Chordio Wireframes

Chordio Wireframes is an AI chat bot tool. PX-bench from Chordio scores coding agents on the product experience decisions they make when adding features to a held-out app, across eight categories including visual craft, convention adherence and accessibility.

PX-bench is a long-horizon benchmark from Chordio that measures the design decisions coding agents make when they add a feature to an existing product. Rather than prompting an agent to build an app from scratch, the benchmark drops it into a held-out host app, a multi-screen product it has never seen with its own house conventions, and asks it to ship a new capability. The resulting build is then scored across eight categories of product experience, each tied to a decision a senior product designer would recognize. The eight categories are intent fidelity, product fit, visual craft, convention adherence, pathway completeness, content and language, resilience, and accessibility. Underneath each category sit concrete rubrics: whether the right container and pattern were chosen, whether design tokens were reused over hardcoded values, whether cancel and undo paths exist without dead ends, whether empty and error states are handled, whether axe-core violations appear, and how the layout holds up under long content and small screens. Every scenario is mapped in advance to a known-good outcome, so the agent's choices can be scored against ground truth instead of judged in a vacuum. The audience is teams building or selecting coding agents: model labs, agent harness developers, and engineering leaders who need to know where product quality breaks and what it costs to ship a feature. Scoring is automatic and anchored in expert-defined rubrics, which removes the slow human review loop that normally stalls evaluation work as models update. The harness is Inspect AI, the UK AI Safety Institute framework, and Chordio states that any published scenario can be independently rerun. Where it differs from generic coding benchmarks: those measure whether code runs, while PX-bench measures whether the result feels like it belongs in the product. Chordio also publishes methodology notes on scoring validity and run-to-run variance, so buyers can weigh how much of a score gap is signal.