That new banner image is AI isn't it?

They have a marketing department taking care of this. As much as I think Jonathan and Ivy are great, I don’t want them anywhere near the code. I absolutely want them making sure that the pictures they put out to promote their product make sense. I want to give my money to a company that values the details, that values getting stuff right.

1 Like

It is a thumb nail image…IMO, you are putting WAY too much emphasis on its importance.


How is this image pertinent to training software?

It seems like people just like to complain about every fraking little thing today. It’s easy on a keyboard shielded by the internet. Nobody would have taken the time to write a letter about such a thing 30 years ago because it’s no big deal.


Maybe. But I’d argue that it’s the image the company is using to try to convince-trick people into learning about their product, with a goal to have those people pay them money for said product. If that’s your job, do it right.

So, you think TR is trying to “trick” people into subscribing with a generic thumbnail image……but you also think Jonathan and Ivy are “great”.

OK….i’m gonna bow out now. This has reached the silly point.

1 Like

And…I haven’t written a letter to the company. I’m ranting about it in an online forum, as god intended.

Anyway, I think putting images like this into the world shows that you don’t care about the details as much as I think you should if that’s your job. I think it reflects poorly on a company I like.

And so I shout into the void. This particular void shouts back sometimes.

They seem like good people. Some of the YT thumbnails have gotten really clickbaity in the past year or so. I understand why, but that doesn’t mean I like it.

Anyway, I think I’ve explained why I care about any of this. Apparently some people don’t. And that’s fine.

1 Like

My day job is cybersecurity architecture for a huge multi-national company. About 80% of my work this year has been directly related to AI/LLM/ML tech in the enterprise. Two focuses have been keeping our data out of public models and only using services that will indemnify us against copyright infringement related to the data thier model was trained on. (I meet with legal on a daily basis :roll_eyes:)

As with everything else cloud related, if you’re not paying for it then you are the product. If someone at TR is using a free service, then that is a risk. If it’s paid, then the service should be the one on the hook for a clean model.

I just trust that Nate and the leadership have established a policy around externally sourced AI/ML and everyone at TR is adhering to it.


There is a flip-side to your theory, and that is that if they are putting too many resources toward the details on stuff that doesn’t impact the actual product (like a thumbnail), then that may be taking away from the resources needed for the important details. And as always, which way it goes depends on who’s running the project and their quality control. Because of the inherent variability, I’ve learned not to judge the product based on the window dressing, in general. Once you have experience with a company, you know, and then the window dressing doesn’t matter that much anyway.

As an example, I design a custom building product, based on the house plans I get from customers. My experience is that 90+% of the time, if there is an artsy hand-drawn sketch/rendnering of one of the perspective views of the house, then the attention to detail in the critical plan sheets is going to be lacking. Due to deadlines, there is almost always a rush to put the finishing touches on the plans, and the architect is apparently spending their time on that rendering, rather than checking the dimensions, elevations, notes on the rest of the plans.


Not a fan. TR seems to be cheapening its brand. Whether that’s AI images, or click baity approaches, I don’t like it.

I hear people saying they don’t need to spend money on photographers/editors/marketing etc, but that misses a couple of points in my view.

  1. They have been trading for over a decade, they already have images. If it was just about saving money to invest in the core product then they could just use old imagery.

  2. In my opinion they have been slow at evolving the core product over the last couple of years even with price increases (outdoor ride levels??), so I’m not seeing a tangible benefit from a race to the bottom brand wise (just my opinion, others will disagree)

AI has the potential to make our lives better - if it is used well. Reducing your marketing spend (jobs) because AI can do a substandard job isn’t one of those ways.

1 Like

That is because click-baity images and titles work.

Who lost their job because TR started using AI images? I must have missed that announcement.

And if they kept using those same images, I can pretty much gurrantee that many people of the same people would be complaining that they don’t invest in new images, etc.

1 Like

I didn’t say a specific individual lost their job. I don’t know how many people TR employ, their employee turnover, or their spend with third parties. I’m sure you were just making a snarky comment, but not sure why you’d expect TR to publicly announce if they were changing their marketing spend…

Ultimately, if production of imagery is done by AI then there is no need to spend any money on a photographer (whether or not they are an employee of TR).

If AI is being used to create creepy/offputting images like those posted above then I’m going to assume TR is doing it for cost purposes rather than trying to produce a better end product.

Click baity thumbnails really put me off. They make a brand look like they are less trustworthy because they aren’t relying on the quality of their content/product to create engagement, they are just optimising for volume of click throughs. You may like them, that’s your prerogative.


You aren’t the target audience….you are already engaged.

So there is zero chance then that they are using their resources to, in fact, improve their product by having their employees focus on the product instead of images?

As I said, they work and that is why people use them. You can rail about it all you want, but you have to get the clicks to get the engagement. Go listen to the Bonk Bros podcast where they talk about the massive difference they get in engagement when they use titles like “scariest gravel race ever” vs. “Big Sugar Recap”.

Doewsn’t matter how good your content is if people don’t click on it.

Well since I never said I like them, this seems like another assumption on your part. I am simply pointing out the realities of competing in a digital world. If you aren’t using proven, successful str2tegies for engagement, you are wasting your time and money.


By that same thinking though, any given human is inconsequential. Literally, that’s what you’re saying - but put slightly differently. Overly dramatic, sure, but again, that’s equally how silly that thinking is.

The problem is, for some topics, the training data is slim pickings. Sports tech is one of those I understand well, and getting accurate ‘training data’ on some products means only a handful of sources are out there. So, I spend weeks preparing a review, then a training set come along, reads it in a fraction of a second, dumps it into a search engine, and nobody needs to come to the site (or YouTube video) again to learn about the product. They ask some empty text box, which then re-gurtitates my review and is unlikely to credit it, and even if they did, nobody would ever click on that tiny link placed in oblivion anyway.

While that might sound silly, in my little corner of the world, that’s exactly what’s happening. Heck, you can even see the style of my writing come out in certain questions you ask ChatGPT about in this space. It’s hilarious.

The reality is, eventually people won’t visit sites anymore for answers. It’ll just be spit out of a prompt. And once that happens, those underlying sources disappear. After all, a processor chip in a datacenter isn’t going out and testing the latest watch to see if it sucks or not. It’s pretending to not steal that content instead. Today there’s a visible lag on that training data set, but in short order that lag will be measured in milliseconds. Within minutes of my review posting, other completely-AI generated reviews will appear as well, fake photos and all. Some probably with unique ‘tones’ and all.

This isn’t meant to be doomsdayer, but a reality-check. Anyone who thinks otherwise is frankly kidding themselves and blind to how fast tech is moving. Am I doing anything about it? Nah, not really. For a site my size, even with millions of views/month, I’m not even a tadpole in the sea of this fight. This fight will eventually be duked out between massive book and news publishers, vs tech companies, ending up on the steps of big courts around the world. Most of which will then be ignored in various countries that just don’t care. At some point the underpinnings of human-fed information will stop, or be severely restricted behind endless paywalls. Nothing is truly for free.

Bringing this back to TrainerRoad, they’ve noted historically they’ve been sensitive to other training platforms over the years stealing their workout libraries and re-publishing them. While I’m not directly equating that to using AI-image models, I do think for a company like TrainerRoad where its entire existence is summarized into a series of easily replicated text files with target wattage values and timestamps, they’d be even more aware of copied content than most companies.

And to be clear, I frankly don’t care if TR uses AI-generated imagery for YouTube thumbnails or not. But I do think they should at least have formulated an internal opinion on where they draw those lines, and how they might feel if/when their upstart AI-driven competitors simply pull in their TR workout database as a training set, and call it done.


Thank you for writing that.


Good info, but I was specifically talking about diffusion model AI and copyrighted images.

An LLM paraphrasing/regurgitating an internet article, e.g. yours, is another matter entirely.

1 Like


With friends in this space, I know for a fact that many of the players are extremely protective of their training data sources, per their legal depts. They don’t want to justify it.

1 Like

Well timed:

Non paywall: New York Times sues Microsoft, ChatGPT maker OpenAI over copyright infringement

They want the YouTube channel to grow. Like it or not, “click bait” thumbnails are necessary on YT to get views.


I know that these sorts of strategies are needed to maximize views. It does keep weirdos like me from clicking on as many videos, on principle. But I know that these things result in more net views. It just makes me sad. That’s a related, it different issue to the fake photos.