I work in engineering, sometimes with startup types that want to develop a “product”. I’m also a coinventor on some patent applications. This response will be based on US perspective and economics.

1. First before all, do a patent search. This is to find out if someone already patented your ideas. If so, you either need to pay them royalties to license the patent(s) or rework your product to avoid the patents. Google Patents is highly accessible for this.
2. Then, if you think you have original, patentable ideas, engage a patent attorney to do a “real” search and to work on filings. This will take money (at least 10s of thousands US) for the initial work. All the major legal jurisdictions are “first to file,” so it no longer helps to mail your notebooks to yourself for proof of date of invention. You have to at least file a provisional application to get a patent priority date. Keep everything top secret until you have that application. Execute non disclosure agreements (NDAs) with any outside firm or individual you talk with. 2b. As an aside, software is not generally patentable any more, on the grounds that math formulas are also not patentable. There may still be some ability to patent software-oriented ideas as business methods or the like. Just because the patent office issued a software patent doesn’t mean it’s enforceable. Courts hold patents to be invalid all the time.
3. I want to impress upon you some view of the real costs of prototype design and what is known as “nonrecurring engineering” (NRE) in the biz. You don’t say, but it sounds like you want some amount of custom electronics coupled with some backend software. Costs can vary considerably depending on circumstance, but I would typically see 100k-300k USD in design and prototype build costs to get initial prototypes with some limited functionality for these components. It could very well take 1 million USD or more to get a more complete product design. This also depends a lot on how you engage engineering talent: turnkey consultants can be the most expensive, or you could save a bunch of up front labor cost by offering equity to a key designer. 3b. I don’t know how complicated this app is, but it’s not unusual for software engineering costs to overshadow the hardware engineering costs, and sometimes by a lot. This might be something to keep in mind if you’re contemplating app development up front paired with virtual hardware plans.
4. A “virtual design” for hardware that is just drawings could be done for cheaper than the prototype quote I gave. I see common prices for that kind of work at least 10k and up to 50k depending on how much initial design work you want or need done (and how the talent is engaged, etc, etc).
5. Manufacturing. Depends of course on what is going into your gadget and how many units you plan to build.
   5a. Custom PCBs can commonly be run in low volumes for relatively cheap. It’s more expensive to solder the components on than just to etch the boards. There are many board houses that let you turn in your design files and get a quote online. 5b. For startup that wants a low volume (~100) of some gadget, you might want to look into contract manufacturers. These will assemble your product per drawings, typically in a non-automated or low-automation fashion. For example, they might have pick and place machines and expensive wave flow solder machines to assemble PCBs, but then the boards are screwed into enclosures by hand. These places might run double or triple the per unit cost of a more automated setup, but it can still be the best option for low numbers of units. 5c. Overseas manufacturing can cut costs through reduced labor bills. The traditional hurdles in the startup environment are long shipping lead times (particularly by sea, 10-12 weeks not uncommon) and the added hassle and complexity of international business dealings. In the US particularly, the recent tariff situation is throwing a monkey wrench right in the middle of this, and I will not attempt to analyze the impact. 5d. A commonly surprising manufacturing cost: if using injection molded plastic for enclosures or the like, the custom molds can cost several 10s of thousands to build and store. This is a fixed cost, so it doesn’t impact the per unit for large volumes, but it is often an expensive hurdle in the total manufacturing process for small startups.
6. Regulatory. It’s highly likely that any consumer facing gadget will need at least some regulatory testing, probably from a dedicated contract test house For example in the US, Underwriters Laboratory (UL) demands safety testing, and the FCC can require “part 15” testing and separate testing if you have a radio (such as WiFi or Bluetooth). As an exercise you could try looking up all of the various logos on the bottom of your favorite gizmo or in the fine print in the manual.

“Because they were prevented from restocking, they are now out of food, but this condition need not continue if they are restocked.”. Is that good enough?

So anything that NASA produces alone with public money is for the public by default ?

Anything that NASA civil servants produce and publish is in the public domain by default. NASA can spend public money on contracts that don’t result in public domain information.

In this case, if NASA spends public money to buy (license) a commercially available compiler from PGI, that compiler doesn’t magically become open source just because NASA is a paying customer.

Works, reports, and software that NASA produces itself are “works of the United States”, so they are in the public domain by law.

However, not everything NASA does is a published work, such as the classified GPS encryption modules on the shuttle or private medical conferences with ISS crewmembers. Additionally, a lot of stuff is actually done by contractors, such as SpaceX or Boeing, and those may or may not be required by contract to release various amounts of data to the public.

I did a quick Google search, and I was unable to find anything contemporary where NASA is maintaining or developing an in house Fortran compiler.