Recommend you change this footer text: Instantly create your dream vacation with Duebase AI, where AI-driven precision meets personal preferences. Tailor-made travel plans crafted just for you, effortlessly.

Thanks!!!

Also under the heading “Comprehensive Accounts Analysis” you have “RetryClaude can make mistakes. Please double-check responses.”

And I'm curious: was this a pivot from another industry, or did you just try out Claude on a dummy project?

> I use Claude to help me with the copy

“Help with” implies you do some of the work yourself. But most of the comments in this thread so far are about glaring copy mistakes. Have you not read any of the text Claude gave you? That doesn‘t inspire any confidence that your products is remotely competent and what you claim.

> Our AI analyzes balance sheets, cash flows, and profitability metrics to deliver executive summaries, risk assessments, and key insights that help you make confident business decisions.RetryClaude can make mistakes. Please double-check responses.

Slight issue on the homepage ;)

#vibecoding

Hey I had an example URL in a README.md with github.com/yourusername. I think this is going to bite me in the ass if anyone checks the commit that fixes it.

From the footer:

“Instantly create your dream vacation with Duebase AI, where AI-driven precision meets personal preferences. Tailor-made travel plans crafted just for you, effortlessly.”

Did you pivot?

Looks like OP still does that: https://news.ycombinator.com/submitted?id=superproton

If we want to be charitable their previous website may have served as a template.

Well, at least this one made it to the front page!

Any stats on hallucinations or just general inaccuracies in the final results it generates?

Seconded. Throwing a bunch of data into an LLM and calling the result financial analysis is one thing, the results being at all trustworthy and valuable is another. I’d like to see an independent piece of analysis compared to this bot’s output for the same company and the original data.

On an even more basic level, throwing data at an LLM and telling people you "built an AI" pretty much blows any chance of anyone taking the results seriously.

There's a copy pasted "Retry Claude" sentence in the copy of the page, the footer advertises unrelated travel services. Oh, and there's no details on what methods the AI uses for analysis, or any measurements at all of an unproven AI analysis tool.

This is pretty bottom of the barrel.

I use Companies House (CH) a fair bit so interesting, tried out:

- There is UX issue here as on first pass it just looked like a view to CH, the scores where N/A, after a minute it came to life with something useful but during that period wasn't sure what was going on, you need to manage the user and let them know its working out scores not show N/A.

- Liked the history of scores as well, perhaps a chart would be good here.

- I need to understand how you score before I would ever consider paying for this, there has to be some explanation of the methodology.

That's great feedback, thanks a lot!

UK company filings are already marked up with machine readable, structured, standardised tags. It's called XBRL. So it doesn't necessarily need a LLM to parse.

Wrong, my friend, that is only for small companies, big companies accounts are only available in PDF. Plus, in any case you have the raw data, here you have a full financial analysis.

Large companies need to do it too, including all LSE listed. Here is a summary by the regulator https://www.frc.org.uk/library/digital-reporting/structured-...

But of course a LLM could theoretically automate much of the analysis stage.

Aren't there services that provide normalized financial data? I don't know about the UK, but I've seen in other markets. If your USP is more towards analytics than data wrangling, it might save you a lot of headache to offload that to someone else.

Good point. There are a few (Creditsafe, Experian) but they're expensive and focus on basic credit metrics, not comprehensive financial health analysis. The real value is in the contextual interpretation - understanding what the numbers mean for investment/credit decisions. That requires domain knowledge baked into the analytics, not just clean data feeds. Plus controlling the full pipeline lets us iterate quickly on scoring algorithms based on user feedback. Hard to do that with third-party normalised data.

If these are already expensive, there must be a lot of value for the data itself? Why not make that your core offering? If you're able to automate the process and provide more data on top, you'll definitely have an edge over them.

Using an AI for this is like using crypto to pay for something: only done for the novelty

Regular coding, the SECs API (or UK equivalent), and some simple spreadsheet formulas have already done the brunt of this for years now.

Fair point - I started with the Companies House API and traditional parsing too. The issue isn't data access, it's consistency. UK companies file in wildly different formats - some use full GAAP, others micro-entity accounts, many have incomplete data. Rule-based systems work for ~60% of companies, then break on edge cases. Financial health also requires context: Is this debt ratio concerning for their industry/size? Are these cash patterns seasonal or declining? How do you weight profitability vs growth stage? I spent months on traditional approaches first. The AI handles the inconsistency and contextual interpretation that spreadsheet formulas can't.

I respect your comment but all of these things can be automated without an AI

The inconsistency of UK reporting is highly interesting though

Thank you Claude, good job!

The heading text is cut off by the top navigation (on mobile safari).