LearnPrompt vs Bito

The two products operate at opposite ends of the same pipeline without overlapping in capability. LearnPrompt is content: 8 paths (AI programming basics, Claude Code, Codex, Agent engineering, Agent Skills, Loop Engineering, Obsidian AI, and Hermes/OpenClaw long-running agents), 47 tutorials each tagged with primary sources and verification dates, and a content unit built around task cards that specify input, output, failure states, and acceptance criteria. It also ships practical writing guides for CLAUDE.md, AGENTS.md, and SKILL.md; a five-component Harness troubleshooting method for locating which layer an agent is missing; a side-by-side comparison of Claude Code and Codex execution surfaces (CLI/IDE/desktop/Cloud); and replayable Showcases that require both the happy path and out-of-bounds counterexamples. Everything is Markdown, open source on GitHub, and integrated conceptually with Claude Code, Codex, Obsidian, ChatGPT, Hermes, OpenClaw, and GitHub. Bito is infrastructure: Governor routes each request to a right-sized model using complexity scoring benchmarked against a live knowledge graph of your services, dependency depth, and blast radius, serves relevant files/symbols/dependencies as context per request, and installs via a single base-URL swap on Anthropic and OpenAI APIs with bring-your-own-keys or gateway passthrough. It adds quality floors and route pinning per key, per-team or per-key budgets with token and spend analytics in one admin view, rolling on/off savings measurement against your own traffic, and AI Architect for feasibility checks, technical design in Jira/Linear, and cross-repo impact analysis. One teaches judgment; the other enforces policy and cuts retrieval waste.

LearnPrompt is free with no paid tier at all — the entire wiki, all 47 tutorials, task cards, and templates sit in an open-source GitHub repo maintained by Carl, with email and WeChat channels for contributing. The real cost is your time: engineering-heavy, Chinese-language, no video, no certificate, and it assumes you've already run a terminal and Git at least once. Bito is freemium, but the freemium distinction matters less here than in most comparisons: Bito does not publish usage rates for Governor or AI Architect, and its own positioning lists 'buyers who need published usage rates without a sales conversation' as a not-for-us segment. So you can't price-compare these two on a table. Bito's commercial pitch is outcome-based instead: its own analysis claims 78% of AI coding spend goes to agents hunting for code rather than generating it, and it reports a customer A/B where cost per task dropped 48% with task success holding at 100%. Rolling on/off measurement against your own traffic exists precisely so you can verify that before committing to a contract. If agent spend is a rounding error for you, Bito's setup — indexing and scoping a large codebase — will not pay back. If you're routing multi-repo agent traffic across teams, the free wiki is irrelevant to your budget and the router is not.