Vicco LabsVicco Labs
0x4A2F ƒ(x) → ∇ 01001 stream MODEL_v2.4 GUARDRAIL=on TRACE_ID=#a7 → context: 4096 → tokens: 1.2M LATENCY 84ms P99 412ms COST $0.0021 ────────────── ƒ(x) → ∇ retry 01001 stream MODEL_v2.4 GUARDRAIL=on TRACE_ID=#a7 → context: 4096 → tokens: 1.2M LATENCY 84ms P99 412ms COST $0.0021
Vicco Labs · São Paulo

Softwarefortheedge
between generative AI
andregulatedenvironments.

Build products where AI has to live with audit, scale, and pressure - and we write down the path.

Read writing
vicco-labs ~ /bench
$ build --target=regulated --ai=generative
// compiling guardrails…
→ auditoria "ok" · escala "ok"
→ p99 412ms · cost $0.0021/req
// articles · series · ai governance
$ read ./articles
20
years inside banking
7
frameworks in the governance toolkit
20×
reduction on a production prompt
Notes from the bench

Latest writing

See all writing →

Field notes on generative AI, governance, observability, and what does - and doesn't - fit inside regulated environments.

AI Security
AI governance, in practice · Part 2AI Security

What can go wrong in an LLM agent: OWASP, MITRE ATLAS, and what the frameworks don't cover

Reading the architecture of a real conversational agent through an attacker's eyes: where OWASP and MITRE ATLAS map directly onto code, and where they leave gaps the team has to close.

9 MAY 2026 · 6 min read
AI GOVERNANCE
AI governance, in practice · Part 1

Seven frameworks for governing LLM systems: the map before the territory

Before implementing controls, you have to understand the terrain. A practical overview of NIST AI RMF, ISO/IEC 42001, OWASP LLM Top 10, and the other frameworks defining what it means, today, to govern generative AI in production.

22 APR 2026 · 14 min read
FT.SEARCH
AI governance, in practice · Part 10FT.SEARCH

QueryBuilder: turning a Pydantic object into a safe FT.SEARCH query

Building FT.SEARCH queries manually is where you discover RediSearch silently interprets '&' as AND — no error, no exception.

20 APR 2026 · 4 min read