All the jokes about OpenAI killing startups with each new release have some validity.
We just removed Pinecone and Langchain from our codebase lowering our monthly fees and removing a lot of complexity.
New Assistants API is fantastic ✨
The laptop animation on is one I’m most proud of.
The secret behind it is our <Sticky /> component that I’ve used everywhere for over 5 years now.
Here’s how it works:
Building a Stock Screener is incredibly complex.
The amount of data you need is only half the equation. You also have to make it searchable and simple to use.
What’s the trick to ours?
We added more financial history to Fey.
This came with the challenge of building a table that could accommodate unlimited columns.
If you’ve ever built HTML tables you know the pain.
Here are some details and how we built it:
How we made Fey feel fast ⚡️
Our Canvas was painfully slow to render — slowed down by seven charts and data fetching.
Despite my efforts, everything I tried kept failing.
Until I decided to go simple and to my surprise it worked:
I know there's a belief that "SPAs are good for apps / dashboards".
There's a SaaS app I use which sends me notifications. Each time I click their links I stare at their fancy spinner for multiple seconds.
Unlike when someone sends me a e.g.: GitHub issue or PR I see the
I've received a lot of questions about how we do the text animation for our new SEC filings feature so here’s the breakdown.
It’s probably hackier and more straightforward than you’re thinking.
Code below ⬇️⬇️⬇️⬇️
Yesterday we released all the logos we use in Fey for anyone to use!
The website is simple, but getting all the pieces in place wasn't.
Here are the most interesting technical details:
Next.js Incremental Static Regeneration (ISR) is extremely underrated.
It's how we instantly generate shareable links in Fey.
The API is as simple as it gets.
Let me show you how it works:
Charts are a core to Fey and everyone asks what we use to build them!
The short answer is we use .
But I'll breakdown some components and code we have to make our charts.
How we changed Fey’s auth to load instantly using Next.js (without SSR).
Our stack
- Next.js latest (pages router)
- Firebase auth
Before we’d run auth before rendering any page which meant we always had 100-400ms splash screen.
Gradient border with CSS:
background:
linear-gradient(black, black) padding-box,
linear-gradient(to right, blue, red) border-box;
border: 1px solid transparent;
The gradient can be any type: conic, radial, linear, etc.
Codepen:
10 years later I still think about the Mac Pro landing page!
When it was released in 2013 I remember being so impressed by the video, assets, and techniques they used.
To this day I still think it's one of the best landing pages ever made.
I wrote about Frontend Engineering and why it's so special. It's a magical role that I love.
Why is it so special and what makes a great Frontend Engineer?
Read my thoughts:
So you want to build a Command?
In Fey, the Command is at the centre of the experience. Getting it right came with a lot of challenges.
Here’s how Fey’s works and details:
Everyone I talk to is moving over to Cursor — the AI-first Code Editor.
Seems like
@anysphere
really nailed it. You can even add your OpenAI key to save costs.
Anyone use it? Recommend?
February was intense.
We went heads-down and focused on operational efficiency, costs, and infra.
The biggest surprise? We were able to use AI to replace a vendor costing $12,000/year.
Overall we reduced our monthly expenses by 60%.
Fey’s Share feature is here 📈
Copy any chart to your clipboard — instantly.
Generate a link with a click of a button.
The journey to build this feature was a rollercoaster!
Here’s how it went down:
We launched "For you" and it's a foundational step towards making finance effortless.
The feed surfaces news, filings, ratings, earnings, and more all based on your watchlist.
Here's how we built it:
The simple tricks are always the best.
The original version of this hero didn't have the faded text as idea only came to us at the very end.
Here's how it's done:
New charts for EPS and Revenue in Fey!
Out of all the charts we've built I think these turned out the best. So many deign details that translated beautifully to production.
A peek into how it's made:
When you use the same tech stack everywhere it unlocks so much productivity.
We recently added a section to .
It was as simple as copy-pasting the code over and applying a CSS transform.
The code is very simple to reuse 👇
@rauchg
@shuding_
I was curious how they did this as well on the iMac page from before. Ended up recreating it here:
Load up all the images, render to canvas, scrub through using scroll. The real magic with these effects comes from the assets. Apple has the best assets.
We recently deployed a massive improvement performance and it's hard to understate how much of an impact it makes to the entire app.
We went from 2-4s to 0s load time! (without SSR)
CMD+T → F → Enter → instant load
In almost all my projects I now add a Breakpoints component that displays the breakpoint name and the pixels it's active from. (bottom right of gif)
Working on creating a lot of developer experience components to really speed up how we work.
Source:
My morning routine since building Analyze
• Open Fey
• See what's popular
• Hit X and get 10-Q summary
• Hit N if I want more news
• Hit C if I want an overview
A delightful way to informed about earnings, markets, and trends.
💖
1. Create a 1px by 1px dot using background radial-gradient. Then define the background size and position to create the infinite repetition.
2. Then apply a mask to that dotted grid so it smoothly fades out towards the edges
That's it!
Sometimes you have to work smarter not harder.
For analyst ratings, we exported a static SVG for the background instead of fiddling with the graphing library.
3 minutes of work vs 3 hours+
With
@nextjs
Route Handlers on
@vercel
, setting up live pricing was super simple on our marketing site 🖤 This used to be a complex custom API from our backend but now it���s a piece of cake.
Here’s a breakdown of how we use the new
@OpenAI
Assistants API in Fey.
We actually just had a call with their product and API engineer discussing some of these details:
AI keeps taking over Fey.
In our For you page, there's a section where we surface a summary of the top news every hour.
It's able to parse news, generate a summary, give us the tickers, and nicely formatted.
How would you automate this before AI?
2024 news in Fey will be unbeatable.
First step out today:
We've revamped our news page with an improved UI and the addition of "Events" for more concise insights.
Alright, time to give credit where it's due:
@nextjs
&
@vercel
really changed how we build Fey. Our team is tiny (it's just me on eng) but we're able to still able ship an experience that's insanely fast and easy to manage.
Will document my progress of migrating frontend to
@nextjs
here.
- Large monorepo codebase
- Client code is CRA/TypeScript hosted on GCP
- Marketing site already on Next
- Goal is to slowly migrate client code to Next (in my spare time)
Fey is fast ⏩
Our tech stack is simple (Next+Firestore)
How do we stay fast?
- pre-aggregate data into single requests
- multi-region db for latency
And then smaller details like preloading, good design, edge APIs.
@viewportui
@tcosta_co
Fey seems to be really fast! A click and boom new data is shown.
What's the back-end data techstack to deliver that amount of data at that speed?
Charts are a core to Fey and everyone asks what we use to build them!
The short answer is we use .
But I'll breakdown some components and code we have to make our charts.
The hardest and most tedious task was accomplished by
@tcosta_co
. He manually created over 1000 company logos in Figma.
This was him heads down, going to company websites, getting the logos (sometimes not SVG) and drawing them in Figma.
Huge task. You have no idea.