Because "the AI wrote it" and "we can't control the cost" are not answers for your auditor, your board, or your CFO

SAN FRANCISCO, June 9, 2026 /PRNewswire/ -- Every day, engineering teams at thousands of companies are shipping code written by AI. Claude Code, Cursor, Copilot -- the tools have changed how fast software gets built. What hasn't changed: those same companies have no verifiable record of which lines were AI-generated, what it cost, or who authorized it. That's not a gap in visibility. That's a liability.

Avinash Lakshman -- the co-creator of Amazon Dynamo and creator of Apache Cassandra, two of the most consequential database systems ever built -- has spent his career solving exactly this kind of foundational infrastructure problem. Today, through his company Weilliptic, he is launching Codensics: the missing cost governance and provenance layer for engineering organizations deploying AI coding agents at scale.

AI-assisted development today feels a lot like the early days of cloud adoption -- powerful, fast-moving, and largely ungoverned. As coding agents write functions, refactor modules, and ship production code autonomously, two questions become unavoidable: can you prove what your AI built -- and do you actually know what it's costing you?

Codensics is Weilliptic's answer: verifiable proof of every AI-written line of code, per-developer token governance with fail-closed enforcement, and a full attribution dashboard for the executives accountable for both -- without requiring infrastructure changes or provider cooperation.

Predictable AI Spend Without Slowing Developers Down

Token consumption scales with team size, project complexity, and usage patterns in ways that are notoriously difficult to forecast. The result is familiar to every finance organization: unexpected invoices, limited attribution, and little ability to distinguish strategic AI investment from uncontrolled consumption.

Most organizations manage AI costs reactively. They review invoices after the fact, establish soft guidelines, and rely on teams to self-regulate. That approach may have worked for cloud adoption a decade ago; it is increasingly untenable for autonomous coding agents consuming millions of tokens per day.

Codensics introduces a governance model designed for both engineering agility and financial accountability.

Organizations can allocate monthly token allowances to individual developers, teams, projects, or cost centers before usage begins. These allowances function as enforceable spending envelopes rather than advisory quotas, giving finance teams a predictable view of expected consumption and enabling accurate forecasting before the invoice arrives.

When additional capacity is needed, organizations can define policy-driven top-up mechanisms. Team leads, engineering managers, or designated approvers can authorize incremental token allocations without waiting for the next budgeting cycle. Every adjustment is recorded, attributed, and auditable.

Codensics also supports budget transfers between developers and teams. If one project underutilizes its allocation while another exceeds expectations, unused token capacity can be reassigned instantly rather than wasted. This creates a market-driven allocation model inside the enterprise while preserving financial controls and full auditability.

For organizations that require strict governance, fail-closed enforcement ensures that once an allowance is exhausted and no approved top-up exists, agent activity stops automatically. Not a warning. Not a recommendation. An enforceable control.

The result is a system that transforms AI spending from an unpredictable operational expense into a managed and forecastable resource. Finance gains visibility into expected spend, engineering gains flexibility to meet delivery goals, and executives gain confidence that AI adoption will not create surprise liabilities at the end of the quarter.

The Audit Is Coming. Can You Prove What Your AI Built?

Git tracks who committed. CI/CD tracks what deployed. Neither tracks provenance -- whether a function was human-authored, AI-generated, which model produced it, under what authorization, and at what cost. For organizations in regulated industries, heading into M&A, preparing for a SOC 2 audit, or defending an IP claim, this is not a documentation inconvenience. It is a material risk.

The EU AI Act, emerging SEC guidance on AI in financial services, and evolving IP frameworks are all moving in the same direction: organizations will be held accountable for what their AI systems produced. The question "can you prove what your AI wrote and when?" is moving from philosophical to legal.

Traditional approaches -- linting rules, honor-system disclosure, commit comments -- provide no verifiable guarantees. What's needed is diff-level attribution, signed by a verifiable agent identity, anchored on immutable infrastructure. This is what Codensics delivers.

How Codensics Works

Codensics is built on three capabilities that no existing developer tooling provides simultaneously.

Diff-level AI provenance. Every AI-generated code change is captured at the diff level, keyed to the commit hash, and tamper-proof and independently verifiable on WeilChain -- not dependent on trusting Weilliptic's logs, the model provider's records, or any third party. When an auditor asks what your AI wrote in Q3, the answer is a verifiable record, not a spreadsheet.

Per-developer token budgets with fail-closed enforcement. Set monthly token budgets per developer across Claude Code, Cursor, and any Git-integrated coding agent. When a developer hits their limit, the agent stops. Not a notification. A hard stop. No ungoverned overage. No surprises on the invoice.

Executive attribution dashboard. Every developer, every model, every commit: attributed, timestamped, and queryable. Built for the people accountable for the answer -- not the developers writing the code.

Codensics integrates with Claude Code, Cursor, and any Git workflow -- no model replacement, no infrastructure changes, no provider cooperation required. Powered by WeilChain, the verifiable execution layer that makes the audit trail tamper-proof rather than merely thorough.

"The internet era taught us that scalability couldn't be an afterthought. The AI era will teach us the same about trust -- and cost. Every line of code an AI writes without a provenance record is a liability that compounds quietly. Until it doesn't."
-- Avinash Lakshman, Founder and CEO, Weilliptic

About Avinash Lakshman

Avinash Lakshman is the co-inventor of Amazon Dynamo and the creator of Apache Cassandra, two of the most widely deployed distributed systems in history. He founded Hedvig, a distributed cloud storage company acquired by Commvault in 2019. He is the Founder and CEO of Weilliptic.

About Weilliptic

Weilliptic builds cryptographic trust infrastructure for enterprise AI. Its platform, WeilChain, provides provenance, immutable audit trails, and verifiable execution records across browser, agent, and coding environments. Products include Receipts (browser AI audit trail), Agent Receipts (autonomous agent governance), and Codensics (AI code provenance and token governance).

View original content:https://www.prnewswire.com/news-releases/the-man-who-built-amazon-dynamo-and-apache-cassandra-is-now-solving-ais-accountability-crisis-302795348.html

SOURCE Weilliptic