veritai.org

Think-Free*

AI is massively disruptive to research, yet getting API access for a university group is near-impossible without paying out of pocket. It shouldn’t be this hard. And too much of AI today asks you to trust a black box—I think researchers deserve better. With open weights and formal verification, you can check every claim yourself. Math doesn’t care who you are or where you work. I hope these tools help you think more freely. — Dirk Englund

Sign in with your MIT or Google account.

Resources

VVUQ-Coupled LLMs mit.nonlocally.org
Play with LLMs coupled to formal verification and apply them to math, science, and engineering tasks—from quantum optics proofs to photonic IC layout. Split-screen with Marimo notebooks, GDSFactory PIC design, Mecademic robot dashboards, and other tools, and 3DOptix optical simulation for interactive computation alongside chat. Lean 4 formalization of Pauli matrices with Mathlib imports and InfoView panel in a Marimo notebook
Notebooks marimo.nonlocally.org
Marimo reactive Python notebooks with live execution. Includes GDSFactory for photonic IC design, VVUQ verification tools, and shared workspace.
LLM API llm.nonlocally.org
OpenAI-compatible REST endpoint backed by LiteLLM. Proxies to Anthropic, OpenAI, Google, and on-premise Ollama instances. Per-user usage tracking and budgets. Requires API key — see API access below.
VVUQ4AI vvuq.dirkenglund.org
Validation, Verification, and Uncertainty Quantification for AI-generated scientific claims. Lean 4 formal verification with MCP integration.

API access

The LLM endpoint at llm.nonlocally.org implements the OpenAI API. Use any compatible client.

1. Sign in at mit.nonlocally.org with your MIT or Google account.
2. Go to Settings → Account → API Keys and create a key.
3. Use it with any OpenAI-compatible client: 

from openai import OpenAI

client = OpenAI(
    base_url="https://llm.nonlocally.org/v1",
    api_key="sk-...",  # from step 2
)

resp = client.chat.completions.create(
    model="qwen3-235b",
    messages=[{"role": "user", "content": "Explain Bell nonlocality."}],
)
print(resp.choices[0].message.content)
# Or with curl:
curl https://llm.nonlocally.org/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen3-235b","messages":[{"role":"user","content":"Hello"}]}'

Available models

From the fleet. Updated by scripts/generate_models_json.py.

ModelProvider
Loading…