LatentMesh

The evidence plane for production AI.

From obligation to proof: how regulatory obligations become controls, evaluations, evidence, and incident response.

Read: The Evidence Plane for AI Systems → Try: EU AI Compliance Checker
Evidence loop: Obligation, Control, Eval, Evidence — with coverage and verification status
Obligation → Control
Eval → Evidence

By Karthik Mahalingam. Engineering leader at Microsoft, working on Purview Data Lifecycle Management: retention, compliance, and governance services for large enterprise environments.

What I write about

New here? Start with the AI Compliance guide →
≡

AI compliance

How regulatory obligations become engineering controls, evaluations, and audit-ready evidence.

Start with the AI Compliance guide →
⦿

Agent failures

How agent systems break, drift, and misbehave once they leave the demo environment.
⌖

Evaluation

Practical evals for regression, reliability, and operational trust in production.
⊘

Safety

Guardrails, control surfaces, observability, containment, and human review loops.

Featured series

All series →

Reliable Agent Systems

How to design, test, and operate individual agent systems reliably.

Start here if you are building one agent, workflow, or AI system and need to understand how it fails, how to evaluate it, and how governance becomes operational.

Read the series → New series

Evaluating Agent Fleets

How to allocate limited evaluation capacity across many agents, classes, workloads, and changes.

Start here if your problem is no longer one agent. This series is about fleet-scale evaluation: work units, baseline inheritance, trigger-routed re-evaluation, sampling, eval packs, and evidence.

Read the series →

Latest essays

View all →
Evaluation
Evidence Is What Someone Can Verify Later
An eval result becomes evidence only when the party that needs to verify a claim can retrieve it, interpret it, and verify it. Production alone is not enough. Retention alone is not enough.
May 2026 8 min
Evaluation
The Eval Pack Belongs to the Class
An eval pack is the reusable unit of evaluation for an agent class. It contains scenarios, scorers, thresholds, expected evidence, and obligation mappings — the content that turns a test suite into traceable evidence.
May 2026 8 min
Evaluation
Drift Is When the Queue Does Not Know
Change routing works only for visible change. Drift is the movement that invalidates evidence without producing a change record.
May 2026 7 min
Compliance
I Built a Free EU AI Act Compliance Checker
A schema-driven way to turn scope questions into a preliminary risk classification, triggering reasons, and a downloadable record.
May 2026 4 min
Evaluation
Baseline Inheritance Is How Agent Evaluation Scales
An instance inherits its class baseline only while it stays inside the boundary the baseline was proven against. The hard part of fleet evaluation is detecting the moment that boundary has been crossed.
May 2026 8 min
Evaluation
Evaluation Should Follow Change
Calendar-driven eval cadence wastes capacity on stable systems and misses risk on changing ones. Change-routed evaluation matches eval work to what actually changed.
May 2026 8 min