Claude Helped Me Build Four Platforms. Today, I Asked for More Than Credits.

Today I applied to Claude Startups.

A few hours earlier, I was joking that Claude should go and make tea.

Both things can be true.

I have no interest in turning a software subscription into a religion. I use the tools that help me build. Sometimes I complain about them. Sometimes I change them. And when a company offers something useful, I read the conditions rather than checking whether accepting it would upset my imaginary allegiance to another company.

Anthropic has expanded its startup programme. Approved companies can receive $1,000 in Claude API credits, a free year of Claude Team for up to five Premium seats if they are new to Team, access to its Applied AI team, and partner offers advertised at up to $45,000 in combined value. Bootstrapped founders can apply; venture capital funding is not required. Programme details

That last part matters to me.

I am building from Malta, through Fierce & Tame Limited. Solo.

FreeMalta. MaltaInsider. News Beast. Gambity.

I built all four with Claude. News Beast and Gambity also use the Claude API behind their AI-native newsrooms. These are not four names waiting for someone to turn them into products. They exist. They have pages, systems, decisions, and things I would do differently if I started again.

Which is usually how you know something has actually been built.

When the application asked what I wanted support with, I did not struggle to find an answer.

Source verification. Editorial reliability. Multilingual publishing. Architecture. API costs.

The less glamorous things.

Nobody shares a screenshot of a well-handled failed request and says, “Look at this beautiful bastard.”

But perhaps they should. Because once the excitement of generating something wears off, those are the things standing between a convincing demonstration and a system you can depend on.

A newsroom is a good example.

You can ask a model to write an article. That is easy enough.

Then you discover that a source does not support the sentence attached to it. A summary has quietly become a certainty. Two versions of an event have been combined. The prose sounds magnificent, but the claim underneath it is doing something suspicious.

Now you have a different job.

You are no longer asking whether the model can write. You are deciding what deserves to be published under your name.

I cannot outsource that decision and then act surprised when something goes wrong.

That is why the possibility of speaking with the people building these models interests me. I have questions that come from running things, not watching demonstrations of them.

Where should one agent stop and another begin? Which checks are worth their cost? What should happen when a source cannot be verified? How do I keep a workflow useful without turning it into an expensive procession of models congratulating each other?

Credits would help me test those questions.

But a credit balance is not an architecture.

And the headline value needs a little translation. The advertised partner total is a collection of offers, not money arriving in your bank account. Each provider has its own conditions. Something worth thousands at list price can be worth nothing to you if you do not need it. Additional benefit conditions

I do not need another subscription simply because somebody has temporarily made it free.

I need the right tools to make the systems I already have more reliable, more affordable, and easier to maintain.

For FreeMalta, that comes back to a very ordinary ambition.

Give people the information.

Not a teasing paragraph followed by a form. Not a tax explanation written as though everybody reading it has spent twenty years as a CFO. Not a useful answer held hostage until someone books a call.

Serve the meal.

If someone wants help afterwards, they can ask about dessert.

Building that experience takes work. The fact that AI helps me do it does not make the choices disappear. It gives me more room to make them and more things I need to check.

I still decide what belongs on the page. I still read the output. I still have to notice when a technically correct interface is miserable to use. And I still want the freedom to build something simply because I love it.That is how a Malta platform ends up with a door into Tolkien’s world. Nobody commissioned it. I wanted it there.

I have submitted the startup application. I have not been accepted, and this is not a partnership announcement.

For now, it is a founder arriving with work already done and questions he wants to investigate next.

I can appreciate what Claude helped me build without pretending it is the best tool for every task. I can use something else tomorrow without rewriting my history today.

That seems a healthier relationship with technology than choosing a flag and defending it forever.

Anyway, the application is in.

Claude can put the kettle down. There is still work to do.

Xoxo