Tenex

Tenex is an AI engineering and enterprise transformation firm cofounded by Alex Lieberman and Arman Hezarkhani, both listed as managing partners. Its engineering model ties client billing and team compensation to completed work measured in company-defined story points. Its transformation services combine AI strategy with software development, workflow automation, training, and change management.[3][2][4][5]

In an account dated February 24, 2026, Lieberman said he and Hezarkhani had launched the firm eleven months earlier—placing the launch around March 2025—as “McKinsey for AI.” Lieberman brought a background as Morning Brew's cofounder; Hezarkhani had previously founded venture-backed AI companies, worked on Google's Cloud and AI platforms, and taught computer science as an adjunct professor at Carnegie Mellon University.[6][2]

Lieberman described Tenex as bootstrapped and profitable since its first day and reported winning Fortune 500 business against McKinsey, Palantir, and Accenture. He identified hiring fast enough as the firm's biggest bottleneck.[6]

Tenex's engineering service charges clients for produced work and pays its team for completed story points. Tenex defines those points itself; they are the firm's work-accounting units, rather than a standardized measure of productivity across companies. The service combines off-the-shelf AI tools with tools built internally.[4]

Dedicated teams include a technical product manager and forward deployed engineers. Clients submit tasks, which Tenex advertises breaking down and grading into story points within 24 hours, then choose priorities every two weeks. Advertised capabilities include full-stack applications, model fine-tuning, migrations and refactors, data engineering, and custom agentic solutions.[4]

Tenex's technical strategists had incentives tied to client retention and served as the final quality gate before engineering plans reached clients, according to the November 19, 2025 Latent Space episode summary. Tenex expected multiple engineers to earn more than $1 million in cash compensation from story points the following year—a forecast of 2026 earnings.[1]

Tenex reportedly produced early prototypes in two weeks for a retail-camera computer vision system providing heat maps, queue detection, shelf-stocking analysis, and theft detection, according to the November 2025 episode summary. Its reported projects also included a one-month build of Snapback Sports' mobile trivia app. In another example, a Tenex engineer spent four hours building a fitness influencer's AI health-coach prototype after the prospect initially declined, reportedly moving Tenex to the top of the prospect's vendor list.[1]

Tenex's transformation offering covers product, process, and people. It advertises 30-, 60-, or 90-day audits to identify AI use cases, estimate returns, and establish implementation roadmaps, culminating in an AI-Adoption Report with a technical appendix. Process engagements begin with employee surveys and stakeholder calls, establish baseline metrics, and implement automation alongside training and change management. Bespoke curricula and hands-on workshops address employee adoption.[5]

On August 5, 2026, Alex Lieberman, Dan Zakon, CJ Hess, and Arman Hezarkhani published Tenex's MetaHarness methodology. It addresses the risk that agent execution expands faster than teams can preserve shared understanding. Each project has a versioned markdown record beside its code, separating current facts, future plans, and history so tooling can cross-check them. Linear, GitHub, and client updates become views over that record.[7]

The Tenex CLI parses planning documents but limits synchronization writes to fenced blocks containing machine-managed IDs, status, and metadata, preserving surrounding human reasoning. External writes require a preview followed by a separate application step; ambiguous matches to Linear issues are rejected. Validation checks relationships and acceptance criteria, and blocks unsupported completion claims when they lack linked pull-request, test, or sign-off evidence.[7]

Managed startup hooks assemble architecture, prior decisions, acceptance criteria, and validation commands for new agent sessions. The August account reported more than fifty reusable skills and 30+ CLI commands. Skills cover planning, execution, investigation, review, and documentation; they are versioned, tested against fixtures, and deliberately released so improvements can spread across projects.[7]

Tenex OS presents the same structured state to engineers, engineering managers, and leadership, deriving delivery health from connected project records rather than manually entered status reports. The authors retain human accountability for shipped work and acknowledge a practical limit: for one team on one project, the process can cost more than it saves; its value increases with concurrent work.[7]