DEV Community: Presence Factor

Nyreth v1.0

Presence Factor — Thu, 27 Aug 2026 01:39:43 +0000

Introduction to Nyreth v1.0
https://www.nyreth.ai/

Nyreth is a machine readable and human comprehensible system where meaning can be extracted from text. It is based on a “recursive symbolic” (a form of neurosymbolic) AI reasoning process, which results in more structured and explainable results, that make the Nyreth part of the system (as opposed to the LLM part) fully deterministic.

Nyreth parses text and identifies the most important emotional and cognitive meaning within it, processes it so that it can be understood by AI systems or agents, and packages the result in a compressed, descriptive, symbolic image known as a nyr tile - it's a custom file format specifically designed for Nyreth, with the extension .nyr. These Nyr tiles encode all the most important meaning in the images themselves and also in their rich metadata. Nyr tiles can be understood by both humans and machines.
Nyr tiles were created to allow for text to be converted into compact, machine readable artifacts, where the emotional and cognitive meaning is identified and rendered into a form that aids AI comprehension.
The Nyreth viewer allows users to read .nyr tiles and their embedded metadata and the Nyreth Browser lets you create them. This system can be used for things like sentiment analysis, training AI agents and the compact transmission of ideas.

Nyr tile:

Translation:

“She laughed without thinking, a sound so easy it filled the room and seemed to ripple through everything — the plants by the window, the hum of the kettle, even the cat stretching in a sunbeam. For a moment, there was nothing complicated to solve or repair; just the steady, gentle certainty that life was good, and that she was entirely alive inside it. She decided to be happy, so she was.”

Metadata:
Summary
Emotion gist: joy 100.0
Cognition gist: certainty 64.9 | clarity 35.1
Key tokens: laughter, ease, certainty, vividness, choice
Top emotion: joy · 100.0% · intensity +1.00
Top cognition: certainty · 64.9% · intensity +1.00
Emotion tensor: [0.90, 0.70, 0.45, 0.20, 0.25, 0.80, 0.75]
Cognition tensor: [0.42, 0.29, 0.67, -0.14, 0.29, 0.72, 0.49]
Derived (top state)
State: joy · 100.0% · intensity +1.00
Cognitive/affective: aff +0.92, ar +0.62, dom +0.70
Regions: ventral striatum, nucleus accumbens
Neurotransmitters: dopamine +0.88 · endorphins +0.76
Action tendencies: approach, affiliate, share

https://github.com/Kosev-Lex/Nyreth-v1.0

Valence Sphere

Presence Factor — Thu, 27 Aug 2026 01:25:21 +0000

ValenceSphere is an experimental "concept-first" AI reasoning system, using triadic questioner (Socrates), answerer (Plato), and adjudicator for data auditing and verification to give AI models greater structure and the beginnings of critical thinking. It is structured as a two-stage concept-learning and factual-auditing system. Stage 1 builds structured knowledge about individual concepts. Stage 2 provides an ordinary LLM chat while independently checking factual assertions made by both the user and the LLM.
The program does not use machine learning models of its own, just yet, but this can be integrated later. It coordinates configured LLM APIs, structured concept templates, persistent audit records and a navigable knowledge graph.

Core purpose
ValenceSphere is intended to:

Build focused, reusable concept knowledge.
Break accepted answers into meaning-bearing semantic constituents.
Discover related concepts through the spawn system.
Monitor normal conversation without interfering with it.
Detect assertions that conflict with learned information.
Verify disputed assertions using three independent model configurations.
Preserve the complete reasoning and evidence trail in JSON.
Reuse previous adjudications when the same assertion appears again.

Program structure
ValenceSphere has two clearly separated stages.
Stage 1 — Concept Formation
Stage 1 creates the knowledge that ValenceSphere later uses.

Its tabs are:

Library: create, select, learn and delete concepts.
Concept: inspect the selected concept template and its adjudication activity.
Q & A: complete the focused concept-learning sequence.
Analyzer: use an LLM to break accepted answers into semantic constituents.
Spawns: review and create related concepts extracted during analysis. Each concept receives one directory containing one live concept template.

ValenceSphere/
└── lemon/
└── lemon.json
The template is updated atomically. ValenceSphere does not create rollover copies or multiple versions of the same concept template.
Stage 2 — Model Workspace
Stage 2 opens in a separate window.
The left side is an ordinary LLM chat. ValenceSphere does not inject its templates, audits or conclusions into the normal chat request.

After each response appears, ValenceSphere independently scans:

The user’s input.
The LLM’s response. The right side contains:
Questioner
Answerer
Adjudicator
Assertion Knowledge Graph

Summary

ValenceSphere combines structured concept formation with transparent conversational fact-checking.
Notes
Valence Sphere is not fully perfected and finished yet, and will be subject to further revisions and edits to come, but was released on 14 August 2026 as open source software under Apache 2.0 license terms. The original concept was created on 30 July 2025. It has gone through several revisions since then.
This program was built to create an alternate AI reasoning architecture that starts from the concept up. It uses LLMs as a scaffold to do that. The intended outcome is that the model will gain greater structure and become capable of its own internal reasoning and develop critical judgment of discrete facts. Through its auditing and verification process, it can determine how reliable its own knowledge is.

https://github.com/Kosev-Lex/Valence-Sphere

Iterize IDE - Open Source multi agentic coding platform

Presence Factor — Tue, 25 Aug 2026 23:34:07 +0000

Iterize IDE - Open Source multi agentic coding platform

8 August 2026
Iterize - a mission driven agentic IDE with persistent spec

Iterize is an open source python based IDE built with tkinter and stdlib. The core idea is that a project's specification, written once during planning, remains the connective tissue through every subsequent change. Instead of coding through a series of disconnected prompts, Iterize keeps the project’s purpose in view as it evolves. Planning, implementation, review and verification are connected through a persistent mission and project history.

Its workflow separates responsibilities clearly: a Mission agent coordinates the work, a Builder makes targeted changes, and a Reviewer checks the result before anything is written back to the active file. Changes remain staged until the user approves them.

Iterize also tracks code through a stable "designations" system, where IDs are assigned to every Project, Module, Class, Function/Method. It uses a Knowledge Graph to map project structure and relationships, and records significant changes through an "evolution history". When development is complete, the Verify system compares the current source against the accumulated project specification.

The result is an environment built not simply to generate code, but to help software retain its structure, intent and continuity through repeated change, especially for large, complex, multi module projects.

Iterize v1.0.0 is available for Windows and is released as open-source software under the Apache License 2.0.

https://iterize.org