#internal-tools

[anupkaranjkar]

The Internal AI Playbook That Big Tech Doesn't Share

While tech giants sell you on simple prompts and basic workflows, their internal teams use completely different AI strategies. These leaked practices from Google, OpenAI, and Meta show the real methods that actually scale—and why public tutorials barely scratch the surface.

Read full article at wowhow.cloud

[trywildcard]

Inside Wildcard 2.0: Content

With launch behind us, we wanted to showcase different parts of our engineering platform here at Wildcard that power the beautiful experiences in our featured iOS app. Today, I wanted to briefly take you behind the scenes of the tools that help our editorial team discover and curate our one of a kind news + media experience.

When we started building these tools from scratch 5 months ago, I wanted the ethos of what we built to revolve around this mission statement:

“Our aim is to provide a beautiful, enjoyable, and robust toolset for our editorial team to curate content for our users.”

Robust is an adjective you often hear when describing engineering deliverables, so I want to focus on the first two adjectives for a second. When building internal tools, I feel that user experience & empathy is often dramatically under-valued. In my experience, taking the time to build tools that are easy and pleasurable to use generates dividends and overall goodwill that typically outweigh any extra functionality that could have been built with that time instead.

To facilitate this, we opted for a more traditional “waterfall” SDLC for our minimum viable product, where a fully fleshed-out design would be in place before any engineering occurred. Our rock-star design team focused on a clean, sleek look where editors could easily assemble “collections” of content about a topic (e.g. articles, videos, tweets, etc), see a live preview of what they’d curated, and publish to the app. Here’s one of our early visual treatments:

With a design in place, engineering quickly rallied around the vision and built out our feature set. Our front-end is virtually all built using ReactJS, which promotes a strong essence of state + code reusability with front-end code. In fact, much of the code we wrote to power live preview was re-used to power the beautiful web interface of our curated content. The back-end is built to leverage around our robust JAVA-based platform and uses REDIS as a caching layer to optimize performance.  

Some of the most important features that we’ve engineered are the ability to algorithmically surface the day’s most trending content to our editorial team to aid discovery, create the image effects that are prominently displayed in the app, schedule a curation for the future (we use Sidekiq for this), and provide suggested cards to editorial as they curate.

Here’s a screenshot of our tool in action:

Now that we’ve launched, what’s next? The biggest theme we’re looking to tackle next are smart ways to publish more and more of the content our users have demonstrated they love. An added benefit of this is that the more content we have, the more fine-tuned we can be about displaying the content that you care about.

If you have any questions about our tools, technologies or methodologies that we used to build these tools, feel free to email me at [email protected]