Who is CA?

The rejection years.

CA spent years trying to get hired as a developer. The applications. The interviews. The rejections. The feedback was always polite and always the same: "You haven't shipped at scale." "We need someone with production experience." "Your background is interesting, but..."

The "but" was where the rejection lived.

CA told himself a story about what that meant. The story was: I'm not good enough. If I have the degrees and the knowledge and the projects, and I still can't get hired, the problem has to be me.

He believed that story for years. It shaped what he applied for. What he said yes to. What he believed he deserved.

Then AI arrived. And the industry started telling developers the same thing it had told him: you're not needed anymore. The work can be done faster and cheaper by something that doesn't need a salary or a break or a life.

CA watched. He recognized the feeling. And he realized the story had been wrong. He wasn't rejected because he wasn't good enough. He was rejected because the door was closed and nobody told him there was another way in.

The pivot.

CA redirected. Technical writing. Freelance work. The thing that paid and didn't require anyone's permission. And in the evenings, he kept building. Not for a career. Because he was curious.

WordPress first. Then headless WordPress, which taught him what decoupled actually means. Then GraphQL, because a client's API used it and he needed to understand it to document it. Then a router to access cheap inference, because he wanted to test AI models and couldn't afford the API bills. Then local models. Then MCP servers. Then agents.

Each thing led to the next. The curiosity never ran out.

The self-taught years.

CA learned to code the hard way. Trial and error. Late nights that turned into early mornings. Long hours spent debugging something that turned out to be one missing line. He knows Python, JavaScript, HTML, CSS, SQL, and C#—though he's used C# mostly for console projects and a few small experiments. He's tried several databases over the years. He runs a few hobby sites with a purpose he created himself. He just finished a couple of apps he's proud of. One has been tested and will be deployed soon. The other needs refinement.

A note for the poetry readers.

If you followed CA for poetry or creative writing, thank you. That was a different chapter. This is the new one. The focus now is tech—specifically, honest testing of AI tools and the failures nobody documents. A poem may still appear occasionally. A short story might slip through. But the main thread has changed.

CA is a writer of different things. Poetry. Creative stories. Travel writing from his own trips. Current affairs. Tech. The throughline is curiosity and a need to explain what he finds. For the past few years, that curiosity has been pointed at AI tools, infrastructure, and the gap between what the demos promise and what actually happens.

The Unsponsored Stack.

The Unsponsored Stack is where the receipts live. Not benchmarks. Not demos. What actually happens when the tool meets real work.

The streaming cutout at 80% because of an MTU mismatch. The credits that ran out mid-refactor and forced a switch to a model that didn't know the architecture. The MCP server that returned empty strings. The delete endpoint that returned success but didn't delete. The vector database installed for twelve documents. The cognitive debt of code that runs but can't be explained. The 90-second cold start of a local model on a refurbished laptop. The $187 in subscriptions replaced by a $12 stack.

CA doesn't get paid to promote tools. No affiliate links. No vendor relationships. No sponsored posts. That's not a virtue. It's just the situation. But the situation turns out to be the qualification. The unsponsored truth is different from the sponsored version. Not because sponsored writers are dishonest. Because they're constrained, and CA isn't.

What CA actually does.

CA tests tools on his own projects, on his own time, with his own constraints. The constraints are different from the constraints of the people who write sponsored reviews. They test on clean setups. CA tests on setups where the credits ran out last week and he's routing through three different providers because he can't afford to pay for one.

They test the happy path. CA tests the failures. The streaming cutout. The empty response. The context loss when you switch models mid-task. The missing index that made the query slow. The delete endpoint that lied about success.

Those failures are what he writes about. Not because they're interesting. Because they're what actually happens. And almost nobody writes about them because almost everybody writing about tools has a reason not to.

The philosophy.

The professional has a deadline. CA has a Tuesday. That's the difference. The deadline says "ship it when it works." The Tuesday says "find out when it doesn't."

The deadline is rational. The professional has a job. The job requires shipping. The shipping is what gets paid for. The understanding is what gets done on the professional's own time, if there is any.

CA has nothing but time. Not in a sad way. In a way that makes the investigation possible. He can spend three evenings on a problem that might not matter. The professional can't.

CA is the one who finds the failure mode. The professional is the one who reads about it later.

The other interests.

When CA isn't testing tools, he's reading about current affairs, writing about his travels, taking long walks, or occasionally returning to poetry. He lives in a Scandinavian country. The long walks are essential. The poetry might return.