Blog
MCP and A2A Are Protocols, Not Governance
2026-09-01 — 7,973 live MCP servers probed, 40.55% with no authentication, all 119 OAuth-enabled servers flawed. What the 2026 security studies mean for agent governance.
Zero Trust for AI Agents: Policy as Code and Provenance
2026-09-01 — How zero-trust identity, versioned policy, and SLSA-style signed provenance give you audit-ready proof that agent code and data stay under your control.
Context Economics: Cache Reads, Context Rot, Retrieval-First
2026-09-01 — Your AI bill is mostly re-read history: 92.17% cache reads over 27 recorded days, list-priced at $12,027.14. The measured case for retrieval-first.
The Distributed-Systems Toolkit AI Agents Inherit
2026-09-01 — Leases, fencing tokens, idempotency, sagas, and controllers: the classic distributed-systems toolkit that keeps parallel AI agents from overwriting each other.
The Narrow-Waist Lesson: AI's Future Was Invented in 1974
2026-09-01 — Cerf & Kahn's 1974 narrow waist and Saltzer & Schroeder's fail-safe defaults explain why agentic AI needs one evidence envelope and one coordination plane.
The Verification Tax: What AI Code Studies Measured
2026-09-01 — DORA, Stack Overflow 2025, Veracode, GitClear, EASE/MSR 2026, ImpossibleBench, and SWE-Bench+ — what the evidence actually says about trusting AI-generated code.
Evidence-Gated Assurance Framework (EGAF) overview
2026-08-30 — The Evidence-Gated Assurance Framework (EGAF) is the public bar for verified software delivery: ten industry profiles and per-commit conformance with evidence.
EGAF Industrial profile for IEC 61508 and rail
2026-08-30 — The EGAF Industrial profile (L3) crosswalks IEC 61508-3 and EN 50128. How the framework works for industrial and rail safety-related software.
EGAF Automotive profile for ISO 26262 software
2026-08-30 — The EGAF Automotive profile (L3) crosswalks ISO 26262 parts 6/8/9 and MISRA C:2012. How the framework works, and why ASIL comes from the adopter's HARA.
EGAF Aerospace profile: L3 safety-critical software
2026-08-30 — The EGAF Aerospace profile (L3) crosswalks DO-178C, DO-330/331/333, ARP4754A, and ARP4761. How the framework works for airborne software today.
EGAF AI-Regulated profile for high-risk AI systems
2026-08-30 — The EGAF AI-Regulated profile (L2) crosswalks the EU AI Act, ISO/IEC 42001, OWASP LLM Top 10, and OWASP Agentic ASI, and pulls AGT-10 into L2.
EGAF Public Sector profile for government software
2026-08-30 — The EGAF Public Sector profile (L2) crosswalks NIST 800-53, CISA SSDAF, NIST CSF 2.0, Section 508, and EN 301 549. How the framework works here.
EGAF Defense profile: L2 assurance for CUI programs
2026-08-30 — The EGAF Defense profile (L2) crosswalks CMMC / NIST 800-171, NIST 800-53, NIST 800-161, and CISA SSDAF. How the framework works — and what it does not scope.
EGAF Health profile: L2 assurance for medical software
2026-08-30 — The EGAF Health profile (L2) crosswalks HIPAA, HITRUST, IEC 62304, 21 CFR Part 11, and FDA premarket cybersecurity. How the framework works for health software.
EGAF Financial profile: L2 assurance for regulated finance
2026-08-30 — The EGAF Financial profile (L2) crosswalks PCI DSS, SOX ITGC, NYDFS Part 500, GLBA, EU DORA, and SR 11-7. How the framework works for finance.
EGAF OSS Maintainer profile for open-source projects
2026-08-30 — The EGAF OSS Maintainer profile is L1 plus OpenSSF Scorecard and OSPS Baseline. How the framework works, and what publishing open source adds to Core.
EGAF Core profile: L1 assurance for verified delivery
2026-08-30 — How the Evidence-Gated Assurance Framework works, and what the Core (L1) profile requires of every adopter. Conformance is per-commit, with evidence attached.
Koinessa Limitations and Status: Candor as Evidence
2026-08-30 — Pre-release, founder-operated, unaudited: the Koinessa limitations report — zero-row ledgers, two corruptions, no multiplier, candor as the trust strategy.
AI Context Economics: Retrieval Without the Multiplier
2026-08-30 — Five measured bounds on agentic retrieval, from $0.32 vs $0.74 per correct answer to 92.17% cache reads — and why no compound multiplier is claimed.
Local-First AI Delivery: Cost Advantage by Architecture
2026-08-30 — Local-first AI delivery as architecture, not ideology: an escalation ladder from micro models to frontier APIs, and the 30-day token receipt behind it.
Evidence-Gated Agentic Development: Inside the EG Loop
2026-08-30 — How the evidence-gated EG loop works: ten stages, a six-state result model, fail-closed verification, and why the repair agent may never weaken your tests.
The Koinessa Brain: Five Planes, One System
2026-08-30 — How the Koinessa Brain unifies scattered engineering and business state: five planes, a single-writer gateway, MCP/A2A security gaps, durable memory.
Seven Lessons From Fifty Years of Systems
2026-08-30 — Parallel agents collide without leases, memory, and proof. Seven classic systems techniques, from fencing tokens to sagas to SLSA, applied to agentic work.
The Verification Gap: Proving AI-Written Code Is Ready
2026-08-30 — AI agents produce code faster than teams can verify it. What the 2025-2026 research actually measured, and how evidence-gated delivery closes the gap.
The Verification Gap in Agentic Development
2026-08-30 — AI agents generate code faster than teams can verify it. The Koinessa white paper's full argument: five planes, four honesty labels, and fifty years of proof.
Koinessa Roadmap: Eleven Directions, Labeled Honestly
2026-08-30 — A roadmap item without a falsifiable exit condition is marketing. Walk Koinessa's eleven directions with live / in flight / built-but-gated / planned labels.
Cost Is the Floor, Control Is the Multiplier
2026-08-30 — Intent-to-deployment provenance, IAM-grade identity, three tenancy properties, and a real swarm receipt: how Koinessa makes agentic speed governable.
Risk-Tiered Autonomy: How Koinessa Gates Agent Authority
2026-08-30 — Koinessa's R0-R3 tier model grades agent actions by blast radius, enforces policy as versioned code, and keeps autonomy raisable with a full audit trail.
Why Function and File Complexity Matter for Verified AI Delivery
2026-08-08 — Complex code stalls pipelines and hides bugs. Function and file complexity become fail-closed evidence for safer, faster, AI-assisted delivery.
What an AI Operating System for Software Delivery Actually Is
2026-08-08 — An AI OS for software delivery is not a chat wrapper. It coordinates agents, verifies work with signed evidence and stores outcomes in persistent shared memory.