Blog
Notes on AI agents, evaluation, and the engineering around software that takes real actions.
How RedDial pressure-tests conversational agents with LangGraph and RAG
RedDial sends difficult customer personas to an agent, then turns the transcripts into an auditable report. This writeup walks through the engineering behind it: LangGraph for parallel runs, LangChain for structured output, RAG for groundedness, and decision trees for scoring.
PermitRail: a permission layer for AI agents that take real actions
Open-source TypeScript infrastructure for policy checks, signed proofs, replay protection, and audit receipts.
Why Docgate Matters ↗
Why document understanding matters for immigrants and residents navigating official paperwork.
Age Gates Without the Privacy Risk: How Selective Disclosure Changes Online Verification ↗
How selective disclosure can change age verification without exposing unnecessary identity data.
The End of Passport Selfies: How EUDI Wallets Are Set to Replace Legacy KYC ↗
Why EUDI wallets can replace legacy identity verification flows.
Verify a User's Age in 3 API Calls: A Developer's Guide to EUDI Wallets ↗
A developer guide to age verification with EUDI wallets.
Why Every EU Business Needs an EUDI Strategy Before 2027 ↗
A practical view on EUDI strategy before the 2027 horizon.
OpenID4VP in Production: The End-to-End Integration Checklist ↗
A production checklist for OpenID4VP integration.
Nuxt 2 to Nuxt 3: What I Actually Did and Why It Finally Worked ↗
A practical migration write-up from Nuxt 2 to Nuxt 3.