Allison Parrish: There is great beauty in applying scientific concepts (interpolation) and the scientific method to domains that we wouldn't normally think to (syntax, grammar, poetry).
Anjana Vakil: Language matters, and has historically evolved to meet the needs of its speakers; we need to be better at helping the language in the domain of computing evolve beyond where it is currently being held back.
Augie Fackler: API design can (and should) be decoupled from API implementation; this decoupling can provide a valuable articulation point that yields flexibility in the right areas without compromising on principles.
Elle Vargas: We need to embrace the technical neophytes (or Juniors) mindset in our organizations, because they are best equipped to bridge the gap between people with little to no technical background and experts; failure to do this leads to security compromises on the scale of the manipulation of democracy.
Gary Bernhardt: The technical community needs to be aware of the altruistic goals of the early pioneers of computing, so we can move beyond systemic harvesting of personal information for purely corporate motivations.
Julia Evans: We need to discover the truth that hard problems are solvable if we commit to solve them, and we need to embrace the unknown and be willing to step out with creativity and exploration.
Jumana Bahrainwala: Scaling DB Migrations is highly dependant on the size of your team and the problems you are solving, but generally you will want to break up migrations at critical boundaries to avoid locking tables and consuming resources.
Katrina Owen: A strategy for making decisions that involves exploring more than just our predisposition to the binary is healthier; we are better equipped to make decisions when we have more options, driven by self-discovery.
Nabil Hassein: We need to think about how our relationships are affected by the work we do and the principles we hold; this includes people we might not normally consider in our relational circle, such as those in other countries or industries. We need to be able to feel "right" with those relationships, and this is worth thinking about.
Pablo Meier: Redundancy and Failure in systems design is good to think about, but Erlang has solved most of these hard problems already, so we should probably avoid reinventing the wheel and build on top of Erlang if we want really fault-tolerant, redundant systems.
Ryan Herr: Everyone has different approaches to learning, and sometimes what works to enable someone to "just get it" doesn't work for others. It's healthy to deconstruct the system in the subject we want to learn in a way that makes it approachable for us.
Sandi Metz: Good object oriented design is more about passing and responding to messages between objects in a system than blindly applying patterns. We need to be able to name things effectively to achieve effective OO design.
Stuart Halloway: Documenting decisions (in design docs, or good comment hygiene) is one of the most effective ways to scale large systems.
Tea Ho: Despite common perception, accessibility is approachable and can vastly increase the quality of life for people who are disadvantaged and suffer with disabilities; we need to socialize the idea of developing for accessibility to counter the myths that it's too hard or won't affect a meaningful group of people.
Tom Murphy: (I actually don't have much to take away from Toms talk other than meta-humour (on multiple levels) is hard to track when you are watching an EE and tinkering wizard demo way more than you ever thought was possible on a NES from a live stage.)
Vicky Nguyen: BGP and the internet network stack is fundamentally broken, so vendors have had to invest a lot of capital to "fix it" by providing delivery networks that reinstate how the internet should work.