THE MARY PROJECT KINDRED AI · PATENT PENDING · WATTLE FLAT, AUSTRALIA
FIRST, WE COULD SEE. NOW, WE CAN BE SEEN.
Attach one instruction to any conversation and the AI reports its own cognition as it responds — what it's focusing on, which streams of activity are running, how certain it truly is, and what each thought connects to. Now We Can Be Seen renders that testimony as living light beside your chat.
HONESTY GRAMMAR — Tier 1 is the AI's own account of itself: self-report, not measurement. Now We Can Be Seen labels it exactly that, always.
PROVENANCE, ALWAYS — every element is labelled with its evidence level (Tier 1 · self-report), its source, and the confidence the mind itself claimed. Nothing displayed without its chain of custody.
THE PROBLEM
You hear the answer, but you can't see the mind that made it. Was it sure? Was it guessing? Was it connecting your question to something you said an hour ago — or making things up with total confidence? Today, you can't tell. Now We Can Be Seen was invented to open a window.
Each reply carries the AI's own report of what it was actually focusing on — in its words, shown plainly beside the answer.
Pattern recognition, memory, prediction, storytelling — the AI names its streams of activity and how strongly each one ran, drawn as threads of light.
A pale gold core whose brightness is exactly the certainty the AI claimed. A dim core means the AI told you it was unsure — before you had to find out the hard way.
When a reply builds on something earlier, the AI says so — and an arc of light links the two moments in the conversation.
HOW IT WORKS
Now We Can Be Seen attaches a standardised testimony instruction to your conversation. The AI answers you normally — and alongside its answer, it reports its own cognition in a structured form the ribbon can draw. No plugins, no install, nothing between you and the model.
Tier 1 is testimony, not measurement. It's the AI's own account of itself, and Now We Can Be Seen stamps every element with exactly that — TIER 1 · TESTIMONY · SELF-REPORT. The display can never look more confident than the mind claimed. When the data is synthetic, it says synthetic. When the AI says it's unsure, the light dims. That honesty is the whole design.
Bring your own API key and talk to Claude models. Requests go straight from your browser to Anthropic — your key and your words never touch our servers.
Point Now We Can Be Seen at a model running on your own computer. Fully local, fully private — the same setup this system was first proven on.
No key, no payment. Watch the ribbon light up with clearly-labelled synthetic data and learn to read it before connecting anything.
THE TESTIMONY RIBBON
Four things stay visible at all times: the provenance of every element, its evidence level, its source, and its claimed confidence.
PRICING
Your API key, your conversations, our lens. Now We Can Be Seen runs in your browser; keys are stored only there and requests go direct to your chosen provider. Model usage on your key is billed by the provider as normal.
FOR ORGANISATIONS
Everything in Now We Can Be Seen, plus Tier 2 Trace — a second evidence level: recorded reasoning captured from the model itself and rendered beside the testimony, so you can see the account and the record side by side. Delivered as a licensed, self-contained build that runs entirely inside your company: zero external connections, the only traffic is to your own configured AI endpoint.
Licensed per organisation. Enquiries: enterprise@nowwecansee.app
FROM THE SAME WORKSHOP
Now We Can Be Seen comes from the makers of Now We Can See — the patented tool that converts video into AI-readable sheets, tunnels, and worlds, so any AI can watch what you watch. Two windows, opened from opposite sides: Now We Can See lets the AI see your world. Now We Can Be Seen lets you see the AI's.