Socria Logos

The full reasoning environment: the conversation and a live map of your thinking, side by side.

Logos is the model that made Socria an environment. Select it and the chat becomes a split screen: your conversation on the left, and on the right a Thinking Map that draws the structure of what you are saying — claims, assumptions, tensions, evidence — live, while you talk. On a phone the two become tabs; the thinking is the same.

The split screen#

The map is not an illustration of the reply — it is a second, parallel reading of you. Its header names the kind of thinking it currently sees (Deciding, Learning, Writing, Math…), inferred from the conversation rather than chosen from a menu, and it moves as the conversation moves. When an extraction reorganizes things, the panel says what changed — “merged 1 · resolved 2” — and the affected nodes flash briefly so you can see where.

What happens when you send a message#

Every message you send fires two independent requests in parallel: one streams the conversational reply; the other re-reads the recent conversation and produces a fresh map. The map is deliberately not derived from the reply — it reorganizes while the answer is still streaming, and if an extraction fails or lags, the conversation never notices. The map is allowed to skip a turn; the thinking is not.

How it talks#

Two to four sentences of plain prose. One move per turn — ask, notice, challenge, connect, clarify, explain, acknowledge, or leave space — never several at once, and not every reply ends with a question: one appended out of habit teaches you to stop reading the last line. Therapist reflexes (“It sounds like…”, “What I’m hearing is…”) are banned as openers, and Logos never narrates the map at you — the map itself is how it demonstrates understanding.

Logos has its own depth control beside the composer. As its footer says: depth changes how deeply Logos helps you think — never how quickly it gives answers.

The boundary#

The prompt’s own worked examples draw the line cleanly:

• “Write my essay” — it won’t.
• “Make my paragraph clearer” — it will.
• “What does this study say?” — it explains. Understanding is not authorship.

When the work is mathematical, this boundary takes its sharpest form — the Answer Guard.

Lines of thinking#

The left rail keeps every line of thinking, each with a small deterministic thumbnail of its map’s shape — you recognize a session by the shape its reasoning took, not by reading a title. A session holds four things together: the conversation, the map, the draft if one exists, and any grounded material — reopening one restores all of them.

Signed in, sessions sync to your account and appear in the same sidebar as your Core conversations, marked with the Logos glyph. Unlocked by key without an account, they stay in the browser. Deep links work either way: /chat?model=logos opens Logos, and /chat?model=logos&s=<session> opens a specific line of thinking.

Bringing your own material#

Paste something long and it becomes an attached note instead of flooding the box; attach images and Logos reads them once, at attach time, so the reply and the map work from the same reading. Every pasted or uploaded note carries an origin you can correct — My thinking, Source material or Context — and the origin decides what the map may do with it: someone else’s argument can become a claim or a source on your map, but never one of your beliefs. Images enter as context — described, never adopted as positions. Grounding a specific node in material is covered under Draft Space & grounding.