Blog

Beyond the Chat Window: Persistent State, Adaptive Schemas, and Auditable LLM Workflows

• The Chat Window Is Not the ApplicationSeptember 30, 2026
• Schemas and Protocols: Turning an LLM Workflow into a Stateful SystemSeptember 30, 2026
• Resumable LLM Work: Manifests, Checkpoints, and Outputs Larger Than a Single CompletionSeptember 30, 2026
• Adaptive Schemas: Letting the Data Model Evolve with the WorkSeptember 30, 2026
• From Semantic State to Formal Analysis: Graphs, Provenance, and Auditable LLM WorkflowsSeptember 30, 2026
• Building the Workflow: A Reconstruction of This SeriesSeptember 30, 2026
• Beneath the Chat Window: Taking the Reins of Stateful LLM WorkSeptember 30, 2026
• What Changes for the User? Performance, Control, and Project-Scale WorkSeptember 30, 2026

From Magical Mysteries to Mathematical Meaning: Backward-solving with persistent LLM state, from wizard-detective role play to assessment-grounded mathematics education

• Why a Magical Mystery Couldn’t Be Generated ForwardOctober 1, 2026
• Start at the Solution: Backsolving a Wizard’s Evidentiary TrailOctober 1, 2026
• A Wizarding World Made of Committed TruthsOctober 1, 2026
• Before the Spell Becomes Canon: Validating Generative FactsOctober 1, 2026
• The Magical Mystery Was an Epistemic MachineOctober 1, 2026
• From Magic to Mathematics: Let Assessment Define the Starting StateOctober 1, 2026
• How Far Back Must We Solve? The Mathematical Class Access BoundaryOctober 1, 2026