You reply with a combination of trolling and endorsing a specific link that you believe represents doing it better. The first thing I looked at, syntax study, was already weaker than prior work I've seen on the subject. The language choice was highly biased while not focusing on enough metrics for comparison of strengths and weaknesses. Between field use and its first "evidence," your Quorum is already weaker than most of what I presented in terms of empirical weight. Due to the trolling, I'm hesistant to even watch the video in case his presentation is something likes yours in terms of content.

@ non-trolling, HN readers also concerned about science in software

Back to science, though, given other readers might be interested in that point. Software development is a combination of human- and machine-driven processes to turn some requirements into working, maintainable code. A subset of this, significant it turns out, can be analyzed for effectiveness with enough different case studies, sample sizes, objective measures, and repeats to make a claim about it with believable evidence. For process stuff, it's not going to be as logically formulated or minutely analyzed as a math equation or something. Not usually, anyway. Instead, it works more like this:

1. A recurring problem exists in the development process or artifact.

2. Measurements are taken of how often that problem occurs in general and for specific scenarios.

3. A hypothesis is formed as to why it exists and a solution to it.

4. The hypothesis is tested by applying the technique to several projects likely to experience the problem with objective records on outcomes and subjective accounts from users for a take on less tangible aspects.

5. Significant reductions in the problem in the objective data provide supporting evidence that the theory is correct.

6. The theory (solution) is tested by others against the problem to find more supporting evidence or counter-examples to the claim.

7. If it's correct, the results continue to speak for themselves in it being a solution with likely modifications as the problem and solution space are further explored.

Note: If it's performed by humans, it's wise to either look out for or control for issues like Hawthorne effect. Must be sure that it's the overall solution that's working rather than people's heightened attention to an issue being studied.

The above 1-7 have applied to my list of techniques to varying degrees ranging from strong analysis + repeated, field results (eg formal specs, code reviews) to clear theory, models, assessment, and massive field results (eg type-systems, usage-driven testing). The evidence is quite clear and favors them as effective with repeated successes that matched the theory. Many other aspects of programming I left off the list because the evidence or studies were not clear. Or were non-existent. Hence me not including the parent's strawmen ("most science") or red herring (Quorum) on a list of techniques with decades of measured results behind them across many software projects.

Science, what little exists in our field, is on my side if the reader is focused on evidence aspects such as the Why, How, and Got Job Done. They collectively make up The Right Thing in robust, system development. YMMV for individual techniques or tools on a given project. Pick best tools for the job, prioritize your development budget, and so on. Common sense. People looking at where to start dealing with recurring issues or overall robustness have an evidence-backed list in my main comment, though. There are also papers and books on each showing where to apply them and most effectively.