Skip to main content

Blog

Safe AI field notes and product essays

Seven connected articles on harm-aware AI, social-sector workflows, and software assets available for acquisition.

Products7 min read

Real people should never be the testing ground

CallCraft is a simulated voice-practice space for crisis-support conversations, built so trainees get their reps before anyone in crisis depends on them.

Read article →
Essays5 min read

Could This Harm Someone? A Build Checklist for High-Stakes AI

I build AI for people in crisis. This is the first question I ask, and the five practices that turn it into gates that can actually kill a feature.

Read article →
Products7 min read

An AI answer can sound right and still be dangerous

EchoCheck is a workbench for judging AI responses in survivor-serving contexts, built for reviewers who need more than a 'looks good' to go on.

Read article →
Products7 min read

Protective order software should fail closed

Protectly is a Texas protective order drafter built so that safety beats speed and privacy beats convenience, every time the two are in tension.

Read article →
Products6 min read

Policy Work Needs a Better Operating Surface

Texas Legislature Tracker pulls bills, districts, legislators, committees, and advocacy drafting into one place, so a small team can get from information to action without stitching five systems together by hand.

Read article →
Essays5 min read

The 7% Problem: Why Nonprofit AI Stalls at "Faster Drafts"

92% of nonprofits use AI. Only 7% got more capable. The gap is not a tools problem, it is a workflow problem, and it is the most important number in the sector.

Read article →
Essays4 min read

The Anthropologist's Advantage in Social-Sector AI

Most software for vulnerable people is built by teams who never sat with them. Fieldwork-first design is a real advantage and a real form of risk reduction.

Read article →