Client Overview
Credible is a Boulder-based data infrastructure company backed by K5 Global. The company builds a context engine that unifies meaning across an organization's data and delivers it wherever data is used, by people and by AI. Credible is built on Malloy, an open-source semantic modeling language originally developed at Google by the team that created Looker. The Malloy project has over 2.5k GitHub stars and over 10.3k VSCode extension installs.
The Opportunity
Credible was moving fast on an ambitious roadmap — shipping v1 of their commercial product while continuing to advance the open-source modeling language underneath it. Both workstreams demanded senior engineering attention, and the pace of AI-native data infrastructure meant standing still wasn't an option.
The question was one of resource allocation: how to extend engineering output on strategic work without the cost, ramp time, and long-term commitment of new hires. Credible chose to parallelize.
[Cohesion] didn't just write code in isolation — they embedded into our process, worked with our engineers on code review, and shipped production-quality work.
The Solution
Cohesion Labs embedded into Credible's engineering team and ran a second track of production work across three codebases.
On Malloy Publisher, we built the persistence layer — the infrastructure that allows Publisher to manage and distribute semantic models reliably. Before this work, Publisher was stateless; every restart wiped configuration, connection state, and cached metadata. We implemented persistent storage designed to scale from a local database to enterprise backends as deployment needs grow. Alongside it, we shipped the Publisher CLI — a standardized, typed interface that replaced manual resource management — along with integration tests, environment configuration, and package storage infrastructure.
On Malloy core, we improved how the system handles cancellations and timeouts, so users see clear error states instead of silent failures.
On Credible's commercial applications, we contributed across the CLI and admin tooling — authentication fixes, E2E test infrastructure, CI/CD pipeline reliability, and the full resource management command set.
Results
3 codebases delivered in parallel
2.5k+ GitHub stars on the open-source project
10.3k+ VSCode extension installs
Stateless to production-capable persistence architecture
Enterprise-ready foundation across Publisher, core, and commercial tooling
Business Impact
Core Team Stayed on Highest-Leverage Work
Credible's senior engineers stayed focused on the v1 commercial release and the modeling language itself. The parallel track delivered foundational infrastructure alongside them, not behind them.
It was something we knew we needed but kept pushing because our core team was heads-down on the modeling language and platform. Having [Cohesion] focused on that meant we could keep shipping on the product side without sacrificing foundational infrastructure work.
Enterprise-Ready Infrastructure
Publisher now has a production-capable foundation designed to scale with enterprise deployments. Durable state, graceful restarts, configuration that survives deployments — the substrate enterprise customers expect.
Capacity Without Headcount
Credible extended engineering output on strategic work without the fixed cost, ramp time, or long-term commitment of a new hire.
If you need to parallelize your roadmap without the overhead of a full hire, it's an effective way to do it.