Writing Again, This Time It Is Different

So, we are starting again. No, bear with me, I mean it. Not the desire to write, or the desire to share. That part was always there. What is different now is the capability.

I am dyslexic and dyspraxic — ironically two of the more annoying words to have to spell when that is the hand you have been dealt. I am also older than I look, and on some days older than I feel. That means I have collected a fair few coping mechanisms over the years, most of them born of frustration, necessity, and the general absence of support when I was younger.

Writing, for a long time, meant friction.

Not the good kind of friction, where an idea is sharpened by editing. The bad kind. The kind where the thought arrives clearly enough in your head and then gets mugged on the way to the page by spelling, grammar, transposed letters, dropped words, or prose that does not quite survive contact with the keyboard. The point gets stolen by red ink before anyone can even argue with the idea itself.

I had Grammarly, and still do. It helped. It still helps. But Grammarly could only correct what was already there. It could tidy the sentence, not retrieve the thought. It could point at the mistake, not help me bridge the gap between the thing I meant and the thing I had managed to type.

That gap matters. If writing is always work before it is expression, then publishing becomes something you postpone. You tell yourself you will come back to the draft when you have more energy, more time, more patience for your own errors. Usually, if you are anything like me, that means you do not come back at all.

What Changed

What changed was not AI in the abstract. It was the combination of AI with a workflow that is actually usable.

An article is now a link, a screenshot, a photo of a book page, a note in a queue, a few rough sentences in an issue. That is enough. That is the seed.

From there, GitHub gives me a place to put that seed. The repository holds the blog. The issue or prompt holds the instruction. The agent has guard rails, conventions, and enough context to do something useful with what would previously have been half a thought and a guilty feeling that I had not written it up properly.

That is the real difference: the distance between thought and finished post is much shorter than it used to be.

I might still write the stub. In fact I usually do, because voice matters and intent matters. But I no longer have to do every laborious bit of turning that stub into publishable text while also fighting my own wiring. The agent can take the rough material, follow the workflow, create the draft, open the pull request, and leave me with the part I actually want to do: review it, tweak it, make sure it sounds like me, and then publish it.

Less Friction, Less Fear

There is something oddly emotional about removing that much friction from a process that used to be hard.

No more forgetting the idea because I did not have the energy to wrestle it into shape on the same day it arrived.

No more rereading my own work with that sinking feeling that I have missed something obvious again.

No more disproportionate anxiety about spelling and grammar drowning out the actual point.

That does not mean no work. It does not mean the machine has replaced writing. It means the bureaucratic tax between having a thought and sharing it has dropped dramatically.

For me, that is not a gimmick. It is accessibility.

I can now move from prompt to draft, draft to pull request, and pull request to published post in minutes rather than in a long, stop-start cycle of avoidance, correction, embarrassment, and eventual abandonment. A workflow runs. I review it. I adjust tone and emphasis. I click the button. A little later the post is live.

That is a very different experience from the one I had when I first started blogging, and from the several later attempts where I tried to start again and quietly fell off.

It Still Has To Sound Like Me

There is an important caveat here, and I do mean important.

I do not want the machine to become my voice. I want it to help me reach my voice more reliably.

That distinction matters. Earlier this year I wrote about an experiment in Refactoring Agent Workflows, where a post produced by an agent was technically good but did not really sound like me. It was too neat. Too polished. Too obviously machine-shaped. That is still the risk, and it is one worth watching for. The value here is not in producing generic text quickly. It is in reducing the friction enough that I can stay involved all the way through and keep the piece mine.

So this is not really a post about AI hype, and it is not especially about tech news either. It is a post about what it feels like when a tool finally meets you where you are instead of asking you to overcome yourself before you are allowed to use it.

For someone with dyslexia and dyspraxia, that is not a small thing.

It is the difference between wanting to write and actually writing.

A Small Moment Of Reflection

This blog has had more than one beginning. That is probably true of most personal sites if you leave them around long enough. You start, you stop, you mean to come back, you do not, and then one day you do.

This time, though, it does feel different.

Not because I suddenly became better at writing in the conventional sense. Not because the fear vanished completely. Not because every post will now be effortless.

It feels different because the path is clear now. The support is there. The workflow exists. The friction is lower. The thought has a better chance of surviving the journey.

For the first time in a long while, that feels like enough.

Further Reading

• Refactoring Agent Workflows — my earlier note on agent workflows, including the problem of a post sounding too little like me.
• AI Burnout and the Loss of an Ideal — a more critical piece on where AI tooling goes wrong when it replaces judgment rather than supporting it.
• Agents on GitHub — the general platform model that makes the “seed → workflow → pull request → review” loop possible.