I will evaluate whether I actually think it would work and have a big impact, and award the bounty accordingly!
Typically, the best startup ideas are counter-intuitive and not immediately crowd-pleasing (otherwise, it would have been done already!).
TagTime: stochastic time tracker. Instead of having to remember to start and stop a timer, TagTime randomly pings you, asking what you're doing at that moment. You answer with tags and it builds up statistical estimates of where your time is going. (For example, it tells me I've spent just over 200 hours on Manifold in the last year!)
/BoltonBailey/commercial-application-of-quantum-p , applied to high-frequency trading. Quantum advantage on this has already been demonstrated at kilometer distances. Ultra low latency trading apparently will pay $100m per millisecond per year. While correlation is not quite as good as faster-than-light communication, surely it is still useful.
Any of the Possible Manifold Foci listed in the following market that are not Manifold's focus could be spun off as their own startups: https://manifold.markets/dreev/which-of-these-possible-manifold-fo
The Maker Scheduler (makerscheduler.com or msched.com yes I'm sitting on the domains in case I ever want to make this): a tool for connecting with people you want to meet with, without the productivity hit of putting something on your calendar.
A deep tech company for non-invasive brain stimulation.
Much of the funding for non-invasive brain stimulation is scattered leading to flash in the pan single use case startups and underpowered studies. However, the field needs a critical mass of funding to solve the hard problems as soon as possible. This organisation would create the tools needed to accelerate research, but ultimately put the technology into medical and consumer products. The goal would be to create the hardware and software platform necessary to be able to tackle the many different areas non-invasive brain stimulation can be used. Stroke rehab to major depressive disorder, attention to sleep, all through one hardware and software platform. The big AI companies started with a similar problem, a need for more funding compared to academia and a long term view of how the technology could be broadly transformative without a specific product in mind.
I think a really nice type of software to make would be a user-friendly automatic file resorter. The reason for this is that most people tend to have very messy filesystems on their personal computers or Google Drives (and I am no exception!). At a larger scale, I think bigger projects like monorepos and such also tend to get messy over time without maintenance, although I think this may be less of an issue than it is for individuals? There are already a few programs that automatically reorganize files, but none(?) are that user friendly to my knowledge.
People take pictures to communicate something. For instance a realtor may want to take a picture of a house to show how nice the front yard is. The owner of a restaurant may want to show off his food. A tinder profile picture has the purpose to attract people of the desired sex. Etc. In each case we have a clear cut metric (house price, number of orders, number of dates). It should be possible to train a few shot (or even zero shot) learner to predict the metric given the picture. If you then run LIME or similar explainability tools you can get a heatmap showing which parts of the image affect the metric. Will the house sell better with or without flower beds? Will customers order the pizza with pineapple more often than the one without? Will the profile with moustache get more dates? This can be the start of a smart image analytics platform. It probably lacks focus though: narrow down to a single use case and understand who the customer is.
BitBoy is a free cell phone app that allows you to create your own cryptocurrency. You download this app, set a few sliders to determine how much future coin can be mined, how much issues to yourself and the initial trading price. One button launches the new currency. BitBoy makes money by issuing itself 10% of all new cryptocurrency minted by this process.
I wanna make an app that has 3 way integration between companies (and their clothing/makeup/fashion products), retailers, and consumers.
But it uses midjourney to generate images of the users trying on various products under different lighting conditions or in combination with other products. I think the combinations and different lighting conditions is huge, although not necessary to the mvp.
Super convenient for users as they only need to create their profile once by uploading a couple of images. Then shop away online to your heart's content trying on lots of clothes.
Immediately ready to monetise (affiliate links, paid promotion of products from companies, paid subscription models from users).
Existing demand and examples (some companies have crappy filters that allow you to upload a selfie to see how a certain shade of lipstick or foundation will look on you).
Biggest blocker is that I dont think midjourney isn't quite good enough yet. Also this would be a huge project. Although if anyone does want to maybe work on it hmu haha. Maybe Manifold can have a drastic pivot.