I built a 516-panel financial terminal in 3 weeks using AI | Hacker News Reader

I built a 516-panel financial terminal in 3 weeks using AI

neuberg.ai

45 pointssaratsai43 comments

I built Neuberg because I’d always wanted a Bloomberg Terminal, but as an independent trader, I just couldn’t justify paying $24,000 a year for it. So I decided to build my own with Claude Code.

It’s a real-time trading terminal covering fixed income, derivatives, commodities, equities, credit, macro, and alternative assets, with 516 panels in total. You can freely drag, drop, and arrange them however you like.

What it does: Scrapes financial news and uses AI for sentiment analysis, including first- and second-order impact detection Offers 516 market data panels, such as Treasury auctions, option surfaces, CLO analytics, GDP nowcasting, and dark pool data Lets you trade directly on Hyperliquid (49 stock perpetuals) and Polymarket (prediction markets) Supports stock trading through Alpaca, both paper and live Includes TradingView-style charting with indicators like RSI, MACD, Bollinger Bands, ATR, and VWAP Features a world map showing geo-located news events and conflict-zone heatmaps

A few honest caveats: Most of the 486 newly added panels currently use seeded or simulated data. The UI is functional, but not all of them are connected to live data feeds yet. The core 30 panels — including news, charting, trading, heatmaps, calendar, options flow, and insider trading — use real data. This is a solo project, so there are still some rough edges. It’s licensed under BSL 1.1, which means it’s free for non-commercial use.

In about three weeks, I expanded the platform from 30 panels to 516, mostly by vibe-coding with Claude Code. Around 90% of the panel code was written by AI, while I reviewed and integrated everything. I’d really love feedback on the architecture, UI, or panel coverage, and I’m happy to answer any questions.

Source code: GitHub - KoNananachan/Neuberg: Real-time trading news terminal with AI analysis, prediction markets

If I understand correctly, you had 30 working panels for the core functionality. You then vibe-coded an additional 486 new panels that use fake data? That's not quite how you presented the project in the title.

What do you mean? It says used AI in the title. That already implies it doesn't actually work.

Strong disagree on AI implying it doesn’t actually work. Staff engineers at FAANG are vibe coding these days and their stuff absolutely works.

> Staff engineers at FAANG are vibe coding these days and their stuff absolutely works.

60% of the time it works 100% of the time

You really want to claim that with the amount of outages lately?

We are trending towards zero nines (not 100) uptime.

I use Gemini CLI and given the might of all their engineers and AI tools, the features they ship are constantly buggy and broken.

It's so easy to pump out borderline useless functionality with AI that nobody is actually testing or designing it.

Most of the value of the Bloomberg terminal is the data, not the number of panels

as someone who worked at TR (now Refinitiv) and then Bloomberg, it's not even the data. it's IB. After IB it's the data. :)

As someone that worked for the spin out from GS, then got acquired by TR, then sold off to Blackstone, I wonder how much of our tech debt is still there for you to deal with 10 years later. :)

I haven't been at TR in 10+ years as well. AT TR, I mostly dealt with Eikon's pipeline for estimates, actuals and the like (IBES data) ingestion, querying and analytics. So can't say much, though some of my old colleagues still seem to be there.

I'm not sure if this is a coincidence or not, but today my github account was removed from the organisation... after 10 years.

As someone who never worked anywhere: what’s IB?

Chat.

https://professional.bloomberg.com/products/bloomberg-termin...

Nonsense, when he'll have one panel per screen pixel, he'll be able to see over 8 million Fibonacci retracements, 40 heatmaps and real-time market sentiment headlines at once on a 4k monitor, then you'll see.

... and speed. They publish billions of ticker items in near real time simultaneously to all their users... the tech is pretty crazy!

it's all about the bas

this.

The thing here is I don’t think you’re going to be able to get your hands on the data that the Bloomberg terminal provides especially not near real time live streaming data which is the reason people pay $25,000 a year. You’re not really paying for the interface, that is easy.

Can you please get rid of all the fake data panels? I will use it but only if I can trust that the data I am looking at is real.

I’m curious about the UI development. What sort of prompts did you give? Ie look like this or look like that? Or did you specify the qualities?

This is great work! Where are you getting your data from? Finding quality data sources is difficult

As far as I know, finding quality data sources is virtually impossible unless you already have so much money that your involvement in the stock market is either because you have insider knowledge or severe boredom owing to your never needing to work a day in your life.

I remember some startup selling a stock market data API as a subscription. I don’t think they exist anymore. So anyone who spent weeks, months of their free time building an app around that API is now completely shit out of luck.

I suspect the real APIs are still running the same code they ran in the 90s and if you have to ask how much they cost, you can’t afford them.

You can buy historical data at least from some reputable vendors, although you're still responsible to understand their collection process (sampling frequency, timestamp conventions, corporate action adjustments etc etc), as even 'obvious' things like how daily stock levels are reconstructed based on intraday data can mess up your analytics really bad. I don't think it generalizes to real-time data.

You can get real-time market data through Databento. Been doing it for years. Top of book suffices for most people.

I open the site and nothing is loading and all I see is a grid

Just curious if “number of panels” is a genuine metric by which financial terminals are judged?

The only real metric by which financial terminals are judged is "Why don't we have any bloombergs? The bond traders are throwing a hissy fit."

This.

Surely it gets noisy after 30 or so, but pedestrian stock trading apps like Thinkorswim have very high levels of customizability and modularity. I think even extensibility in Thinkorswim’s case. Java I think. Anyway, I would think any users of OP’s app are not using 90% of the 500+ widgets.

Panel is an arbitrary UX delimiter, so fundamentally no, unless you're really pedantic in defining (upfront) panel as a meaningfully semantic unit across apps.

Not really. The quality of data and latency are paramount, that's what you are actually paying for.

This is cool, but my experience with trading stocks is that you need to spend money in order to be on a remotely-level playing field with the whales. Ignoring that so much stock market data costs money, there’s also the matter of direct fiber connections to exchanges, brokers who batch your order with others resulting in subpar fill prices relative to bid-ask spread, the pattern day trading rule that only affects poor people, the simple fact that the 1% have insider information and use it to trade stocks for a guaranteed win… really having a dashboard of charts is the least valuable piece, though I admit it is a piece.

I could be mistaken but I thought the reason Bloomberg Terminals are so expensive is because they come with the expensive, real-time data feeds.

The real moat of the Bloomberg terminals is the messaging feature that connects you to the other traders.

Fascinating. It’s called Instant Bloomberg and it’s just AOL instant messenger but only between Bloomberg Terminals. So that is absolutely the “data that insiders rely on that pedestrians will never have” and I’m sure it offers tremendous advantage and may explain why stocks trend in a direction for extended periods of time absent any public news.

I mean there’s no way average joes like us stand a chance doing anything but dollar cost averaging into index funds.

E: imagine if there was a law passed that required those IMs to be public in near-real-time (releasing the transcripts days later defeats the purpose).

I disagree on us average joes not standing a chance. It's pretty easy to ride the wave of popular stocks or commodities simply by following the macro trends in the news and staying informed on global developments.

It's more risky than index funds, but being able to ride a bit of the Nvidia hype and then de-risking those proceeds into an index fund is nice.

You'll probably never win with day-trading or HFT though.

(not financial advice!)

That's a cool slot machine reskin

Even if the data were accurate, I'm going to stick with the cheapest vanguard broad market mutual fund. Almost nobody can beat that, if you have a decade or more.

https://en.wikipedia.org/wiki/Jeffrey_Tarrant#Buffett_bet

So refreshing for me to see a high data density display. It highlights for me how the norm is to display more “white space” than information.

Reminds me of the palantir clone: https://www.youtube.com/watch?v=rXvU7bPJ8n4

This is the year when solo software firms start to dominate.

I just get a black empty grid (brave browser on iphone)

very cool , i wish it had multi screen mode

wait to release until it uses real data?

Reply on news.ycombinator.com