Skip to content
VZU
VZU Industries · Media / News

Proofreading. Editorial intelligence. The editorial bar held .

VZU's media / news playbook. Proofreading. Editorial intelligence. Workflow optimization. Multilingual content. Archive digitization. Powered by Oracle (prompts) and Pixel (canvas) on the VZU OS runtime. The brief is the contract. The work is the work. The audit trail is per-article.

Industry

Media / News

95%+

NZZ first-pass accuracy

Direct answer

What does VZU do for media / news?

VZU's media / news playbook is editorial AI, proofreading, and content production on the VZU OS runtime. Oracle writes the prompt. Quill runs the proof. Sentinel audits. Real deployments: NZZ proofreading at 95%+ first-pass accuracy on a 2,400-articles-a-week Swiss news desk, F500 Broadcaster editorial AI at 85%+ accuracy.

Use cases

Five deployments. Real numbers.

  • Proofreading

    NZZ (Swiss national newspaper). 95%+ first-pass accuracy. 2,400 articles a week. German, French, Italian. The Oracle agent wrote the prompt. The Quill agent ran the proof. The Sentinel agent audited the corrections.

  • Editorial intelligence

    F500 Broadcaster. 85%+ accuracy. The Oracle agent wrote the editorial prompt. The Quill agent ran the editorial check. The Sentinel agent audited the changes. The editorial bar held.

  • Workflow optimization

    NZZ. The Hunter agent pulled the article. The Quill agent ran the proof. The Atlas agent routed the article to the right desk. The audit trail is per-article.

  • Multilingual content

    14 languages. Regional accents. The Oracle agent wrote the prompt per language. The Quill agent ran the proof. The Sentinel agent audited. The translation is on the audit trail.

  • Archive digitization

    NZZ legacy archive. The Hunter agent pulled the scanned article. The Oracle agent wrote the OCR prompt. The Quill agent ran the extraction. The Sentinel agent audited. The archive is AI-native.

A case study · NZZ

2,400 articles a week. 95%+ first-pass accuracy. The editorial bar held.

Chapter 01 · The brief

NZZ, the Swiss national newspaper. 2,400 articles a week.

NZZ is the Swiss national newspaper. The proofreading desk handled 2,400 articles a week across German, French, and Italian. Each article was a first-pass proofread, a style check, a fact check, and a desk handoff. The desk was 14 people. The work was 90% mechanical. The desk wanted to free the proofreaders for the 10% that needed a human.

The brief from NZZ was specific. Build a proofreading pipeline. Hit 95%+ first-pass accuracy. Hit 90%+ on style. Build a per-article audit trail. The editorial bar had to hold. The work had to scale. The brief is the contract. The work is the work.

2,400

Articles a week

Chapter 02 · The architecture

Oracle, Quill, and Sentinel on the VZU runtime.

The Oracle agent wrote the prompt. The Quill agent ran the proof. The Sentinel agent audited the corrections. The audit trail is per-article. The editorial bar held. The 14 proofreaders went to the 10% that needed a human.

“95%+ first-pass accuracy. 2,400 articles a week. The editorial bar held. The audit trail is per-article. The work is the work.”

Chapter 03 · The result

95%+ first-pass accuracy. The editorial bar held.

The pipeline shipped in 8 weeks. 95%+ first-pass accuracy. The editorial bar held. The 14 proofreaders went to the 10% that needed a human. The audit trail is per-article. The brief is the contract. The work is the work.

95%+

First-pass accuracy

The Oracle agent · A real production prompt

Watch the Oracle agent ship a proofing prompt.

The Oracle agent writes, versions, evals, and deploys production system prompts. 50+ eval cases. Iterated until 100% pass. The audit trail records the version, the eval, the deploy. The brief is the contract. The work is the work.

oracle@vzu-os — trace.log live
  1. 01
    oracle read brief

    read brief

  2. 02
    oracle thinking: voice agent prompt. 4 intents. 50+ eval cases.

    thinking: voice agent prompt. 4 intents. 50+ eval cases. : voice agent prompt. 4 intents. 50+ eval cases.

  3. 03
    oracle mcp.prompt.write

    v1: 1,247 tokens, 4 intents : 1,247 tokens, 4 intents

  4. 04
    oracle mcp.eval.run

    47/50 pass, 3 fail (escalation edge cases)

  5. 05
    oracle mcp.prompt.iterate

    v2: add explicit escalation policy : add explicit escalation policy

  6. 06
    oracle mcp.eval.run

    50/50 pass

  7. 07
    oracle mcp.deploy.staging

    staged to staging.atlas.vzu.ca

  8. 08
    oracle ship. 1 week. 50 eval cases. 1,247 token prompt. done

    ship. 1 week. 50 eval cases. 1,247 token prompt.

oracle@vzu-os $

Audit trail · 6 entries

streaming
  • T+00:00:00 brief.received 01
  • T+00:00:30 prompt.v1.drafted 02
  • T+00:02:00 eval.v1.run 03
  • T+00:04:00 prompt.v2.iterated 04
  • T+00:06:00 eval.v2.passed 05
  • T+01:00:00 brief.completed 06
  • KPI · 01 now

    1 week

    Median prompt duration

    71%
  • KPI · 02 now

    50

    Eval cases

    5%
  • KPI · 03 now

    100%

    Eval pass rate

    10%
  • KPI · 04 now

    1,200

    Avg token count

    32%

The trust strip

What the runtime has shipped for media / news.

  • NZZ · 95%+ accuracy

  • NZZ · 2,400 articles/wk

  • F500 Broadcaster · 85%

  • NZZ · 3 languages

  • F500 Broadcaster · editorial

A case study · NZZ

Case study

95%+. First-pass accuracy.

01

The brief

A Swiss national newspaper was publishing 2,400 articles a week across German, French, and Italian. Every article needed to be proofread before publication — style, consistency, language-specific conventions. The proof desk was the bottleneck. The work didn't scale.

02

The analysis

The Oracle agent needed to write the prompt. The Quill agent needed to run the proof. The Sentinel agent needed to audit the corrections. The proof desk needed to keep the final say on every change. The runtime needed to be 95%+ accurate on the first pass to be worth the editorial overhead.

03

The build

Oracle wrote the proof prompt. Quill ran the proof across 2,400 articles a week in 3 languages. Sentinel audited the corrections and flagged the changes for the proof desk's review. The proof desk approved the suggestion, edited it, or rejected it.

04

The result

95%+ first-pass accuracy. The proof desk stopped being the bottleneck. The same pattern was applied at an F500 broadcaster — 85% editorial automation across the newsroom.

The runtime

Every VZU brand ships on the NetWit Agentic OS runtime. Cloud, hybrid, on-prem. SOC 2 + HIPAA + ISO 27001.