_pywj1o

Personally I have my hands full with the shell itself, and I don’t have the UI skills/background.

Actually we have gotten 100 K euros in the from of 2 grants from NLnet

I think somebody should apply for a grant to make the shell GUI. I might write a blog post about it. (Feel free to e-mail me andy @ oilshell.org if you have UI experience and are interested in our grant experience )

BTW Oils is the ONLY POSIX compatible shell that has a “headless” interface can be factored like this. I just put the tests for the protocol in the CI yesterday – the GUI would communicate with a pure C++ native binary that’s 1.4 MB now. It compiles in <30 seconds.

(It’s also the only POSIX compatible shell that will have a JSON-like data model, i.e. real data structures and not just strings)

idk, maybe try to figure out a way for job applications to not be horrible for both sides.

I would take the money and turn it into a foundation. If you invest it and pull out 3% a year you have $30 million for projects hopefully in perpetuity. @andyc can get his million and so can 29 other people. Or maybe you have bigger projects like building housing or buying medical debt. The biggest project would be staying a normal human when you have so much money that you are giving away. The second biggest one is not letting the bureaucracy grow to the point where consume all the budget or distracts from the purpose.

I’d finance another few seasons of Firefly. :D

A front-to-back open source, open hardware version of The Guide from Hitchhiker’s Guide to the Galaxy. Basically an iPad mini-sized entirely open computing device that works as a personal computer, tablet, e-reader, and with its own federated social network and Wikipedia-like news and encyclopedia. You can dock it an use it for personal computing and games, you can carry it and use it for communications and reading, reporting, and the entire stack would be open source, federated, P2P, encrypted, etc.

It irks me that in 2023 we still need a phone, tablet, and laptop/PC when you could realistically have one device to rule them all if it wasn’t in Apple/Google/Microsoft, etc’s best interest to keep all these things separate. In theory you could do (most) of this today with an iPad mini with cellular but it’s artificially limited so it doesn’t have the phone app and the iPad OS isn’t any great shakes if you want to use it as a PC. (Though to be fair to Apple iPad OS is slowly creeping in that direction.)

Obviously the device would be fully repairable, to reduce e-waste. If you want to upgrade to a faster device every three years or whatever, it’d be possible, but not encouraged or required. Ideally people could use it until it’s just worn out physically.

What the Internet and personal computing should’ve been, basically. I don’t know if you could accomplish this with $1bn, but it’d be amazing to try.

a stakeholder marketing risk management customer focused stock management crypto accelerated tool for real state agents to optimize the synergy of non fungible transactions. Humanity needs infinite corporate CRUDs, we’re like bees producing honey, but we make corporate CRUDs instead. That’s the meaning of life.

Kick stack as many cooperatives as possible, asking them to chip in once they are profitable to keep the trend going. If a lot of people learned to talk to each other and work together as equals, it would greatly improve our overall political climate.

In the world at large, there are much more important things to spend a billion dollars on than software, like, I dunno, malaria prevention or vaccine research. In software, I would try to make an open, complete, easy-to-get-started-with way to build everyday Web apps. The motivation:

• $product_from_work is built/operated/supported a mid-sized team (10 folks) and we’re well-served by Django, an RDBMS, and some AWS services. It still bugs me how much of what we built is still “factorable” stuff we and a zillion other teams each have to do themselves.
• Integrated “code it and go” services often lock you into a vendor and have a huge markup on the underlying hardware for steady-state uses (Lambda, Vercel–I know, the markup is how they get the billion dollars), and/or you hit a cliff where you can’t do what you need in their environment. You need it to be easy to start with, but flexible when you outgrow it.
• I’ve got a soft spot for how people who aren’t expert programmers could strap together something useful for their work in FileMaker or such, or folks familiar with a framework can churn out a CRUD site quick. People tend to pooh-pooh it as not-really-programming but if you’re trying to make the potential of computers available to as many folks as possible, degree of difficulty is a bad thing!

The project is to assemble and package up the stuff that’s missing around CRUD frameworks:

• Front-end batteries: there’s a range of widgets and form idioms that are almost universal out on the Web but still tend to take a lot of reimplementation in every app, from “decent multiselects” to complex interactions between fields (field Y is required only if box X is checed). You need to make decent forms easy w/out creating a cliff where people need to leave your framework in order to do anything custom.
• Back-end batteries: the development and operations envrionment, from things like repo, deployments, a Web cluster, monitoring and alerting, your DB setup (incl., say, backups and scchema migrations), backend and cron jobs, miscellaneous services like logging, caches. People are often either stringing this together or getting a managed environment at the cost of actual dollars and/or vendor lock-in.
• Administration and multitenancy: Many apps have at least a few types of user: customer administrators, customer end-users, and developers/operators of the application, and many can operate different instances with some degree of separation. CRUD frameworks’ authentication also needs some awareness of modern world (2FA, outside identity providers).
• Openness: The whole thing should be open source and able to run on bare metal. Not that you couldn’t use a service for blob storage, CDN, or a managed DB, but even if you don’t, you get a complete stack. As a corollary, if some bundled/“blessed” component doesn’t work quite right for you, you can integrate a different one. (This also complicates the “back-end batteries” part; you need a fallback to AWS’s or GCP’s approaches to things.)
• Using existing stuff: Despite the billion dollars, no one is writing a new database or wholesale reinventing the CRUD framework. That loses the benefits of the existing ecosystem and gives you the chance to remake old mistakes or make new ones.

I specifically think it’s not right to tackle instant/infinite/transparent scalability (vertical RDBMS and frontend/worker cluster, or one per tenant, scale reasonably large!) or to replace the programming models people are actually using. Those eat your project and are not really where the barriers to entry are.

Now if you manage to get all that done, I have a few silly specific ideas where some elbow grease could probably make things better:

• There are particular parts of the stack where I think the right way to do things may involve something not super common in CRUD frameworks now, like templating, frontend performance, and what to do about reporting when you get medium-sized.
• Once you reduce the overhead to coding, you can start to think about a super-easy entry point–“low-code” as they say these days–that ramps up to really writing code. I know various projects try to do this on top of existing frameworks.
• If you’re actually operating a large instance of this you have the motivation to work on scalability/efficiency. That isn’t necessarily Big Ideas, ‘just’ the (interesting!) loop of profiling or organically noticing pain points and figuring out how you get past ’em.

But, again, I think a satisfactory factoring of the things every mid-sized app already needs to do is an enormous already!

A massive scalable mutation testing framework for Erlang or Elixir. Use a very big cluster to provide extremely precise coverage analysis beyond line coverage. With the goal of making quality less stressful for developers.

Something that watches right-wing news, stores it, captions it, and puts the scripts on a publicly searchable database so people can easily find out what X person said about Y, Z years ago, and provides the clips as needed.

For $1,000 I’d look into vCard 4 support for my validator.

$1,000,000 would fund an expert working group to develop a modern, open address information interchange format to replace the clunky vCard and similar standards with something sane and easy to use, all the way from inception (what’s good & bad about the existing standards) to tooling.

$1,000,000,000 might be enough to develop something on the order of the next cURL or SQLite, by hiring top people in relevant fields at unbeatable salaries with the resources they need and a well-defined mission. Maybe a replacement for LDAP and Active Directory which would be easy to get started with and AGPLv3 with some sort of extension clause so extensions would have to be AGPLv3 as well. The project would make many more billions on support and extension requests from millions of businesses.

I would develop my Qweyboard idea[1] to be suitable for production use to replace the Velotype/Veyboard for orthographic stenography. That would probably only take 50-500 k over its lifetime though so I should either get more ambitious or donate the rest to charity…

But I think giving humansthe ability to transcribe speech live may also be somewhat of a historic anomaly that we’ll outgrow. So maybe it’s not as important as I think.

It would definitely be something in education. There is so much incredible untapped potential in young people around the world. I’m just not quite sure exactly what it would be. Perhaps an AI-based tutor – but one that actually, really works and is rigorously tested.

A system that slowly converges humans towards an efficient heterogeneous geographical distribution within given borders. The entire landscape is evaluated for resources and potential density. Those who are part of the system receive benefits for contributing to infrastructure and living in designated areas.