Good day, colleagues 😑

No one writes perfect code, especially the new o1 model 😳

Bugs have always existed,
dormant bugs are always lurking,
and they will be with us for some time in the future.

As well as debugging.
I wrote a bit about this -- feel free to read and enjoy, or hate, whatever 🌟

🔗LINK 🔗

https://fabric.so/ is complete dogshit 💩

The guys are drastically wanting your money and selling their product like a charming “one place for your notes and documents.”
Well, while they did a kind of good integration with some LLM for linking and summarizing your saved notes, documents, and links (linking “similar” stuff also works acceptably every other time),

I just encountered a horrible bug where the desktop Fabric app did not save my text note. 😢

I’ve been working on a new blog post for a couple of hours within its notes text editor, finished a big section, and boom, everything’s gone 🙂

Just becuase of that, no matter how good they integrated AI features - it is the dogshit 💩
Why? Because it is unnacaptable what such a critical and fundamental functionallity is not working propperly.

Well, at least they reacted pretty quickly to my bug report on Discord.

You know what? My note might not have been saved BECAUSE I DID NOT CLICK THE SAVE BUTTON INITIALLY, RIGHT AFTER CREATING THE NOTE!
Moreover, the developer mentioned that something similar happened to one of their team members a couple of weeks ago:

it does indeed sound pretty bad - the reason I asked about the Save button is it happened to somebody in the team a week or two ago, they did not click Save and so the content was wiped when they closed the modal; in which case it would be a UX problem more than a true bug (not sure what the case is for yours, and we are going to look into it)

They tried to restore something, but would it be surprise if I will say that nothing was restored?

lmfao

Textual modelling VS visual.

Have you ever come across infographics that are not just pictures but truly provide a comprehensive understanding of the system being modeled? 🤔

The kind of representation that’s easy to edit, maintain, and upgrade, and that a new engineer on the team can grasp without any huge additional explanations or clarifications from more experienced colleagues — and without formal text additions.

Personally, I’ve never encountered such diagrams. In fact, I believe it’s probably impossible to create them.

What I have seen are either poorly designed, ugly, and confusing diagrams or very beautiful and impressive visuals that evoke a good first gasp but quickly lead to a choke. 😶

These pictures may capture your admiration for 10 minutes, but as soon as you dive deeper, they raise a bunch of questions. Questions that can only be expressed in text to be properly addressed and further processed.

Text carries far more meaning per unit of space than a picture, and it’s much easier to correct and expand. A picture, on the other hand, can only be redrawn.

Take, for example, a flowchart representing a loop versus just the word “loop” itself. Both convey the same meaning, especially to someone who already knows what a loop is.

You might argue that a flowchart makes it easier to explain the concept of a loop to someone with no programming background. But that’s not true. 🙅‍♂️

Show a passerby a flowchart of a loop without annotations — they likely won’t understand a thing and will fall into fantasies and assumptions, depending on their background.

However, a concise text description of a loop, with a couple of examples and analogies, can communicate the concept to almost anyone.

From that moment on, the word “loop” will carry the same meaning for them as the clunky, angular flowchart with all its blocks, enabling its use to explain more complex matters. The word becomes a powerful container of essential meaning, staying flexible and applicable.

This is a simple example, but it extends seamlessly to diagrams and models of any size and nature — from the smallest to the largest. The bigger the system, the more relevant our problem becomes. Imagine a large system and the challenge of modeling its concepts and domain entities using both pictures and text. The latter approach feels promising and applicable. Whether it involves formal natural language specifications or specifications through common programming languages, it remains, ultimately, just text.

However, it is obviously not easy to reflect all domain intricacies in the code itself. That’s the point I was referring to here

🔗

Formal textual specifications are easy to write. I mean, even if it requires a lot o mental efforts of course, it is easier to reflect in text, than in rock painting. Because text carry more meaning in any sense, be the end of the day - even per square centimeter, and they are much easier to work with and moving forward.

We can even write multiple versions of the same system specifications for different stakeholders, using their “native languages.”

Good luck doing that with pictures. You will end up redrawing them constantly or struggling to find a way to communicate with others and share knowledge, instead of doing actual work.

Systems thinking and engineering are collaborative efforts. A picture alone does not effectively transfer domain-specific information from one person to another. Textual specifications excel at this.

Domain-specific languages are built from textual entities, not emojis.

It feels to me that there’s really nothing to argue about 😞

Hello,

Recently, I had a great time exploring Meta Threads.

The recommendation algorithms are fantastic, and the community is surprisingly positive.
It is much easier to grow audience in Threads than here.
So I’ll be posting frequently in my Threads feed, feel free to subscribe if you using it.

I’m not sure if it makes sense to replicate all the messages from Threads here,
but I’ll definitely continue to announce new blog posts in this channel and share what I find to be the most interesting news and topics.

Just like now — I’m sharing my new post with you!
🔗The global state is not a Monster 🔗

Uh oh no way! 🤡


https://open.substack.com/pub/garrisonlovely/p/is-deep-learning-actually-hitting?utm_campaign=post&utm_medium=web