william argo

william argo

AI Behavioural Evaluation
#99314246 followers0 following
đź‘Ź 11 day streak
#19,640Last month
0 KP
>10,000All time
0 KP

About

I research how AI behaves when it meets real work. It started as fascination with what these systems could do, then turned into documenting the surprising failure modes nobody was talking about. The limits and the hype pushed me into writing papers on how these systems could be improved, and the testing methods I use grew out of that work. None of it was planned. It came from seeing gaps everywhere and building what the field was missing. I share my research and insights through Inquisitor Labs.

Badges

Tastemaker
Tastemaker
Gone streaking 10
Gone streaking 10
Gone streaking
Gone streaking
Gone streaking 5
Gone streaking 5

Maker History

Forums

•

6d ago

Drift Anchor – Prompt Edge Ext - Six handy prompt slots for working with any web‑based AI.

AI models forget rules, context and critical info, breaking workflows. Drift Anchor is a small Edge extension with six persistent prompt slots so you can group instructions and re‑inject them instantly. It works with any web‑based AI such as Claude, Chat GPT, Copilot, or even local models. It reduces retyping and frustration and keeps behaviour stable. Don’t fight a drifting AI model – anchor it with a prompt injection.
•

23d ago

Vectored Conversational AI Testing - Structured testing for free‑flowing AI conversations.

Vectored Conversational AI Testing is a behavioural methodology for evaluating AI through structured, multi‑turn dialogue. It wraps free‑flowing conversation in a repeatable, auditable test framework and surfaces drift, recovery, escalation, and failure across full conversational arcs. Built for developers, auditors, and governance teams, it complements the LLM Inquisitor methodology.
•

2mo ago

Evaluate & Test AI Using Real-World Work - AI can't be tested with magic prompts -only real work.

A practical field manual for evaluating and testing AI systems using real‑world work, now aligned with the requirements of the EU AI Act. Learn how to expose hidden failures, assess risks, and build reliable AI systems before deployment.
View more