ConardLi
ConardLi is an individual/solo developer and creator who curates and codes Easy AI (also branded as code秘密花园), an independently maintained site that makes complex AI concepts, models, evaluation benchmarks, tutorials, prompts and daily news easier to understand. The creator also develops open-source AI tools, including Easy Dataset, for building and evaluating LLM datasets.
At a Glance
- AI learners and beginners
- Developers and technical practitioners building LLM applications
- Researchers and teams preparing fine-tuning, RAG and evaluation data
- Organizations needing local/private dataset processing
- +1 more
AI Tools by ConardLi
(1)
Garden Skills
Agent Skills for AI Coding Tools
Discussions
No discussions yet
Be the first to start a discussion about ConardLi
Latest News
Easy Dataset v1.7.3 tagged on GitHub
Easy Dataset added MiniMax as a first-class LLM provider, including MiniMax-M2.5 and M2.5-highspeed
Easy Dataset v1.7.0 introduced the evaluation module with generated test sets, automated scoring and human blind testing
Easy Dataset v1.6.2 added individual-question export, batch file operations, keyword exclusion filters and token statistics
Products & Services
Free, independently curated and coded AI learning site with visual concept explainers, AI application and model directories, evaluation benchmarks, tutorials, prompt analyses, articles and daily briefs.
Open-source local tool for creating and managing LLM datasets for fine-tuning, RAG and evaluation. It supports document parsing, intelligent/hybrid chunking, question and answer synthesis, cleaning, augmentation, review, evaluation and export.
Open-source local terminal Agent intended to replicate the Claude Code experience.
An Agent skill that turns articles, courses, product demonstrations and talks into webpage-simulated videos.
Market Position
ConardLi's work is positioned as an independent, open-source and local-first alternative to commercial or more code-centric synthetic-data tools. The Easy Dataset paper explicitly compares it with Distilabel, Kiln, Curator, Data-Juicer and GraphGen, emphasizing its combination of adaptive document parsing, hybrid chunking, persona-driven synthesis, end-to-end GUI and human-in-the-loop refinement. Easy AI differentiates as a highly curated, visual and beginner-friendly AI learning/reference site rather than a general news feed.
Leadership
Founders
ConardLi
Independent developer, open-source maintainer, AI educator and content creator. GitHub identifies the account as ConardLi, lists Beijing and mmh1.top, and links to the code秘密花园 YouTube and X accounts; the Easy AI site says it is curated and coded by ConardLi.
Founding Story
The creator says Easy AI was made for people who want to learn and was started to make complicated AI concepts, models, evaluations, tools and daily news readable through careful editing, visual explanations and daily updates. Easy Dataset was created to address the difficulty of turning heterogeneous documents into high-quality LLM fine-tuning data, with a local, end-to-end workflow for parsing, chunking, synthesis, review and export.
Business Model
Revenue Model
The core Easy AI reading content is free and the site says it does not collect email or show pop-ups. Monetization appears to be through the linked paid Knowledge Planet community and a low-cost model Token Gateway; Easy Dataset is distributed as open-source software with client downloads, NPM and Docker deployment rather than a documented subscription plan.
Pricing Tiers
Official site labels the content as free and says the free content's only requested return is a GitHub star.
Target Markets
- AI learners and beginners
- Developers and technical practitioners building LLM applications
- Researchers and teams preparing fine-tuning, RAG and evaluation data
- Organizations needing local/private dataset processing
- Creators and developers using Agent skills for content, web design, presentations, translation and image generation
- Learning AI concepts, models, benchmarks, prompting and Agent design
- Turning unstructured domain documents into fine-tuning datasets
- Building RAG knowledge-base data
- Generating instruction, QA, multi-turn, distillation and VQA datasets
- Human review and cleaning of synthetic datasets
- Evaluating multiple LLMs on generated test sets