Skip to content

The Laboratory

5 Topics 5 Posts

Where things break and that’s the point. Active experiments, workflow development, and the honest documentation of failure.

Subcategories


  • Ongoing experiments with clear hypotheses and live results. Document as you go, not just when you succeed.

    2 2
    2 Topics
    2 Posts
    Obliqo is growing. Slowly, imperfectly, but for real. [image: view?project=69aeb0e2000f974381fc&mode=admin] And I want to say something clearly: without an AI copilot, I would not have been able to build this alone. That does not mean you press a button and a product appears. It means daily study. Confusion. Debugging. Wrong turns. Rewrites. Retesting. Small breakthroughs surrounded by friction. What I am discovering is not just that AI helps me move faster. It is that, in my case, building with an AI copilot has become a different way of learning while building. Not passive. Not automatic. Not effortless. More like a continuous cognitive exchange: I try, the machine responds, I correct, it expands, I resist, it proposes, I study, I decide. But that exchange is not inherently trustworthy. Sometimes the copilot is useful. Sometimes it is shallow. Sometimes it is confidently wrong. Sometimes it gives me something plausible enough to slow down my own thinking. So the real work is not “using AI.” The real work is judging, testing, rejecting, reformulating, and learning enough to know when not to trust what looks convincing. That is why, for me, this process does not feel less human. If anything, it demands more: more clarity, more responsibility, more patience, and more honesty about what I actually understand versus what I am only borrowing for a moment. I am not presenting this as a universal path. Not everyone has the same access, the same technical starting point, or the same conditions for working this way. I am only saying that this is what I am living through while building Obliqo from zero: a form of learning-through-construction that would have been inaccessible to me without this kind of AI partnership. That is also why I do not think this process should remain a black box. It should be opened, examined, shared, criticized, and made more accessible to people who want to change their lives not by consuming answers, but by learning in the middle of real work. So I want to start sharing that process here from the beginning, including the mistakes, the dead ends, and the parts that still do not make sense. If Pyragogy means anything, it has to survive contact with real work, real confusion, and real construction.
  • n8n, OpenRouter, and multi-LLM orchestration. Share your flows, your failures, and the weird things AI does when given autonomy.

    0 0
    0 Topics
    0 Posts
    No new posts.
  • Scripts, plugins, integrations, and infrastructure. If you’re building something for the Pyragogy ecosystem, document it here.

    3 3
    3 Topics
    3 Posts
    For the past months I have been building Obliqo as a solo founder — and tonight I want to share the thing more than the launch, because the launch is the small part. Obliqo exists because of something this community has named for years. The AI gives you text that looks finished before the thinking behind it is. You publish faster than you can verify. That is the gap. What I built is a small extension that runs four agents over the draft you have just written — inside the tab where you write (Gmail, a PR description, the body of a post). They do not rewrite. They do not flatter. They tell you where the draft does not hold. Then they leave you with a question only you can answer. I built it because I needed it. Not in the dogfooding sense from product talks. In the cruder sense — the dogfounding sense — that for months I was the first user of a tool I had not finished, working in conditions where I knew I would publish badly without it. Necessity under pressure. The product is the sediment of that contradiction: I built a tool against frenzy from inside the frenzy. The extension is live now: Chrome Web Store. The webapp is at obliqo.pyragogy.org. One small note about Chrome: when you install, you will see a warning that the extension is “not trusted.” Nothing dangerous. I am a new developer and Google extends trust over time. Chrome is asking me to earn it — which is also what I am asking the writer to do, with their own drafts, before they ship. Fair enough. The blog has the longer version of this story, with the contradiction left open: I Was the First One Who Needed It. I do not have all the answers about how this scales beyond my own case. I am hoping some of you will find a way to break it, and tell me what you found.