sielski.uk

I can't sleep, I ship

I saw a post on LinkedIn stating that the person who started using AI works less. That’s true.

Especially if you have AuDHD, you’re terrified about the current work market, got overloaded with more heads and projects to manage at work, crushed by the UK housing market, rising cost of living, and inflation making your earnings smaller each year.

That’s me. Today.

One place I don’t want to use AI is here. My dear blog, diary, relic of the 90s brought back to life.

But in all that madness, there is a glimmer of hope. Every time before, when I felt pressure, it was good. It motivated me to challenge myself, leave my comfort zone and grow. The problem was that while it was a good way to get on the next career ladder, it didn’t really scale well in becoming fully self-reliant, time-rich, and entrepreneurial.

Quite early, I realised that building stuff is easy. The problem is in finding problems worth solving. Those that people will be willing to pay for a solution. That required networking, lead generation, promotion, outreach, and research. All the things a nerdy AuDHD individual struggles with. All the things that usually you need to use a lot of time, hiring people to help, or tools that are expensive.

I took a fair share of risks in my life, but bankrolling such an endeavour without know-how and a partner to guide me wasn’t one I’d likely take.

AI has changed the odds here, though.

It’s not only easy to ship - as we did with Martin, building an app for maths- but also early-adopting parents’ main complaint is the lack of use-time limits (added now; see MathBuns). But it’s also easy to ship something that replaces all those expensive tools.

I have the benefit of knowing how software is built, secured and maintained. A big part of the cost is for the IP and their own marketing. For a starting, and experimenting micro-entrepreneur, buying a licence for Buffer, MailChimp, ManyChat and a few others is a massive weight. But if you can ship their replacement relatively quickly, safely and piggy backing on your Claude Max plan, that changes the equasion.

And that’s what happened for me. That’s what Eidan delivered for me, together with Glue, which will be released as open source soon too. I can not only automate a lot of the work, but also delegate it and scale. Mix-and-match sources and models to get different results: Work with Eidan (as a harness and memory) and Claude (as a model) to find the potential market and offering. Go to Preplexity (not connected to Eidan, yet) and dig the supporting data. Back in harness, cross-examine findings using different models like GLT 5.2, DeepSeek and whatever else. Set Claude to generate me content briefs, and get OpenAI to render brilliant assets. Then use Glue (soon available for you) to turn content into a subscription funnel, and measure engagement.

Something that would require hundreds of pounds and months of preparation now takes a few evenings and a bit of licence; I use it anyhow for my day-to-day management.

I still fail, but I fail faster. I learn faster. I improve faster.

And I’ll leave it there. While writing that unstructured post, Claude made two massive iterations in Eidan and Glue, lifted my content pipeline and reviewed a few other changes. Eidan’s node running on my Raspberry kept grinding on my Max plan to fully use it, to deliver a less hands-on approach from me. Glue sent welcome DMs to a few new followers on socials. MathBuns game is still pending review on the App Store…

I did not have any of that a month ago.

Let that sink in.

While sinking, you can just: