Pagecord's Updated AI Disclosure

Photo by Miguel Ángel Padriñán Alba on Unsplash

Introduction

I wrote about hosted blog options¹ a while back. Included in the options that I thought were interesting was Pagecord. It looked like one of the nicer options: easier to use, well-thought-out, and some other nice features. At the time I was a little uncomfortable recommending it, but not having a specific reason I kept it in the options that I talked about.

Now, it appears that the issues I had started to examine are worth discussing in light of Pagecord’s updated AI disclosure. (Which took a funny twist as I was pulling all my notes together from back in June / July when I first looked at Pagecord.)

Who Is Olly?

Note: the following is not a criticism of Olly. It’s just a discussion of his background that anyone can find if they dig around a bit. He doesn’t try to hide any of this information, but he doesn’t really talk about it either.

Olly Headey was part of a group that founded a software platform called FreeAgent² back in 2007. FreeAgent was basically an accounting system designed for individuals and small businesses to handle all their needs.

FreeAgent grew quite rapidly.³ In 2012, they acquired 60mo to expand into the US market, and by 2013 were listed by Deloitte as one of the fastest growing tech firms. In 2015, they raised over £1M, and in 2016 became a publicly traded company.

In 2018 FreeAgent was acquired by the NatWest Group (formerly the Royal Bank of Scotland Group) for £53M. After leaving FreeAgent Olly appears to have tried working on several other projects, including FeedGrab.⁴ And now, of course, there is Pagecord⁵

Pagecord

Pagecord is a project I was initially very interested in when I was searching for a new minimalist blogging platform. However, between noticing a few interesting things about the project, and exchanging a few emails with Olly I found there were a few flags that bothered me, and now I’m seeing more to be cautious about.

Comparison

One of the things that I mentioned to Olly during our discussions was that I thought his comparisons of Pagecord to WordPress, Substack and Medium were a bit odd. I don’t see any of these platforms as falling into the minimalist blogging platform that Pagecord is targeted at.⁶ He also targets Pagecord as an Indie Blogging Platform.⁷

However, there is a lack of comparison to other true minimalist blogging platforms, such as: Bear Blog⁸, Pika Page⁹, micro.blog¹⁰, and WriteFreely¹¹ / Write.as¹². I would think it would be more appropriate to compare Pagecord to these platforms. I personally think potential users of Pagecord would be more interested in comparing these platforms than Substack or Medium.

It feels like the current comparisons are more for SEO targeting and promotion than it is genuine interest in minimalist or indie blogging, or providing information to potential users. Of course all of these pages are designed in a way that makes Pagecord look like a better choice than other platforms.

Promotion

If you go to several of the major search engines (Startpage, Ecosia, Duck Duck Go, Google) and search for PageCord you will find that most of the results feel more promotion targeted. The top results tend to be from (a) Pagecord’s own site, help documents, or (b) sites like ProductHunt, Trend Hunter, Ventrue Gaps, Open Alternative, and so on. Occasionally, you will find a few Reddit posts¹³ which itself isn’t an issue. However, they do make clear that Olly has an interest in finding a specific audience for marketing Pagecord to:

Thanks, yeah there’s a certain Tumblr vibe to my blog! The ideal customer profile is… emergent! 😅 I’m not 100% sure right now who to target, it’s all a bit too generic (likely a problem from a targeting point of view).

There’s quite a broad range of profiles for the current paying and free customers, but they’re all B2C right now. I would say most are also non-technical, or at least are not using their blogs for technical posts. Some are founders, some are hobbyists.

I actually think it makes a good product blog for startups (e.g. https://blog.feedgrab.net), but that’s super-niche as most companies probably want a bunch of customisation which it isn’t designed for.

I guess I’m targeting people who want to blog/write but don’t want to have to think too much about design. Just write some words, put them on the internet. Also people who don’t want to pay for blogging because they’re just occasional posters, or people who want custom domains and email newsletters but can only afford the lowest possible price - $20/yr in this case where competitors are $60+/year

The above was his answer to who he was marking Pagecord for about a year ago. Notice that he singles out his B2C users, and points to it as a companion blog for other products (using his own FeedGrab as an example product). Also of note, just a year ago he was pricing his blog at $20, and now it’s up to $40¹⁴.

None of this is necessarily bad, however it does seem to go against the ethos of the IndieBlog, and minimalist blogging ethos.

Open Source?

Pagecord is open source, in the literal meaning of the phrase. There is a GitHub repository where the source code is located. However, the ability for users to use the contents of the repository to self-host is another matter altogether. I initially contacted Olly after trying to get the Docker image of Pagecord running on my local system so I could look at a few features that weren’t accessible until a user subscribed. When I contact Olly about the issue, it turned out the Docker image hadn’t been updated for a while so there was a mismatch in the configuration.

And this was when I got an insight into Olly’s thinking on this “open source” project. The Docker image was really being supplied for local development and testing with Pagecord. It didn’t seem the intent was for users to set up their own hosted instances of Pagecord. This was something that I had been considering: setting up my own instance.

But, beyond this, it seems that there are dependencies baked into the project that would likely make it difficult for others to self-host. For example, in the documentation in the GitHub repo, it lists:

• Ubicloud for hosting the Postgres database
• Cloudflare R2 for image processing
• Hatchbox for deployment
• Caddy for domain and SSL management
• Cloudflare for DNS and Edge caching

None of this is an issue, but it does (potentially) add a lot of overhead for those wanting to self-host Pagecord. I think this loudly points to the fact that Olly didn’t design Pagecord so much as an open source project as he wanted to develop a commercial product using open source tools.

Compare this to micro.blog, which is a wrapper and set of extensions around the open source Hugo static blogging engine. This means that if you were using micro.blog, you could always set up Hugo, export your content from micro.blog and pretty much drop it into Hugo.

Another comparison is the relationship between WriteFreely and Write.as, where WriteFreely is open source than anyone can download and host themselves (of which there are over 550 instances running in the Fediverse as I write this article). Write.as on the other hand is a hosted service built on WriteFreely. Write.as only accounts for 2.75% of the user base for WriteFreely. And, of course, WordPress has been available for download and self-hosting for years, as well as there being hosted services, including WordPress.com built around it.

In addition to this, it appears that Pagecord is a single developer project. There are only three contributors listed on the GitHub repo: Olly, a bot to handle dependencies, and one other person who made two contributions over a year ago.

Again, none of this is bad, but it suggests that Pagecord is more of an open source project only in a literal sense, and not really in the sense of the spirit of the open source community.

AI Use

Up until this morning, I wasn’t going to write about Pagecord, despite having done most of this research over a month ago. But this morning got an email update from Pagecord that changed my perspective on this (and, there is an interesting twist to this section). Let’s be clear, Olly has been reasonably upfront about his use of AI in the coding of Pagecord. However, this morning an email was sent out noting that Pagecord was now using OpenAI as a content moderation tool¹⁵:

2. Content moderation

Pagecord is a user-generated content platform so it attracts spam blogs, and is open to being abused. The T&Cs cover this, and Pagecord has duties under the UK Online Safety Act. There is one automated check, on free and trial accounts only. When a post is published, the title, text and up to five images are sent to OpenAI’s content moderation service, which flags whether the content contains banned or illegal categories (violence, porn, hate speech etc).

This check looks at published, public content on free or trial blogs only. Private drafts, deleted posts, email addresses, subscriber lists and account details are not sent. I review flagged content manually.

If this is to be believed, anyone that is using a free / trial account is having their content submitted to OpenAI for moderation purposes. Olly states that the content is not being used to train OpenAI’s LLMs. While that is his belief, it requires that the users of the free tier of his service believe that OpenAI is honoring their agreement to not collect this information for use in training their LLMs. And, the problem with this is: there is no agreement between the users of Pagecord and OpenAI under which they can check to make certain that OpenAI isn’t using their content.

The (I think funny) twist to this is Olly wrote on his blog¹⁶ today:

I’ve ditched the OpenAI spam blog check in Pagecord. It wasn’t working very well really, so I’ve decided to manually review all new blogs from now on until the numbers get so large as to be unwieldy (i.e. never). The /ai page has been simplified.

This makes me wonder something… How long was the OpenAI spam blog check implemented? In order to make a decision that it wasn’t working well, it would seem that you would need to have a fairly large dataset to work with. That would suggest that it was implemented for at least a few weeks, if not a month or longer.

But, there is still something that users of Pagecord should be aware of if they are going to use this platform:

AI search indexers are not blocked because they allow your posts to be found and cited. If you want to opt out of this, you can write a custom robots.txt that blocks everything, or disable indexing entirely.

There are over 400 hundred AI web scrapers out there¹⁷, and I doubt that those who are concerned with their content being scraped is thinking that all of these bots are just going to “cite” the content. And, in order to be able to cite the content, doesn’t LLM have to be trained on the content?

This was part of what I talked with Olly about during our email exchange a while back. At that time the user didn’t have the ability to edit the robots.txt file on their site. To Olly’s credit he added this feature. I also talked to him about allowing the user to inject their own code into the HEAD section of their pages. (The reason for this was I read about using canary GUIDs to tell LLMs to not scan a dataset as it could poison the LLM¹⁸. Whether this still works or not, I don’t know. But it still seems like a reasonable thing to implement.) Olly disagreed with the idea of allowing the user to modify the HEAD portion of their pages at the time, but said he might consider it in the future.

Social Integration

This is a really small item, but it’s something that I’ve come to expect with many of the platforms I’ve looked at. Pagecord has an option called Fediverse Author Attribution, but I didn’t see anywhere that posts were being published on the Fediverse. Turns out, this feature does not use ActivityPub to publish posts. The Fediverse Author Attribution feature just adds a fediverse:creator meta tag to the header of posts.

Instead, Olly recommended using Echofeed¹⁹ to post to the Fediverse, Bluesky, etc. While this is a workable option, it seems suboptimal to me. I dislike having to rely on third party services for things that should be integrated into the platform. And, Echofeed shutting down is an exact case in point as to why.²⁰

Conclusion

The point to this article is not to call out Olly, or to warn people away from using Pagecord. This is about awareness. I think making an informed decision on the choice of platform that we use is incredibly important these days. Many of us are using Fediverse platforms like Mastodon, GotoSocial, Lemmy, etc because we wanted to get away from platforms that have control over content, and because we disagreed with the direction of the predominant social platforms.

Pagecord has been building this ethos of being an open source, open platform. However, looking at Olly’s background, his statements about Pagecord, the way the project is being handled, and so on, just leaves me feeling like it’s not being accurately presented to our community.

If you are happy with Pagecord now that you’ve read this background information, then I’m happy for you.