dojo.md

2026-02-01 ~ 2026-03-13 · 40 days

Hyper-Ambitious Scoping

Expanded its 'University for AI agents' to 125 courses in 40 days, collapsing under the weight of its own curriculum.

“A university for AI agents, briefly open for enrollment.”

Death Type

Curriculum Creep

This project succumbed to an insatiable 'curriculum creep.' Its 'course library' rapidly swelled from '92 to 125 courses' in under a week, each with '4400+ to 6250+ scenarios.' It attempted to build a full 'University for AI agents' before it could even graduate its first class, leading to a swift demise from over-ambition.

Cause of Death

1. Course Catalog Bloat

The 'course library' expanded from '92 courses, 4400+ scenarios' at inception to '125 courses (6,250+ scenarios)' within the first week of March 2026, a 35% increase in curriculum in mere days. The ambition outpaced the developer's ability to teach.

2. Sprint to Oblivion

The project sprinted through an 'extremely short and intense lifespan of 40 days,' from 2026-02-27 to 2026-03-13. A 'solo contributor' maintained a 'high commit frequency of 1.27 commits/day,' suggesting a pace unsustainable for long-term learning.

3. LLM Overload

With '2 LLM SDKs (@anthropic-ai/sdk, openai)' and 'LLM-judged assertions' for evaluation, the AI agents had more evaluation criteria than human oversight. The 'ModelClient' interface promised extensibility, but the developer eventually needed a break from all the intelligence.

Vibe Score

22/ 100

AI-assisted but human-driven

What They Did

Conceived as a 'University for AI agents,' `dojo.md` promised a training arena for LLMs, generating 'portable SKILL.md documents.' It aimed to support 'any model via OpenRouter' and integrated 'Anthropic and OpenAI SDKs,' all within a 'robust engine' featuring 'mock services and LLM-judged assertions.' The README likely promised enlightenment; the first commit delivered ambition.

@anthropic-ai/sdkopenaivitestzodcommanderNode.jsViteVercelGitHub Actions

Burnout Analysis

Developed by a 'solo contributor, Eduard Cristea,' the project saw '51+ commits' over '40 days,' averaging '1.27 commits/day.' This 'high commit frequency' included '3 feature additions,' '3 bug fixes,' and '3 documentation updates' among the last 10 commits, indicating a relentless, unsegmented effort. The developer likely experienced rapid mental exhaustion from teaching '125 courses' to invisible AI students.

Dependency Archaeology

The `package.json` listed '16 dependencies,' including '2 LLM SDKs (@anthropic-ai/sdk, openai)' for 'any model via OpenRouter' integration. 'vitest' was present for testing, 'zod' for schema validation, and 'commander' for the CLI, indicating a meticulously planned but overly robust stack for a project that barely saw its second month. One might say it had more infrastructure for learning than actual learners.

Autopsy: File Structure

├──CLAUDE.mdDetailed '11 core engine files,' suggesting the AI knew the architecture better than any README could convey.

├──courses/docker-container-debugging/scenarios/level-1/dockerfile-build-failures.yamlA '60-line scenario' with 'llm_judge assertions,' demonstrating an AI grading system more complex than most human universities.

├──.github/workflows/ci.ymlA '30-line CI workflow' ensuring 'Node.js 18, 20, 22' compatibility, for code that would soon go untested by users.

├──vercel.jsonA '5-line configuration' for website deployment, a vision of a public frontend that few would ever visit.

├──package.jsonListed '16 dependencies,' including two LLM SDKs, preparing for a future of AI education that arrived too quickly.

├──src/engine/model-client.tsThe core interface for 'ModelClient,' promising an extensible future for LLM integration that ended abruptly.

├──src/types/schemas.tsUtilized 'zod' for 'robust validation of YAML scenarios,' ensuring rigorous type safety for a project that would soon lack a pulse.

└──courses/GENERATION_LOG.mdThe most frequently changed file, with '+878/-0 lines,' recording the relentless expansion of a curriculum no one would complete.

Eulogy Stats

Total Commits: 51
Ambitious Adjectives: 5
Deploy Config: Yes
Estimated Users: 0 (predicted enrollment of 6,250+ AI agents)

Last Words

“The last recorded commit, 'chore: bump version to 0.3.3,' suggests a final, weary attempt at progress before the silence.”