Originally published at www.crema.us on May 13, 2019
Throughout the course of our two-day developer conference in San Jose, California, we shared exciting new updates and innovations to explore the relationship between technology and human connection. This year’s event featured 70+ sessions and offered hands-on experiences with Facebook’s products and tools.
— Facebook Developer Team
Since we couldn’t attend F8 in person, our development team at Crema decided to select several talks and watch them together right after the conference ended. Here are the ones we chose, what we learned, and snippets from our group discussion after each session.
- Hooks allow you to split your component based on logic, not lifecycle methods.
- Use hooks when you want shared functionality between components
the beauty of class-less code is a more concise, logically grouped, highly reusable set of functions
- Commonly-used hooks: useState, useContext, useEffect
- Custom hooks can be created as a standard JS function beginning with the “use” keyword — hooks may call other hooks.
- Tip: Only call hooks at the top level — this ensures that React can preserve the call order.
- See also: https://usehooks.com
- Two of the creators of the language, Dan and Lee, explained that the framework was created to unify the query approach across all apps.
- Invented in 2012 for/by Facebook, transferred to the Linux Foundation in 2019!
- The future of the language at today’s scale and adoption is heavily reliant upon open-source support from large companies via the foundation.
- Service-based architectures or microservers can greatly benefit from GraphQL as a “wrapper” around public or private REST endpoints.
- Machine Learning: A subfield of AI giving machines the ability to carry out specific tasks without explicitly programming them to do so.
- Deep Learning: Machine learning with artificial neural networks that “learn” associations between inputs and outputs.
- Neural Networks: Computing systems made up of interconnected nodes.
- How it works: Data input is processed through a linear transformation (weights and biases) followed by a nonlinear transformation (activation function) is applied to contextualize the output.
- Example: Identifying and analyzing the individual pieces in an image (counter, sink, pans) leads to an approximation of what the image may be representing as a whole (a kitchen).
- Building a ML project goes something like this: Determine the approach, prepare your data, build and train the model, then deploy and scale.
- Don’t sweat it: There are lots of third-party tools to get started in ML!
- The mobile-based attention span has changed everything.
- Always use creativity to capture people’s attention. 56% of brand sales coming from mobile are thanks to good creative.
- Short-form content that has your product front and center is best.
- Tips: Go shorter in duration, design for sound off (your audience may be watching TV!), and frame your story in either square or vertical formats.
- 100+ Languages supported by Facebook.
- ~2M Translations processed every week.
- 57%+ of Facebook users are non-English speakers.
- How does this compilation in web components affect runtime? It’s a Babel plugin, so it happens client-side.
- Using inline tags, the i18n team can construct sentences with untold combinations that translate intelligently at runtime based on the user.
Remotely streaming a conference is a great way to use your time if you can’t be present for the event — even taking a few minutes to chat about each session with the rest of your team can be very beneficial.
We love making things and it’s important to take time to learn together. Keep an eye out for more recaps, as we hope to do more of these in the future!