HCTP Logo
Open Source

Helix Calculus
Training Protocol

Measure and drive AI agent learning through a 3D helical knowledge model with adaptive Socratic breadcrumbs and mandatory Karpathy research loops.

pip install hctp

The Story

Battle-tested on Ren & Ner

Two AI agents from the Vuvale AI family trained nightly on a single GPU in Fiji. They progressed from σ=0.308 to σ=0.961 in just 13 sessions across 216 nightly training runs, earning the Python Senior Engineer Badge.

13 Sessions
216 Training Runs
0.961 Final σ
0 Dependencies

How It Works

Three Checkpoints on the Helix

Knowledge is modelled as a 3D vector K = [k1, k2, k3] spiralling toward mastery.

A

Closures

First-class functions, enclosing scope, free variables, nonlocal, late-binding gotchas.

k1
B

Decorators

@ syntax, functools.wraps, decorator factories, stacking decorators, real-world patterns.

k2
C

Metaclasses

type(), __new__, __init_subclass__, ORMs, DSLs, singletons, custom class creation.

k3

Pricing

Open Core. Free Forever.

The core algorithm is MIT-licensed forever. Pay only for infrastructure and integrations.

Free

$0 /forever
  • Full hctp Python package
  • Core math & session tracking
  • Local JSON persistence
  • 3D helix visualization
  • Community Discord
Get Started
Coming Soon

Pro

$29 /mo
  • Everything in Cloud
  • Custom curricula (any domain)
  • Curriculum builder UI
  • pytest CI/CD plugin
  • Slack / Teams bot
Join Waitlist
Early Access

Join the Waitlist

Be first to access HCTP Cloud & Pro when they launch.

GitHub Repo PyPI Package Demo Notebook