LearnPrompt vs Sakana AI

The two product surfaces don't overlap at all. LearnPrompt is a knowledge artifact: 47 tutorials in 8 paths (AI coding basics, Claude Code, Codex, Agent Engineering, Agent Skills, Loop Engineering, Obsidian AI, Hermes/OpenClaw long-running agents). Its unit is the task card — input, output, failure state, acceptance criteria — plus writing guides for CLAUDE.md, AGENTS.md and SKILL.md, a Harness five-component troubleshooting method, and Loop Engineering for carrying state and verification into the next run. It teaches you to run tools you already pay for elsewhere; it does not execute anything. Sakana AI ships runnable systems. Fugu orchestrates frontier models across multi-step tasks, and Fugu-Ultra v1.1 adds a Claude Code interface; a Fugu conductor model trained on Gemma 4 matched performance in validation, deliberately reducing dependence on one upstream provider. Namazu is a Japanese-specialised LLM, standalone API since August 2026, also powering Sakana Chat (alongside Fugu) and Sakana Translate's new-generation translation model. Marlin generates analyst research reports on demand; Fugu-Cyber targets cybersecurity workloads; the Recursive Self-Improvement Lab does autonomous AI research. For a Chinese-speaking developer, LearnPrompt's practical overlap with Sakana is essentially zero — the only shared name is Claude Code, one as a topic, the other as an interface. LearnPrompt assumes no data-residency obligation; Sakana's entire pitch is data residency and export-control compliance inside Japan, with domain work at SMBC Group, MUFG and Daiwa Securities.

LearnPrompt is free, permanently, and open source on GitHub — there is no tier, no trial clock, and contributing tutorials or agent workflows via email or WeChat is the only 'purchase' available. Your real cost is the hours you spend, offset by not needing video courses, certificates, or a paid cohort. Sakana AI publishes nothing: pricing_type is contact, so every deal is a custom enterprise agreement scoped to seats, workloads and compliance needs. That structure is deliberate and the company states the trade-off itself — it is not for companies that need self-serve signup and published prices before talking to sales, and not for buyers whose procurement can't absorb a custom contract. Practically, that gates Sakana to organizations with legal and security review capacity: regulated Japanese finance, defense and intelligence. Recent public-sector traction (an August 2026 defense ministry contract for AI survey and demonstration work, on top of existing SMBC Group, MUFG and Daiwa relationships) signals the buying motion is relationship- and compliance-led, not self-serve. If you are comparing budgets, the honest framing is that LearnPrompt costs time and Sakana costs a procurement cycle — there is no price point at which a solo developer would weigh Sakana against a free wiki, and no enterprise would run its data-residency strategy off a tutorial site.