Why should you read our blog?

Slop and Trust

If you happen to still prefer actually using a search engine and you in fact scroll down to the small section after the AI summary, there is a good chance that the first few results, sometimes even the first few pages of results are also AI-generated content.

The problem is not per se the AI, but the (potential) lack of review and nuance. AI slop is not called slop without reason; sloppiness is a lack of diligence. With basically everyone having easy access to language models, the internet is flooded with AI-generated content that has not received any due care. It has become trivial to generate generic content for your blog and the market encourages this as sheer quantity increases traffic (and market share) and thus potential ad revenue and exposure.

The resulting problem for me is the lack of trust. Whenever I see markers for AI writing like “not just this, but that” constructions, em-dashes and a lot of fluff with not a lot of actual content, I start to doubt the credibility of any factual content of the text. More often than not AI-generated articles have provided apparently simple solutions to complex problems that omit important context, are not production-ready or straight up don’t work.

Just ask the model directly?

Given those issues with modern web search, I can easily relate to people just directly consulting their favorite language model. You can have a conversation and receive explanations in almost any depth perfectly tailored to your current need. Great stuff!

Whats the problem? While hallucinations have noticeably gone done over the last years, language models (for now) still remain just that, language models. Even with improving architectures and harnesses trying to add reasoning capabilities to the language model, the language model cores still just compute the most probable next word taking a previous sequence of words.

Don’t get me wrong here, by using an apparently simple “guess the next word” logic at the core, models are already performing wondrous feats like solving long-standing mathematical problems and finding critical vulnerabilities in open source software that has been reviewed and iterated on by generations of developers. But still sometimes a language model just fails to understand a pun, count the number of “R”s in Strawberry, or recommends using a certain API in a framework, that just doesn’t exist. Most of the time I can prompt to reconsider, but that requires me to see the mistake in the initial result.

Why do we at esveo write blog articles?

The first articles were written to just internally document common pitfalls and learnings that we encountered in our daily work. The target audience for one of those early articles were colleagues and sometimes even just the author themselves. This evolved into playing our part within the developer community, being vocal about topics that move the community and sharing ideas, practical problems and solutions. Articles no longer just cover the developer perspective, but also business and management perspective which for one, opens up to a bigger audience and more importantly, fosters mutual understanding between those two interconnected groups.

Of course the blog also aims to increase our visibility and transparency to the market. Business clients as well as potential future employees can get valuable insights into how we work and decide whether that is something that fits to them and their needs.

How do we use AI (for our blog)?

This is a question with no single static answer. AI usage varies a lot from writer to writer, and also changes over time. While I myself have used AI to help structure blog articles in the past and to disentangle some monster sentences, I went back to enjoying the process of writing myself. AI also can be really great to get over the hurdle of a first paragraph. Generating something to start from can help to get into the flow of writing (which most of the time leads to discarding the initially generated paragraph anyway, so who cares). I don’t think AI in writing is generally bad. AI is just another tool, and used correctly it can be used to help in the creative process and smoothen out bulky writing. But sometimes bulky writing also gives the text character, makes it a human product.

One important thing that we all share when writing articles is that the idea for the topic and main points to deliver are something that comes from us and not from the AI. It also helps that we have built a healthy and open communication culture at esveo where we both value and practice constructive criticism. This way, bigger inaccuracies are spotted quickly

Oftentimes we do use AI for helping to translate articles, as all of our articles are available both in German and English. Most of us prefer to start with an English version, as most technical vocabulary is English anyway and later on translate back to German. I personally feel like proof-reading and adjusting an AI-generated German translation is easier to manage than the other way round. But even then, the first translation is most often just a starting point for proper manual phrasing.

Quality over quantity

We started the blog with weekly releases. Over the last years we noticed, that due to other projects limiting the available time, quality started to suffer due to the pressure to generate more content. This is why we went down to new posts every other week. Everyone in the company is encouraged to write about two articles per year, which is a low enough number so that everyone can preserve their creative freedom and no one is forced to write an article, when there is nothing to write about.

So where is the value in our articles?

I’ve discussed this a lot with colleagues and friends and we always came to a similar conclusion: There is still some intrinsic value in information being compiled by a human with practical experience. For technical articles, we can select the specific aspects that have proven relevant in our daily work, e.g. what part of date data structures we actually care about or how we went about optimizing our slow MySQL database. We also have articles where we share specific experiences, working in a new kind of project, realizing that we’d like a grid view on data in every application or documenting how we at esveo came to be who we are. All these articles contain parts of our personality which allows you to kind of get to know us. If I had to distill our blog down to one core value, it would be this: through the blog we share our practical, personal, and human experiences.

Getting back to the multitude of AI or AI-looking result in your web search, it is even more important to know sources that you can trust, and we strive to be on that list of trusted sources. So for now, we’ll keep publishing new posts on our blog to keep you updated on topics that matter to us as developers, consultants, and first and foremost humans. Let’s see where this journey leads. We’re happy to have you with us!