I have spent a lot of time removing friction from the beginning of my work.
When an idea arrives, I can say it into Capture without deciding where it belongs. When I want to build, I can bring the useful context into Hermes and start from the real thought instead of reconstructing it later. Those two parts have become fast enough that I rarely lose an idea because I could not be bothered to open the right app.
That should have made everything downstream easy. It did not.
The easier it became to catch thoughts, the more clearly I could see the work I was still avoiding. I had solved the moment of capture, and I had made building much faster, but writing and publishing still had their old weight. Outreach and engagement were worse. I could preserve a thought and turn it into software, yet still leave the finished idea sitting where nobody else could encounter it.
I noticed this while walking and talking into Capture. I had been recording several thoughts about writing and friction. They were landing in a thread about friction, while I also had a writing thread that looked almost empty. For a moment, I thought Capture had failed to connect the same thought to both places. Then, while I was still speaking about the problem, it did exactly that.
The small correction was funny, but the larger problem remained. Saving a thought in the right thread does not write the article. Even perfect cross-referencing would only make the source material easier to find. I still have to decide what I mean, which parts belong together, and whether the result deserves to be published.
The blank page was the wrong place to begin
This is where my perfectionism usually enters the process.
Sitting down to write creates a strange pressure. A blank page appears to demand a finished argument from the first sentence. I can spend too long trying to find the correct opening before I have allowed the rest of the thought to exist. Speaking is different. While I walk, I can approach the same idea from several directions, repeat myself, discover that two thoughts belong together, and keep moving when a sentence comes out badly.
The raw speech is not an article, and I do not want to pretend it is. It contains detours, transcription errors, unfinished connections, and statements that sound more certain than I mean them. But it gives me something much better than a blank page: material that came from an actual line of thought.
The walking journal is only source material
That led me to a possible writing ritual. I can take a long walk and speak slowly, using Capture as a walking journal. Instead of demanding a polished article from myself at a desk, I can develop the argument while moving. Capture keeps the fragments and their context. Later, I can bring the thread into Hermes, use AI to help research and structure it, and then read and edit the article myself.
The point is not to automate my voice. I do not want a model to convert a transcript into a generic essay and publish it under my name. The point is to move the difficult work to the place where I do it more naturally. I think better while moving and speaking. I judge better when I can see the full draft. The system should let each part happen in the conditions that suit it.
The video from Eden about writing long-form articles gave me a useful way to sharpen this process. His argument is that most of the work happens before the sentences: researching, collecting unfamiliar ideas, testing the promise, and outlining the structure. If the writer is staring at a blank page, the source material and the outline probably are not ready.
That fits what I am discovering with Capture, but I would adapt it to my own process. The walking session gives me the personal observation and the live reasoning. The thread gives those fragments continuity. Research can then challenge the idea or add a concept that I would not have found by repeating my own thoughts. The outline decides what the article promises and whether the ending actually delivers it. Only then should the prose be shaped.
This matters because fluent AI prose can hide weak thinking. A model can make an underdeveloped idea sound finished. It can produce a neat title, three balanced sections, and a conclusion that feels inevitable even when nothing new has been said. That is exactly the kind of writing I do not want.
A better system should preserve the unevenness that reveals where the real thought is. It should show me which observation came from my life, which claim needs evidence, which connection appeared only after several walks, and which line is merely an attractive sentence. Research and structure should deepen the material before editing makes it smooth.
The title matters for the same reason. I do not want the clipped style that has become common in AI-written posts, where every phrase is separated into a little declaration and the title sounds like a machine arranging suspense. A good title should sound like a person naming the actual experience. It can create an open question, but the question has to belong to the article.
Publishing is part of the product loop
The same standard applies to publishing. I do not need a different idea for X, Substack, and the Capture site. People choose different places to read. One considered article can live in several places if each version points back to a clear canonical home and the presentation belongs to the platform.
The Capture site is the natural home for writing produced through this process. These articles are not documentation and they are not generic product marketing. They are public Threads: a finished piece shown alongside the source moments that formed it. A reader can see the polished article, but also the earlier fragments and the current summary. The page itself demonstrates what a Thread is without turning private notes into content.
That distinction is important. I am not going to publish every capture. Most rough thoughts should remain private, and many are not worth anyone else’s time. The public artifact should be chosen and edited. Showing a few source moments is useful only when they help the reader understand how the thought developed.
Once the article exists on the Capture site, the same HTML can move to Substack and the same text can become an X Article. Automation can remove the repetitive formatting and copying, but publication should still pass through a deliberate approval. Removing friction from publishing does not require removing judgment from publication.
The part I cannot learn from analytics
The final friction is finding people.
This is the part I have historically been weakest at. I do not naturally follow up, keep contact, or ask people repeatedly to try something. Building another feature is comfortable because the work responds immediately. Outreach creates uncertainty. The person may ignore it, misunderstand it, or use the product in a way that reveals the premise was wrong.
That is also why the work matters. Analytics can tell me that people reached the site and opened the app. Privacy-safe events can tell me that a capture sorted successfully without recording what the person said. They cannot tell me whether Capture became useful in someone’s real day or whether the thought became anything afterward.
I do not think the answer is to bribe people into posting about Capture for extra daily captures. That would optimize for public claims before I know whether the product earned them. A better exchange is based on real value: if someone creates something through Capture and wants to share it, the product can help them publish the artifact and show the process. They receive a useful piece of work and a clearer way to present it. Capture receives proof grounded in use rather than a forced endorsement.
For now, the next step is smaller. Use the walking journal to produce one article at a time. Bring the thread into a research and outlining pass. Edit until the argument sounds like me and the promise is honest. Publish the same finished work on the Capture site, X, and Substack. Then talk to the people who use the playground and learn what happened after their first successful capture.
I built Capture to keep ideas from disappearing. The work now is to make sure the ideas do not merely survive. They have to leave the private thread, become something worth reading or using, and meet another person who can disagree with them.