Michelle - Your Briefing - May 29

Concepts decoded

The handful of terms that have come up that we never properly explained. Every one gets a plain-English version and a line you can use in a sales conversation.

"Heads", the word Joann used that lost you

A "head" is just a specialist inside the model. Picture our engine as a single employee with eight different specialties: one is great at reading anxiety, one at catching sarcasm, one at spotting when a mood is sliding, and so on. Each specialty is a head.

When Joann says "we need data for all 8 heads," she means we have to train each specialist with enough examples to be good at its one job. We have solid coverage for about half of them today and are filling in the rest.

"Middleware", the layer in the middle (you had the right word)

You called it the middleware and worried it was not the right term. It is exactly right. Here is the picture you drew on the call, and it is correct: we sit in between the user and the platform that paid us, quietly handing the platform a better, emotionally aware response before it replies. We never talk to the user directly. We make their AI smarter.

"Drift", the thing the demo is actually showing

Drift is the user's emotion sliding over a conversation. In the Maya example, the user starts out praising her coworker, then slowly turns sour. A normal chatbot misses the slide. It either mirrors her (ChatGPT) or just flattens her back to neutral (Claude). Ours sees the slide happening and responds to where she is heading, not just her last sentence.

E-DNA, why "it gets better the longer you use it"

E-DNA is the memory of who a person is emotionally: their triggers, their rhythms, the people in their life who set them off. The longer someone uses a product we power, the more their E-DNA learns, and the more useful the AI becomes. A competitor starting from zero cannot catch up to a year of someone's E-DNA. That is the longitudinal value you loved, and the switching cost that locks our partners in.

Your questions, answered

Pulled straight from what you asked on the demo call. These were the most useful questions in the room, by the way. If they confuse you, they will confuse a CFO too.

Central to our whole pitch. We do not own the user's data. We never even touch it. The user's emotional profile lives inside the partner's app, encrypted, controlled by the user. They can hit delete and it is gone (that is the "dramatic ending" of the demo). What we keep is the learning, not the data. As Patrick put it well: we do not keep the conversation, we learn the rhythm.

Two reasons, both in our favor:

• They chose coding instead. The frontier labs hit a fork: lean into emotion (where the lawsuits and the GPT-4o headaches live) or lean into coding (where the easy money is). They all sprinted toward coding. The emotional lane is wide open, on purpose.
• It is a different science. Emotion is not their architecture. It needs longitudinal memory, emotional safety, and identity, things their systems are not built around. They would need years to retrofit it.

And the reason ours is different from their failed attempts: the empathetic model everyone misses (GPT-4o) was dangerous, pure empathy with no judgment, which is what drove the lawsuits. We are building emotional understanding with guardrails, the safe version they walked away from rather than fix.

Yes, B2B is the play. We sell to platforms, they reach the consumers. (A consumer companion app is a tempting future move given the "bring back 4o" crowd, but it is a huge investment and a big strategic turn. Parking it, not killing it.)

On the device question: early on, the emotional profile rides along inside whatever app the user is on, on whatever device. The "everything lives on the phone" line in the business plan is a later phase, once we move the profile onto the device itself so it can follow the user between apps. For now it works wherever the partner's app works, phone or laptop. Cross-device syncing is a known next-phase problem, not a gap we are hiding.

The one-liner Joann landed on says it perfectly: "We make chatbots have a higher EQ and prevent you from getting sued."

Where things stand

A quick state of play for the things in your lane.

• The demo (v2) is the clean one you liked. The next version adds the side-by-side against ChatGPT and Claude, a live "what the engine is doing" panel, and an E-DNA dashboard to show the long-term value. All things you and the team asked for.
• Data is in good shape for phase one. Joann has the sources she needs. The multimodal (voice) data conversation with Patrick's contact, Emily, is the one that matters next.
• Pivot (our build partner) delivered what they promised. Joann is running it end-to-end and is now the day-to-day contact.
• "Independent" is now a headline word in the business plan, per your call. It is our halo effect, keep leaning on it.
• Cal / Vera. You are right to keep him warm for 2 to 3 weeks. We will have a real product to show by then.

Want to self-serve any question?

You now have two tools so you never have to wait on me or Joann.

Matt @ EAII Tech Team