I never aspired to be an author; but I have something to say

Latte Other Software & Development
, , ,
1

My name is David Condrey and before I founded WritersLogic I struggled for years with depression, anxiety, PTSD, drug addiction, and homelessness. I eventually got off the streets and become quite successful but I never learned how to be a good author. Ever since I was a kid I always wanted to be a writer though; and I recall graduating high school and considering whether I would join the military or go to college and I thought to myself if I just go to college what in the world am I going to write about…

So I joined the Marine Corps and I got out and for the past 30 years I’ve lived a life worth writing about. Unfortunately, the part that I forgot was taking the time to learn to write well enough to have it published and read by other people.

If you search me online you’ll find that I’ve been writing far longer than AI has been around. Everything from posts on Quora and StackOverflow, to Wikipedia articles, to poetry contests, and research-papers. But I didn’t start sharing the personal stories that I really cared about, at least not much - there are a few - until AI came along. Maybe because I doubted myself too much. Maybe because the stories were too painful and I couldn’t put them down on the page by myself. Maybe because I had rewritten and rewritten the essay so many times I was just sick of it and if I didn’t get help polishing and publishing it I was afraid it would end up in the garbage.

I started WritersLogic immediately after I responded to an article that the CEO of Medium had posted asking people what they should do regarding the influx of AI. I responded with a thoughtful comment about how writing that costs the person something is theirs regardless of how it was created, but writing that cost nothing was what should be audited. The internet is full of clickbait everywhere you look and the same people who ridiculed my writing as AI slush would turn around and write content like “How to Spot AI Writing” and other such uselessness.

I know a lot of people hate AI. And I will stand up and be the first to agree that someone who pastes a simple prompt into ChatGPT or noter LLM and claims ownership of the output should not claim to be the author of that content and should probably keep it to themselves. But I also believe that the desire to be a writer and the desire to be an author are not necessarily the same thing.

Sure, it would be wonderful to sell a novel and have my book sold in bookstores and be able to tell everyone I was an author. But to do that I would have to change my writing to appeal better, agree more, and comfort the biases and assumptions that allows my readers to sleep comfortably at night. I always wanted to be a writer, not an author. I always wanted to be a writer and instead of learning how to do it properly, I lived a life worth writing about.

Yes, I know, I’ve made plenty of mistakes, and yes I know I can still go to college and yes I know there are rare occasions when someones writing reaches the masses even if it were done haphazardly and yes I know there are many exceptions to the assumptions I’ve stated. I’m just telling you my story, my perspective.

So when I was accused of writing AI slush rather than complain about it I did what I always do. I thought maybe they’r right, maybe I have no place publishing my stories. But then I saw comments from people telling me what I wrote had changed their life. Someone said to me that they had never felt understood until they read my article. Another said something I wrote prevented him from relapsing. If it weren’t for AI none of that would have been published.

When I use AI, I prompt ChatGPT or Claude with my completed essay as well as I could write it. Most of the time it comes out as a chain of thought with run on sentences, incorrect grammar, and a limited vocabulary. I ask the LLM to polish the writing to be more coherent and presentable. Sometimes I ask it if things what I’m writing is going to come across wrong and it gives me suggestions on how to be more clear. Often times I tell it to revert what its written because it wants to make my writing more agreeable and more soft.

By the end, I’ve usually spent days or weeks on the first draft and another full day going back and forth with the LLM polishing it and giving me the self-confidence enough to allow others to see it instead of chucking it in the trash.

I came to the conclusion that this AI argument is not all that different than every other technology that has come out throughout history. Anytime something becomes easier you will have those who abuse it, you will have those who reject it, and you will have those who use it responsibly as the tool it was meant to be. Because any technology is not designed to replace us, but to augment our shortcomings and widen our perspectives.

I suppose I could argue that typing this on a computer was far easier than if I had written it down in a notebook with a pencil. And I suppose I could walk to work rather than pollute the environment by driving a car and buying a new iPhone every time the next model becomes available even though the only thing better about it is a slightly better camera I won’t ever notice and a slightly better battery that will just be consumed by the extra apps that I install. And I suppose I’ll just catch up on my algebra rather than use a calculator and mail this to you as a letter so you receive it next week along with all the junk mail that reminds me of all the clickbait written by real humans before AI was ever invented.

That’s why I started WritersLogic which for the past 10 months has been a registered corporation run by one person, with zero outside funding. And I used LLMs to help me write the software, even though I have been a programmer since I was a teenager. Over the past 10 months I’ve worked with experts in the Internet Engineering Task Force who have reviewed an entirely new cryptographic primitive I developed which uses the position of physical memory as its recorded to prevent a calculation from being sped up by a faster computer - this is the underlying technology that I then used to build WritersProof - an offline desktop application that records not the keys that you type, but your editing behavior - the millisecond timing it takes you to type a R after a P, how often you go back and revise a word, or how you slow down to write a complex passage and speed up as you write the prose. It takes this information along with a hash of the document (a long string of digits that look completely random) and it builds a tree of cryptographic evidence where each branch is dependent on the branch before it and any editing of that evidence after its’ published destroys the whole tree. WritersProof allows a writer to write as they always have and if they are ever accused of writing AI slush they have something stronger than drafts that can be easily forged and post dated - they have a cryptographic evidence package that could not be forged, post-dated, or faked. This evidence does not prove that the words came from your own head - no one can prove that - but they do prove that you sat down and took the time to type out each letter and each word.

For a short essay this carries only so much weight. I could easily have an LLM write an essay and then print it out and retype the whole thing into my editor. I can’t do it via splitscreen because WritersProof will notice, and I can’t type it all the way through with uniform pacing because WritersProof will notice that too - at least not as long as I want someone to believe I wrote it. WritersProof will record the evidence either way and never claim whether its generated or authentic from your own head. All it does is record the evidence.

If you are a writer like me snd you use Ai responsibly you can paste th content into an LLM and then paste the altered text back into your editor and WritersProof will ask you to attest to where that content came from and the differences in checkpoints will show how much different that content is versus was.

That’s the whole thing. And while developing it I was invited to help write the text-provenance standards published by Coalition for Content Provenance and Authenticity, and the Trust Over IP Foundation. All this from a guy who used to sleep behind a dumpster.

I also wrote the most popular MCP server for Scrivener that makes it easier for writers who want to use AI responsibility. And BluePencil, an application that gives more conventional users the ability to do the same thing that many AI users do, except BluePencil does it deterministically without any AI involved.

All of these tool are available for free. I also published a small desktop interface that makes the mcp server easier to use which I charge a small fee of $4.99 for. And there is a pro subscription level feature within WritersProof that allows you to save your evidence remotely in the cloud.

I never wanted to be an author. And I never wanted to be an entrepreneur. I just wanted to prove that what I had to say came from me, regardless of the tools I used to create it.

2 Likes
2

I’m very confused about this tool.

I just pasted the whole opening paragraph of Pride and Prejudice into the tool available from here https://www.w3.org/Mobile/training/mobile_course/resources/pandp/chapter1_no_vp.html. I then refreshed the page, pasted the same paragraph in, and the tool says "Signed. Your text is unchanged — an invisible watermark is woven in."

However, and this is where I’m confused:

i) It allowed me to paste in a whole chunk of text, written by another, but attributes that to me. How does this prove authorship?

ii) According to the tool, can I now prove ownership to that paragraph because I “wrote” it? If that’s the case, I now claim ownership to one of the best opening paragraphs ever.

I thought the idea was the tool would prove that I wrote something, and clearly it only records that I pasted something into the pretty text box.

I then went to https://writerslogic.com/sign and signed the following paragraph, totally created in GPT:

It is a truth universally acknowledged, that a small family of elves in possession of a comfortable hollow in an ancient wood, must be in want of a little adventure. The Scriveners, who had dwelt for generations beneath the roots of an exceptionally venerable oak, were not a family of great consequence among the woodland folk, though Mrs Scrivener was firmly of the opinion that her daughters deserved considerably better prospects than the forest had thus far afforded them. Mr Scrivener, who preferred the company of books to that of neighbours, and whose chief ambition was to be left undisturbed with his pipe and his ink, regarded these concerns as largely unnecessary; but when news arrived that a wealthy elf of excellent standing had taken up residence in the hollow of the great elder tree beyond the brook, even he was obliged to concede that the peace of their little household might soon be at an end.

Again the tool accepts this, it doesn’t prove authorship at all. I can still send this out claim it’s mine, when I didn’t actually write any of it.

See why I’m confused…?

5 Likes
3

I share @synapse56’s concerns.

Where a tool like this might be useful is when you need to prove you had a finished draft of something on a certain date (e.g., because someone else is claiming to have written a script you wrote)… but even then there seem to be a number of issues, such as:

  • there are easier ways more trusted by the industry (such as WGA registering; leaving a sealed copy with your lawyer; heck, even just posting yourself a copy and never opening it).
  • anything you digitally add to a file can be digitally removed or digitally faked. This has always been true, but it’s especially true in an AI world.
  • without understanding the processes involved in detail, it sounds a lot like even minor editing can accidentally but irreversibly strip / corrupt any watermarking you’ve added.
2 Likes
4

Um, I hate to break it to you, but BBEdit → Text → Zap Gremlins… and your ‘proof’ is gone. Passing any ‘claimed’ text through OCR would also remove the proof. Security through obscurity provides a false sense of security at best, and it usually only requires a tiny amount of reverse engineering to make it fail completely. Also, being able to claim others’ work as your own doesn’t seem like it’s gonna end well regardless.

To be honest, the whole thing feels like a vibe-coded product to solve a problem that as others have pointed out can already be solved in better (and more reliable) ways.

4 Likes
5

Mr Condrey, I’d like to share with you some advice I was given roughly four decades ago on how to improve one’s writing: read as many well-written novels—award-winning, well-reviewed, generally respected, whatever—as you can fit into your schedule: the more, the better. You don’t need to focus on a genre, or even stick to ‘classics’; it’s best not to limit yourself. Read everything you can get your hands on and that you have time for. After a time—it may take a couple of years—you’ll find that you’ve improved your vocabulary and absorbed an ability to not only write better sentences, but also you will have internalized a capacity for guiding the progression of words to sentences to paragraphs to chapters to a full account of whatever you intend to express in your own voice. You will have learned by example. I suppose one could call it a passive instruction in creative writing, but it is far more effective than any college course which is by necessity far more limited in scope.

2 Likes
6

Reading a lot of books is, of course, highly recommended. However, at some point – if you don’t have innate ability – you have to analyse. And if you’re a keen writer, or keen to be a writer, this should be a delight.

For example, if a writer has caused you to cry, or make your heart race, or turn all the lights on – even in daylight – then understanding how they did that will provide you will a vast well of knowledge to draw from.

So why not pick an author you love – or just one story – and pick it apart; sentence by sentence; word by word, sometimes. Try to understand what the write is doing and why they are doing it. Learn their rhythm. Learn their ticks – you can borrow from those later.

1 Like
7

Many people I know … actually, most of them … are much better at expressing themselves verbally than in writing. This can be put to good use. If you tell your stories to yourself (verbally) and record them, you’ll gather a lot of material and do yourself a favor, because spelling, syntax, and things like that don’t matter. You can always figure out later how or who will transcribe these audio recordings.

8

To this good advice, I would just add: write.

A lot.

And edit it yourself, don’t let a machine do it for you.

Then write some more.

Learn to hear your own voice, and listen to the stories it wants to tell.

3 Likes
9

@writerslogic — thanks for the MCP server, it’s great! And for the inspiring story, wonderful to see you overcoming adversity and making the most of this new paradigm. Lots of people are sensitive here about AI, and rightly so. As I understand it, your goals about quantitative measurement of text provenance should be something everyone here can get behind. As a scientist, I know how hard it is for non-native English scientists to express their work faithfully, the cards are stacked against them. LLMs can greatly help them, but the fear is that it then hallucinates details of the science, adjusting meaning in a way not intended but hard to separate out, not just improving the legibility. Being able to track the provenance would really help them to ensure the meaining is theirs…

10

In the original post, I explicitly said that recording a writing process cannot prove the words originated in someone’s head. I also explicitly rejected the idea that someone should enter a simple prompt and claim authorship of the output. I also described my workflow, which is to start with an essay I’ve written, followed by editing that I review, question, reject, and often undo. Several statements in my original post seem to have been completely ignored, and I wonder how many people who responded actually read it.

Your review demonstrates a real limitation of C2PA provenance. The /sign page signs the supplied text; it is not the tool that witnesses the process. What you reviewed was only the watermark/fingerprinting utility, which binds the text to a cryptographic key using C2PA provenance mechanisms (https://spec.c2pa.org/specifications/specifications/2.2/explainer/Explainer.html), where the focus is on checking the integrity and provenance of a claim, not establishing whether that claim is true.

WritersProof (https://writersproof.com/) is a completely different product linked to from WritersLogic. It collects evidence while you work and packages that evidence for verification. You can edit the document any way you want, and WritersProof will record evidence of how the document evolved. If you would like to try to audit that application, your feedback would be much appreciated.

Yes, BBEdit, OCR, or any number of things can remove invisible characters. But removing a watermark, altering a signed record, and fabricating a credible history are three different operations. The watermark is one way of carrying a record. An evidence package does not disappear when someone removes those characters. Plus, any edits to the document after the evidence package is exported would require a new export for the revised version.

“Invisible” describes how it appears to the reader and has no bearing on the signature’s security. And knowing how to remove a carrier does not give someone the ability to alter an existing signed record while preserving its valid signature.

I agree that registering and depositing a draft with a third party can be useful. The WGA registry describes its service as documenting a dated claim of authorship. The U.S. Copyright Office specifically states that mailing yourself a copy is not a substitute for registration. WritersProof augments your idea by adding process evidence that can be audited.

And on the “vibe-coded” criticism: I stated in the original post that I used AI to develop the software. Much of the work I have produced is open source and open to the scrutiny of anyone. I have also had this work reviewed by security experts. I recently read an article about Linus Torvalds even using AI now.

Several replies repeat general advice about learning to write while overlooking the workflow I described and the distinction between signing a document and recording how it was written.

I appreciate the advice; however, having assistance with polishing an essay does not mean I have surrendered responsibility for the final product. Provenance can help document where assistance entered the workflow, but it is not designed to certify whether an edit preserved the meaning or correctness of the statements.

Like I said in my original post: with every new technology, you will have those who abuse it, those who reject it, and those who use it responsibly as the tool it was meant to be—to augment our shortcomings and widen our perspectives.

For someone like me, who lived through trauma but struggled to put it into words, these tools can make the difference between finally sharing a story and throwing it away.