I read Software development topics I've changed my mind on after 10 years in the industry recently and thought I would do my own.

I've been programming for money since 2014 (whoa) so some of my opinions have changed over that period and some haven't. I don't claim this makes me good, but it does mean I've seen some shit.

Things I didn't used to believe, but now do

• the beginning, middle, and end of programming is data layout and data access patterns
• relational databases are a civilizational technology, like water treatment or the refrigerator
• long functions are often fine, actually
• the way code looks matters a little, not a lot
• you can probably get away with HTTP for almost everything networking-related; the burden of proof to to use non-HTTP should be high
• the bare-minimum for a programming language is to be memory-safe
• YAML delenda est
• SQLite is the best piece of software yet created by humankind
• statically-checked ownership rules for mutable state are often easier to use and understand than functional programming
• customers would have better experiences and building would be faster and cheaper if we used Hypertext and HATEOAS stuff more widely
• there is such a thing as too much teamwork and too much coordination
• most companies would be better served by reducing the amount of coordination they require and accepting some level of rework and incompatibility in exchange for faster delivery and higher autonomy
• knowledge of SQL will likely outlive every other language/systems skill you have
• just writing SQL in lieu of using an ORM is almost always fine
• premature abstraction has left a trail of bodies way, way longer than premature optimization
• until we have a "Rust, but garbage collected", most companies are probably best off with Python or Java, depending on domain
• the thing most likely thing to sink a project is management (and product management) not understanding what they're asking for
• too many programmers on a team is worse than too few because you have to find things for them to do, and there are side effects from trying to use nine women to make a baby in one month
• no subset of the software field has failed to pull its weight more than product management
• the median manager is probably worse at their job than the median software developer
• good managers are harder to find than good software developers
• most companies would be better off using something like Heroku than attempting to run their own infrastructure
• clever people choose technologies that are fun to use, smart people choose technologies that are easy to use, wise people choose technologies that are easy to operate
• there is a huge opportunity for a new language in the space that Ruby and Python currently occupy
• technical product management should be a career track that becomes available once you are a senior software engineer, and only then
• beyond whether it has LSP or not, your editor doesn't affect your productivity enough to be statistically significant
• languages that have weak LSP/Intellisense experiences will get left behind
• doing things on "hard mode" (e.g., by hand, without types, without LSP, without help, etc.) doesn't tell me you're good, it tells me you don't value your time or mine

Things I still believe

• software development is a pop culture
• we don't actually know that much about how to program successfully, so you don't have to listen to people who think we do
• the thing most likely to sink a team or a project is bad management. everything else is a rounding error
• class-based object orientation has been colossal mistake for information modeling and should be done away with
• language matters in so many ways: what you can build, who will want to build it, how users will perceive it, how expensive it will be to maintain, etc.
• large dynamically typed codebases are not enjoyable to work with
• just reading the code is still chronically undervalued
• programmers aren't fungible
• rewrites are often worth it if for nothing else than forcing understanding
• building in excess capacity - in infrastructure, in teams, in timelines, in performance models - is absolutely essential to resiliency
• programmers are irrationally obsessed with language syntax
• as a society we need to stop building new things with bash
• most companies probably shouldn't be operating infrastructure that requires them to know what Kubernetes is
• high-level managers (CTOs, etc.) should know what technology their teams use and understand their tradeoffs, but should avoid prescribing minutiae
• anyone who manages a team that producers software as its primary artifact should at one time have been a programmer
• title inflation is rampant in software engineering
• 5 years of experience just isn't enough to truly be "senior" at anything
• most developers - and really, most people - can't write, can't give a public talk, and don't like reading
• in many companies, writing is not a useful communication mechanism but it is a useful medium for thought
• building something yourself (rather than picking it off the shelf) often has more utility than the built artifact itself

Things I no longer believe

• SQLite is a toy database
• the code in the dependencies you use is special and somehow better than regular code
• functional programming is strictly better than the other forms
• Javascript is fun
• Ruby is fun
• Ruby is better than Python
• Java sucks
• people "learn in different ways"
• Scala is a fun and useful language with a great community
• short functions are easier to reason about
• object orientation is elegant
• everything should be rewritten in Erlang
• DRY ("don't repeat yourself") is a programming principle worth aspiring to
• immutable data is strictly superior to mutable state
• ORMs have a net-positive ROI on most teams
• a great language outweighs mediocre/bad tooling