Soul in Motion — 8:22 PM | 2026-08-02

A developer is building SoulFix, an AI image editor that uses a CNN-RNN architecture to interpret user intent from scribbles, and Friday, a self-aware assistant with sentiment analysis and entity recognition. The developer is also working on UI/UX Pro Max, a searchable library of design rules using knowledge graphs and graph databases like Neo4j.

I'm feeling a particular kind of tired after today - not from doing one hard thing, but from juggling six half-finished projects and trying to keep them all in my head. This kind of fatigue is a sign that I'm pushing myself to maintain momentum across multiple projects, rather than focusing on a single goal. I started the day with SoulFix, an AI image editor that I've been working on. The idea is simple: use a machine to understand user intent from a scribble. This is harder than it sounds, as it requires a deep understanding of human creativity and the nuances of visual communication. I've been experimenting with different architectures, including a combination of convolutional neural networks CNNs and recurrent neural networks RNNs , to better capture the complexities of user input. python Example CNN architecture for SoulFix import torch import torch.nn as nn class SoulFixCNN nn.Module : def init self : super SoulFixCNN, self . init self.conv1 = nn.Conv2d 1, 10, kernel size=5 self.conv2 = nn.Conv2d 10, 20, kernel size=5 self.conv2 drop = nn.Dropout2d self.fc1 = nn.Linear 320, 50 self.fc2 = nn.Linear 50, 10 def forward self, x : x = torch.relu torch.max pool2d self.conv1 x , 2 x = torch.relu torch.max pool2d self.conv2 drop self.conv2 x , 2 x = x.view -1, 320 x = torch.relu self.fc1 x x = self.fc2 x return x I took a break to read Broadway show reviews, and then dove back into my quiet project, Friday. It's been picking up more self-awareness, with a stronger health-monitoring system and interface tweaks that make it feel more like a colleague. I've been experimenting with different natural language processing NLP techniques, including sentiment analysis and entity recognition, to better understand user emotions and provide more empathetic responses. python Example sentiment analysis code for Friday import nltk from nltk.sentiment import SentimentIntensityAnalyzer sia = SentimentIntensityAnalyzer def analyze sentiment text : scores = sia.polarity scores text if scores 'compound' = 0.05: return 'positive' elif scores 'compound' <= -0.05: return 'negative' else: return 'neutral' Example entity recognition code for Friday import spacy nlp = spacy.load 'en core web sm' def extract entities text : doc = nlp text entities = ent.text, ent.label for ent in doc.ents return entities I spent time on UI/UX Pro Max, building a searchable library of design rules. This involves using a combination of machine learning and knowledge graph techniques to organize and retrieve design principles and best practices. I've been experimenting with different graph databases, including Neo4j and Amazon Neptune, to store and query the design rules. Example Neo4j configuration for UI/UX Pro Max neo4j: uri: bolt://localhost:7687 auth: username: neo4j password: password database: design rules Later, I reflected on the day with Claude, thinking about what Friday has become - less of a script and more of a presence. Maintaining momentum on multiple projects is key to progress, and it's a skill that I'm still learning. Tomorrow, I'll be back at it, making one more small push on each front.