EMPIR

The publishing platform for AI/ML minds.
Native LaTeX, syntax-highlighted code, and built-in monetization.

Inline LaTeX math Server-rendered code blocks AI/ML subfield discovery Stripe Connect payouts
# Policy gradient with entropy bonus
def compute_loss(logits, actions, rewards, β=0.01):
  π = F.softmax(logits, dim=-1)
  log_π = F.log_softmax(logits, dim=-1)
  H = -torch.sum(π * log_π, dim=-1# 𝐻(π) = −Σ π log π
  return -(log_π[actions] * rewards + β * H).mean()
Launching soon — founding writers welcome