Software Development

There's many reasons why you'd want to work for trading firms and most of them like Valkyrie will hire juniors.

Everyday you are the race car engineer of software like how they will rebuild an engine and retune the car after each race to squeeze out more optimization that's what you do at many of these firms. There's also work in building high performance models and fancy consumer dashboards.

Jove Intl, Oxford Knight, Reload Search, and ex-employees who have quit to become recruiters. They hire internationally for Chicago, NYC, Austin, London, Sydney, Hong Kong, Mumbai, Shanghai, Tokyo and now Singapore seems to be the location everyone is opening offices these days. LinkedIn is trash but works fine for harvesting recruiter offers.

I got to try their fancy recruiter tools and paid access to LinkedIn to see the minefield of bad candidates they have to filter.

They told me all resumes are reviewed manually there's no automated ATS or applicant tracking system. USACO training or competitive background in anything from sports to math stands out, so-called modern C++ meaning C++17/20/23 stands out but is not necessary as any firm will train you. You never sell yourself as a junior, companies will hire anyone who knows how to sell themselves and can pass the technical competence filters. You always pretend to be employed writing software even if this means setting up a personal page declaring yourself a freelancer on Upwork actively looking for contract work.

Getting more and more monopolized by big tech corps pressuring students with 2-week expiring offers (aka exploding offers) that are against campus hiring guidelines but they do it anyway. Think about how much money is spent on campus engagement to find (highly paid) interns and they have to hope that intern comes back after graduating and many don't. This should tell you they really want to find junior candidates willing to stay to become senior staff so if they find you as a rough diamond in a sea of mediocre trash it's an easy investment for them.

I don't know if they still do this but one Chicago company was notorious for firing off 3 problems to solve on HackerRank seconds after you submitted an application. The first phone screen is usually a recruiter asking are you legal to work and if not what credentials do you have because in some countries like the US you will need a reason for them to issue you a work visa for a job that can't be given to a local so a degree or equivalent work experience. If they want you they'll finesse these requirements if not try Asia.

The new fizzbuzz filter these days is to implement atoi and tests how careless of a developer you are meaning you don't ask any questions about requirements and start writing code immediately. You may be asked something like this as a screen. Here is another typical remote style interview screen which is the first of many, many more interviews. If you pass these screens you may be invited on-site meaning they will pay for you to come out there and job shadow for a week while being hammered with interviews everyday if it's a big US company however I've noticed in Asia the hiring process is much shorter.

As a personal anecdote the on-site interviews after all the screens seem designed as a trap to once again see how careless of a developer you are. The problems I've seen have an obvious area of optimization to take advantage of to squeeze out some pointless constant factors but will open a potential bug or performance sink elsewhere if you aren't taking your time thinking through the whole original problem space. The bad candidate will pounce on that and show off their meaningless optimization while the good candidate will ask what happens if I do X, is it a good tradeoff for the risks involved? Prime example they will try and get you with is cache invalidation and we will learn all about caches here so you won't be the one who invalidates the cache on their last day of on-site interview.

Hiring someone is an investment it will typically take 3 months for them to be productive and if you're being paid $200k/yr that's $50k they have spent hoping you will turn out to be a good hire so now you know why there is all these hoops to jump through.

Once you are in and develop a routine meaning you are no longer panicking try and walk over to the department where all the money is being made and find a mentor to teach you by asking them to give you the work they don't like doing because you want to learn it all from scratch. All you have to do is be the reliable person they turn to whenever some (tedious at first) work needs to be done and if this work over time becomes increasingly more skilled then you have found the right mentor and you are set for life because they will eventually want you to work with them once you have proven to be indispensible. If they get promoted you get promoted, if they go off and start a new company you are the first hire. This can be tricky because of signing multiple NDAs even between internal departments but if you are serious about wanting to learn they will tell management to make it happen. I actually took all the above advice from a comedian many years ago and my life would not be the same if I hadn't watched that throwaway clip at 3am because I always assumed there was no way I could ever do this kind of work.

If you're not interested in finance there are a few companies that mentor developers sometimes even direct from local high schools and build them to senior developers. One is in New Hampshire called Northwoods Software. If you live in Egypt, Georgia, North Macedonia, Albania, Latvia, or Bosnia you can work remotely for Scandiweb (Latvia). The pay is only 500-600 Euros per month but you don't know anything yet so are being paid for direct mentorship to become a senior developer. I've never talked to any developers outside of FAANG that weren't interested in hiring good entry-level juniors even though they didn't advertise these positions.

The prof of this course with a grad student created gitless.com see this research page on MIT's software design group. Read how they audited the conceptual design of Git and found things like staging were confusing.

Many companies today use a customized Mercurial fork or rewritten clone instead of git like Meta's rust written EdenFS and Sapling SCM because git was originally developed to take advantage of linux file system optimizations ergo is conceptually attached to that file system while Mercurial can be easily adapted to be stuffed in a database, distributed or run on any OS. In a Mercurial style of version control, you probably don't care about individual commits so you don't even add comments to commits you instead care about entire branch development.

For these reasons I don't think you should bother to use git at all unless someone is forcing you. Look up how easy and simple the Mercurial guide is for a lone developer why use anything else. If you must then try MIT's version control lecture from it's course The Missing Semester of Your CS Education or read 'Git for Hackers. Open source development outside of FOSS you generally fork the project to your account on Github, commit changes there then make a 'Pull Request' or PR to the original repo asking if they will merge your changes. The FOSS way is different using Savannah for example here is a patch it's all done in plain text using a traditional mailing list style of attaching files and people vote/review it.

I tried a local experiment where the goal was to ask people about their business and what problems they were having to see if software could help. You just let others talk about themselves. I ended up with dozens of leads and if I was a software mercenary I would have picked up a lot of work. Unfortunately I'm not a software merc but kind of thinking I should be now that I've seen how easy it is. Everyone wants to work with someone they've personally met not a stranger on a website and almost every conversation ended with 'do you know anyone who can do this' and them giving me their contacts.

I also called people out of the blue from LinkedIn from various local companies and just told them hey I'm trying to find out about your work tell me everything about it and they talked to me too which was not what I expected.

Here's what I found out by just walking into places and meeting people. Every one of them had a problem with existing software they wanted to modify and they had never met anyone locally who could independently write software who wasn't some SaaS sales guy pushing an existing product.

Look up whatever web design awards for 2024 and see what people are doing. It's always going to be Europe because you know why but anyway examples let's review them. All of these use minor animation to show an incremental view as your scroll. Here is one site they nominated.

You will want to bill daily not hourly. Keep increasing your rates too, if you have so many clients you can't accomodate them all then you're not charging enough of a daily rate.

Search for agencies and you will find many of them. Agencies are often never hiring because they find people to work there by approaching the best in the business and asking them to join the agency. This is how architect agencies work, ad agencies, etc. There's probably agencies in your city they will be more receptive to meeting and working with you than some remote agency.

Try software marketplaces like Shopify or Gsuite and read all the reviews to find out what paying customers hate about a paid addon then make a better one. Notice how bad form creating software is in Shopify app marketplace that seems like something you could start selling right away and compete. If you start making money make sure you have 24/7 service and it's not a worthless chat app.

A real startup requires venture capital. Here's Peter Thiel giving a very good crash course in making a startup to Stanford students. Start small and monopolize something small. Here's an old article of Sam Altman of Open AI (now a monopoly too) talking about how Y Combinator's success is due to it being … a monopoly.

I researched the Y Combinator startup school so you don't have to and they tell you as a founder everyday you are only talking to customers. If you have 5 customers you talk to them everyday and ask them about their needs and make the wagies implement those features. If you have 50 customers you still talk to them everyday. You phone everyone you can in the business niche you are targeting and find out what they want eventually they will start paying you. That is literally it there's no other magic or anything it's just talk to absolutely everybody and build growth by listening to them and making a product that others will want because you listened.

Who buys software companies? Salesforce. They seem the most aggressive company right now buying up everyone.

The USACO guide has a good introduction to big-O and a resource we can use where running time is actually explained is the beginning chapters of Functional Data Structures by Prof Ragde at Waterloo try reading the first few chapters up to 2.2.

The formal definition is:

\(f(x) \le c\cdot g(x)\) for all inputs \(x \ge x_0\) as x (inputs) goes to infinity. If that inequality holds then f(x) is in O(g(x)) which represents a worse-case scenario upper bound. A 'best-case' (wishful thinking) scenario for insertion sort is you don't have to do anything but attach the element to the already sorted list and the worst-case is the entire list is in reverse order and every element looked at and moved.

f() function is a cost function of time or resources derived from a cost model. It's where you count up every constant operation and loop costs by analyzing source. This is not always easy since what costs do you assign when the code triggers garbage collection or triggers memory allocation because an object has been created deep in some library code.

The g() function is chosen from a list of math functions that represent a 'family of functions'. A family means all functions that are linear, all functions that are quadratic, all functions that are linearithmic etc. It actually is a specific list as we'll soon see.

Why are constants hidden in big-O? If you read SICP their example is 16 and 8 both differ by a factor of 2 and 8 but we don't care which factor. In real life the constant factors do matter if they are hiding a lot of operations. A classic example of this is when Knuth was writing the Stanford GraphBase (contents are in Vol 4A of TAOCP) and his real-world testing of popular graph algorithms found that 'Fibonacci heaps' that were analyzed using a technique called amortized analysis weren't any better than Kruskal's old algorithm because the analysis was hiding large constant operations. One way to reduce constants is to do consecutive rounds of a linear algorithm such as cleaning up a graph removing cycles because O(n + n) or O(2n) is still O(n).

Big-O doesn't represent growth since if f(x) = x² it does not grow to be quadratic, f(x) already is quadratic, thus f() is in \(O(n^{2})\) or upper bounded by every quadratic function. Typically in algorithm analysis the inputs are supposed to be 'near infinite' which is why n is used instead of x in f(n) to signal the difference in inputs but here in the USACO guide we can see that if the input is small then we can get away with some very bad runtimes even exponential.

Before we start trying to solve problems watch this Erik Demaine lecture and skip to 09:00. We want to know what polynomial time is, what NP means, what satisfiability means and this is a good enough ~30m crash course in complexity as these concepts come up all the time in algorithm literature. He tells us most decision problems can't be solved by an algorithm which ruins the AI superintelligence dystopia. When he translates Super Mario Bros to a 3SAT decision problem we can see that almost every algorithm you can think of can be transformed to a decision problem. You don't have to keep watching this but there's an interesting intro to constraint logic/graphs. Now you can explain to someone P=NP means all the nondeterministic polynomial time algorithms that always 'lucky guess' correctly or have a magic oracle can be solved by an algorithm in deterministic polynomial time which seems impossible.

The rest of this course I won't do here but the book he wrote that goes with the lectures looks excellent covering topics in economics too.

If you look at the archived contests we are given all the test inputs/outputs as a .zip and a fully worked out solution with a proof of the bounds. This means you can use any language you want however I am going to do it all in C++ with Knuth's literate programming tool cweb because it mimics exactly how you would build something in a collaborative setting and after using it for a few problems it's kind of awesome and I finally get why Knuth shills this style of programming so much.

TODO

Home