Archive - Veylan Solmira

Sentience-Relevant Homologies Across Substrates

Note: This research proposal was originally submitted for the Future Impact Group’s Foundational AI Sentience Research fellowship program.

Oct 14, 2025 • VeylanSolmira

Why LLMs Understanding Corrupted Text Isn't Surprising

Oct 11, 2025 • VeylanSolmira

Testing Bezos: Universal Condemnation, Zero Consistency

Somebody was non-critically amplifying Jeff Bezos’ latest statements on AI bubbles and how, excitedly, even when ‘industrial bubbles’ pop they leave…

Oct 5, 2025 • VeylanSolmira

Secure Infrastructure for Multi-Agent Adversarial Testing

Current security evaluations test individual AI systems in isolation, yet deployed systems will soon operate autonomously and collaboratively.

Oct 3, 2025 • VeylanSolmira

A Comprehensive Consciousness Indicator Evaluation Framework for AI Systems

The Moral Urgency

Oct 1, 2025 • VeylanSolmira

Testing the Limits of Interpretability: A Research Agenda for Stochastic Parameter Decomposition

Two major aspects of mechanistic interpretability research are employing existing methods to learn about AI models and advancing those methods…

Oct 1, 2025 • VeylanSolmira

September 2025

Sep 29, 2025 • VeylanSolmira

12 Hours in Night City

[This is a part of my artistic expression as a philosopher and writer.]

Sep 28, 2025 • VeylanSolmira

How Does One Calculate Probabilities in Real World Examples When Applying Expected Value Calculations?

Introductory note: I wrote this in 2023 as part of the In-Depth program with Effective Altruism.

Sep 24, 2025 • VeylanSolmira

AI-Enabled Biosecurity Risks

The following presents my approach to thinking about AI-enabled biosecurity risks.

Sep 22, 2025 • VeylanSolmira

Environment-Dependent Model Organisms with Integrated Welfare Assessment: A Biological Framework for AI Safety

Sep 21, 2025 • VeylanSolmira

August 2025

Exploring the Distillation Paradox: When Models Transmit Traits but Not Capabilities

Anthropic found that distilling from a teacher model to a similar student model can copy unrelated tendencies, but other recent work found that…

Aug 19, 2025 • VeylanSolmira

#nojs-banner { position: fixed; bottom: 0; left: 0; padding: 16px 16px 16px 32px; width: 100%; box-sizing: border-box; background: red; color: white; font-family: -apple-system, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol"; font-size: 13px; line-height: 13px; } #nojs-banner a { color: inherit; text-decoration: underline; } This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts