---
title: "Work"
description: "Three cases from one build: voxcog, taken from prototype to a launch-ready platform and handed over in August 2026."
url: https://torn.studio/en/work
locale: en
---

# Work

Every case is built the same way: a short answer at the top, figures a reader can count in the client’s own repository, and a line on what the work proves.

## Work

- [voxcog: AI agents that hold up in production](https://torn.studio/en/work/voxcog-ai-agents-in-production) — Torn Studio built voxcog’s agent runtime: 55 tools behind one agent that fetches them on demand, three model tiers so every call goes to the cheapest one that can do the job, and an evaluation gate that runs in CI on every change.
- [voxcog: product decisions with a traceable trail](https://torn.studio/en/work/voxcog-evidence-chain-product-decisions) — Torn Studio ran voxcog’s product work from prototype to MVP. Every feature was planned as a written spec — 108 of them are still in the repository — and every directional call was written down with its options and what the choice cost. A pivot in June 2026 narrowed five pillars to one pipeline.
- [voxcog: a design system in code, and user testing](https://torn.studio/en/work/voxcog-design-system-and-user-testing) — Torn Studio built voxcog’s interface on a design system in code, with OKLCH tokens and a documented rule set. A research round produced seventeen findings; eleven held, one held partly, and five fell when checked against the code. Twelve changes shipped.
- [Changemkr: four systems became one platform](https://torn.studio/en/work/changemkr-four-systems-one-platform) — Torn Studio held the product role at Changemkr and owns part of the platform. Four inherited systems became five deployed services around one schema authority of 63 tables, and eight AI environments share a single LangGraph service with versioned prompts. This page describes the work up to June 2026, where the studio’s period ended.
- [27 findings from a consultancy review, worked through](https://torn.studio/en/work/consultancy-review-27-findings) — A Nordic consultancy tested an AI-driven planning canvas and reported it as untrustworthy. Torn Studio turned the report into 27 traceable tickets, root-caused them before writing any code, fixed them and verified against a running stack. At the verification on 25 June 2026, 22 were implemented, 3 partial and 2 deferred.