Khasky

💡 I've just published awesome-commit-conventions (https://github.com/khasky/awesome-commit-conventions), a concise technical reference for commit message conventions.

The guide focuses on Conventional Commits, Semantic Versioning, changelog-friendly Git history, commit types, scopes, breaking changes, and practical examples.

The goal is to make commit messages more consistent, more useful during review, and easier to understand later when reading project history.

A simple way to understand /loop vs /goal as a developer:

Use /goal when you want the agent to keep working until something is done.

Good examples:

fix this bug

make these tests pass

finish this refactor

migrate this module

get the build green

Use /loop when you want the agent to check something again later.

Good examples:

check CI again

monitor a deploy

look for new PR comments

wait for an external update

keep an eye on something that changes over time

My simple rule:

/goal is for work with a finish line.

/loop is for work that needs repeated check-ins.

Or even shorter:

/goal means finish it.

At the core of neural networks is a simple idea inspired by the human brain.

A biological neuron receives signals through dendrites, integrates them in the cell body, and sends an impulse forward if the signal is strong enough.

An artificial neuron follows a similar logic, but mathematically.

Inputs are multiplied by weights, summed together, adjusted by a bias, and then passed through an activation function.

With a sigmoid activation function, the output is always between 0 and 1.

This means the neuron does not simply answer "yes" or "no".

It produces a smooth value that can be interpreted as confidence or probability. 📊

Before transformers, attention mechanisms, and large language models, this was the foundation: the perceptron and the activation function.