OnlyIdeas keeps a research paper, its equations, figures and passage-level questions in one reading space. Start with the original, then ask for help where you need it.
A Thought Should Be Easier to Save Than to Forget
AiMemo starts with a sentence, then helps turn it into an editable note, task or event. Here is a small way to try it without reorganizing your whole life.
The App Demo Is Recorded. Now Come the Subtitles.
How I use LazyEdit to turn my L & N and Bunko recordings into checked multilingual captions and a reviewed publishing handoff.
Reading Chinese and Japanese Classics Without Constantly Switching Tabs
Bunko keeps the original passage, pinyin or furigana, and parallel study text on one page. A small reading routine for Chinese and Japanese classics.
I Mix Up L and N, So I Built a Practice App
Building a small English, Mandarin and Cantonese practice loop: explainable pronunciation feedback, short listening rounds, and the audio details that make them work.
Why Your Chinese Document Becomes One Giant RAG Chunk
A small regression test for Chinese and Japanese document ingestion: catch whitespace-based chunking, add a code-point cap, and preserve source offsets without confusing characters with embedding tokens.
Why MCP Client Credentials Fail When OAuth Discovery Is Missing
Separate the MCP resource, authorization-server issuer, and token endpoint when a machine client has valid credentials but discovery never reaches the token request.
Fix GitHub MCP ‘Requires Authentication’ in Docker Sandboxes
Separate the Docker sandbox secret, Copilot CLI login storage, the built-in GitHub MCP server, and workspace overrides when authentication fails or disappears after exit.
Claude MCP Authentication: Public, Local, or Remote?
Choose MCP authentication from the deployment boundary: a deliberately public demo, a local developer server, or an OAuth-protected remote service for Claude.
Ten Checks Before You Deploy an MCP Server
A practical pre-deployment review for an MCP server: pin the protocol era and code revision, compare advertised and live capabilities, exercise valid and invalid calls, map real authority, and finish with an evidence-backed go/no-go decision.