RSS Amplifier

OSINTBRIEF · Jul 15, 2026

Technical Stock Screener & A.I.

0
Sign in to vote or save

Eric Engle · OSINTBRIEF

Synopsis: AI generally fails at paragraph breaks. AI doesn’t know if it’s training data is fiction, non-fiction, advertisements, song lyrics, poetry, or dialogue. LLM AI is a next word predictor, not a next paragraph predictor.

And a technical analysis stock screener. Here’s a sample output for Today:

https://amazon.com/author/quizmaster I also write books. They are all on kindle unlimited.

You’ve probably noticed it, even if you couldn’t quite name it:

Choppy.

Abrupt line breaks.

Rhythm that reads like an ad.

That’s no accident.

Large language models are not trained to think in paragraphs. They are trained to predict the next word. Not the next idea, not the next section, and certainly not the natural cadence of human writing across paragraphs. Just the next token in a sequence.

But when humans write, we use paragraph breaks to signal meaning by structure. Proper paragraphs organize thought, control pacing, and guide attention. A well-placed break can create emphasis, contrast, or relief. A misplaced one breaks pace and ruins logic.

AI often fails to break lines and make paragraphs effectively. This is because LLM AI predictively approximates what it has already seen. And what it has seen? Advertising. Song lyrics. Poems. Headlines. Lots of one-liners. Some are very short. Must be good! Clanker gonna clank.

In those formats, one-line paragraphs dominate. They are designed to hook, to interrupt, to keep the eye moving.

Like this.

And this.

And this…

Sick of it yet?

Given the training data, when AI generates text, it often defaults to one liners. Most obviously, three short sentences, of about equal length. In other words: Slop. An endless stream of short, fragmentary phrases, often as bullet points and frankly incomplete thoughts masquerading as coherent prose.

Once you see it, you cannot unsee it. Triplicities, short “sentences”, and endless one line “paragraphs” are the hallmarks of AI slop. With plenty of bullet points, lists, and boldface. Too bad people don’t write that way.

So… if You are using AI to write anything meant to be read seriously you need to actively override AI’s tendency to write slop. Everyone can see your AI writing. They don’t bother to tell you. They just click out and unsubscribe.

Real writing is structured, logical, organized, and has a rhythm that reflects the message. It varies sentence length. It allows ideas to develop. It uses paragraphs as containers of thought, not just as visual spacing, concrete poetry style.

Respect your reader: never write slop.

Last time we built a stock screener for fundamental analysis. This time here are two screeners using technical analysis. Next week will be flask.

The first screener is very aggressive and will strike “gold”, but rarely.

The second screener provides a report and recommendation. All FOSS, no local GPU required.

Read the original on osintbrief.substack.com

Comments

Nothing yet. Say the first thing.

    Sign in to join the conversation.