From the workbench
Thoughts on AI, software engineering, and building products.
Sentinel-R3 is Glyph Software’s 27B reasoning and tool-use model for authorized penetration testing — rebuilt on a hybrid-attention base with a 262k context and a nineteen-tool agent surface.
sentinel is a red-teaming agent CLI that maps attack surface, audits for vulnerabilities, and writes the report — with every command it runs confined to a container it cannot escape.
Sentinel-R2.1 is Glyph Software’s 9B reasoning and tool-use model for authorized penetration testing, security research, and attack-path reporting.
Monza is an open-source platform for spinning up short-lived, isolated development containers that AI agents can use to write, execute, and test code safely.
Release of Reasoning Router — an NLP model designed for reasoning-depth classification
We're excited to introduce Glyph Software LLP — building scalable, AI-powered solutions for modern businesses.