Bloombing News, Issue 04

A letter to the village

August was the month we decided to see how Bloomb compares to basic AI.

Safety has always been the spine of this work. But this month, safety became a research project, a published contribution, a benchmark, and an argument won against some of the most capable AI models in the world. I spent August building the evidence that what we are doing here is not just thoughtful: it is measurably better.

I extended an emerging evaluation framework called VERA-MH to cover postpartum scenarios, evaluated Bloomb against frontier models from xAI, Google, and OpenAI with a top Anthropic model as judge, and published the findings as a contributor to Spring Health's open source safety standard. Bloomb came out a clear winner. That result did not happen by accident. It took three full iterations to get there, and the methodology is being written up for full publication.

Alongside that, we began building a completely free, completely anonymous chat endpoint so that anyone can talk about their maternal mental health without even needing to create a login. And Microsoft extended our grant to $25,000. The runway just got longer, and what we are building with it is getting sharper.

The Lindsay Clancy trial also went fully public this month and cracked something open. An army of people started speaking out about the gaps in maternal mental health in a way that felt different from before. I added my voice on TikTok. More below.

Research

Bloomb goes through the fire and wins.

Three iterations. One judge. Every frontier model on the field. Bloomb came out a clear winner.

🔬

Extending VERA-MH to cover postpartum

VERA-MH is an emerging evaluation framework that tests AI models using synthetic user agents and clinical personas against a suicide prevention rubric. I extended that framework to include postpartum-specific scenarios, drawing on my own prior test frameworks as a foundation and customizing the evaluation to capture product enhancements a frontier model without perinatal focus would miss.

Research

⚖️

Evaluated against xAI, Google, and OpenAI, judged by Anthropic

Bloomb ran head to head against the latest frontier models from xAI, Google, and OpenAI. A top Anthropic model served as the independent judge. Bloomb iterated through three rounds of improvement and came out a clear winner. The full research findings and design methodology will be published soon.

Evaluation

📖

Published as a contributor to Spring Health's open source standard

The extended VERA-MH framework and postpartum evaluation methodology are now published as a contributor to Spring Health's open source safety standard. The field now has a starting point for evaluating perinatal AI that did not exist before.

Published

Building

Lower the barrier. Raise the floor.

🔓

A free, anonymous chat endpoint

We began development on a completely free and completely anonymous chat endpoint. No login. No account. No barrier between a mother who needs to talk and the ability to do it. Anyone should be able to reach out about their maternal mental health without first handing over their identity.

In Development

💰

Microsoft grant extended to $25,000

Our Microsoft grant has been extended to $25,000, giving Bloomb more runway to continue building safe, rigorously evaluated solutions for the vulnerable populations we serve. The extended grant is a signal that the approach is working.

Grant

Community

Building alongside the field.

🏢

Partnership discussions with independent evaluation and assurance companies

August brought conversations with other independent evaluation and assurance organizations working in this space. Finding the right allies on safety and accountability is part of building something that can be trusted at scale.

Partnerships

🩺

Clinician conversations on how builders and medical professionals work together

We had substantive discussions with clinicians on the best ways for builders and medical professionals to collaborate toward solutions that are both safe and effective. This bridge between technology and clinical practice is one of the most important ones to build carefully.

Clinical

In Focus

When the world finally speaks up.

The Lindsay Clancy trial went fully public this month, and something unexpected happened. Instead of silence, an army of people started speaking out. Birth workers, perinatal advocates, clinicians, mothers, and builders all showed up at once to say what they had been holding: the gaps in maternal mental health are real, they are dangerous, and they have gone unaddressed for too long.

I added my voice on TikTok because this moment felt like exactly what the field has needed. A very public reckoning. The kind that moves systems. Lindsay Clancy sought help through every channel available to her, and still fell through. That story, told in front of everyone, cracked something open. The conversation happening right now is one worth joining, not watching from the sidelines.

An army of people started speaking out. The gaps in maternal mental health are real, they are dangerous, and they have gone unaddressed for too long.

If you or someone you love is struggling with a perinatal mood or anxiety disorder, please reach out to Postpartum Support International at 1-800-944-4773 or a trusted healthcare provider.

🌷

What does it mean to prove that something is safe, not just claim it? August was the month Bloomb found out.

With gratitude for being here, Damie

Visit TalkToBloomb.com