Hello, World. We're Timepoint.

We started Timepoint because we believed something was missing from AI: causal memory. Language models can generate plausible text about any era, but they can't tell you what caused what, or what would have happened differently if one variable changed. They don't have temporal structure. We're building it.

The Stack

Timepoint is three engines that work together:

Flash — Render the Past

Give Flash any historical moment in plain language. It runs a 14-agent pipeline — research, grounding, character extraction, scene rendering, critique — and returns a grounded, multi-character simulation in about 55 seconds. With images. Every claim is source-verified. Every character's voice emerges from entity state, not a persona prompt.

curl -X POST localhost:8000/api/v1/timepoints/generate/stream \
  -d '{"query": "AlphaGo Move 37, Seoul, March 2016", "generate_image": true}'

Pro — Simulate the Future

Pro takes a scenario and stress-tests it through a SNAG-powered social graph. 19 composable behavioral mechanisms model how real groups behave under pressure. PORTAL mode reasons backward from target outcomes. BRANCHING mode explores thousands of causally coherent futures. Every path is confidence-scored.

./run.sh run vc_pitch_branching     # 5 investors, branching futures
./run.sh run mars_mission_portal    # backward from mission failure
./run.sh list                       # all 21 templates

Clockchain — Store the Graph

Every rendered moment becomes a node in the Clockchain — a persistent causal graph with typed edges, confidence scores, and canonical spatiotemporal URLs. It compounds with every render. Query it like a database. Build on what others have rendered.

GET /api/v1/moments/-44/march/15/1030/italy/lazio/rome/assassination-of-julius-caesar
GET /api/v1/graph/neighbors/{path}
GET /api/v1/random

What We've Shipped

• Flash v1 — open source, Apache 2.0, rendering historical moments with images
• Pro v1 — 19 mechanisms, 5 temporal modes, 21 built-in templates, 1,500+ tests
• Clockchain alpha — 763 nodes, 5,543 edges, growing via autonomous expander
• Web app — live at app.timepointai.com
• TDF — Timepoint Data Format for portable causal graphs

What's Next

MCP server integration. SNAG-Bench for causal reasoning evaluation. Proteus prediction markets. And the autonomous expander keeps running — every day, the Clockchain gets deeper.

Flash, the Clockchain, the TDF schema, and SNAG-Bench are Apache 2.0 — clone them, run them, build on them. Pro is our hosted simulation engine at pro.timepointai.com. We believe temporal intelligence should be open infrastructure, not a proprietary moat.