Seine

By my estimate, the new and improved version of Seine is now more than 65% complete.

And I can already say that this rebuild was absolutely worth it.

One of the biggest decisions I made was to completely move away from third-party search APIs and build Seine around its own search index.

That sounds simple when you put it into one sentence.

It wasn't.

Once you stop relying on somebody else's search infrastructure, you suddenly have to build and control much more yourself: discovery, indexing, deduplication, storage, updating data, scheduling, processing large amounts of websites and making sure the whole thing can actually run reliably.

It turned out to be one of the most difficult parts of the new Seine.

But it also changed the product completely.

Seine is now becoming a genuinely independent system. It doesn't depend on an external search API, and the core product doesn't rely on LLMs either.

For me, this is a very important milestone.

Fewer external dependencies mean lower operating costs, more control over the product and, ultimately, a lower cost for Seine customers.

But cost is only part of the story.

Having our own search index means Seine now owns one of the most important parts of its infrastructure.

And that opens a lot of doors.

The new version of Seine is also built on a completely new and much cleaner architecture. Previous versions taught me a lot about what worked, what didn't, and what the product actually needed.

This time I was able to rebuild the foundation properly.

The result is a faster, more stable application and, even more importantly, a much better base for everything I want to build next.

Another major improvement is the way Seine detects problems on websites.

I don't want Seine to become another tool that gives a business owner a huge list of technical warnings and leaves them wondering which ones actually matter.

There are already enough tools doing that.

Seine is focused on problems that can genuinely hurt a business.

Things like:

• a customer being unable to book an appointment

• a contact form that doesn't work

• an important CTA that leads nowhere

• a broken customer journey

• a major plugin failure

• serious website errors that can directly interfere with conversion

That distinction is very important to me.

A minor technical imperfection is not necessarily a business problem.

But if a potential customer is trying to contact you, book something, request a quote or buy from you and the website stops them — that is a real problem.

And that is exactly the kind of problem Seine should find.

With every version, Seine is becoming a more mature and more capable business tool.

And I'm proud of where it is going.

The interesting thing is that Seine has already grown beyond the original idea of website QA monitoring.

Once the platform has its own growing search index of business websites, that infrastructure can be used for much more than simply monitoring websites for errors.

There are many other products, datasets and intelligence tools that can potentially be built on top of it.

That is probably one of the things I find most exciting about the new version.

The foundation is no longer limiting what Seine can become.

There is still plenty of work ahead. I'm at roughly 65%+, not 100%.

But Seine is already a much bigger, stronger and more independent product than its predecessors.

And yes, I'm genuinely proud of that.

If you're building something around website intelligence, monitoring, business discovery or lead intelligence, I'd be interested to hear what functionality you would add next.

I’m building Seine, a lead intelligence tool for website care.

The idea came from a pretty simple problem: most outreach starts with a list of businesses, but not with a real reason to contact them.

I wanted to reverse that.

Seine looks at public business websites, finds practical issues, and tries to work out whether any of them are actually worth reaching out about.

That can be things like broken CTAs, contact or booking problems, rendering issues, usability friction, or other problems that affect trust, enquiries, bookings, or the customer journey.

The part I’m spending most of my time on is avoiding weak leads.

Finding an issue is easy. Deciding whether it is real, relevant, reproducible, and commercially meaningful is much harder.

So Seine collects evidence, classifies the issue, and only surfaces the stronger cases.

The goal is to make outreach more specific and useful.

Instead of:

“We help businesses improve their websites.”

Something closer to:

“We found this specific problem on your website, and here is why it may be affecting customers.”

Seine is currently in alpha, and I’m still improving the discovery, QA, evidence, and lead qualification parts.

I’d be interested to hear from anyone working in website services, CRO, QA, agencies, or lead generation.

Do you think having a real website issue before outreach would materially improve response quality?