GitHub's daily trending list today is a snapshot of where developer energy is flowing in mid-2026: practical tooling that reduces friction at the model-input layer, AI video automation, and infrastructure for agents that need to remember things across sessions. The top repos by star velocity are all Python or TypeScript, reflecting where the ecosystem is building most actively right now.
microsoft/markitdown: Converting Everything to Markdown, at Scale
Microsoft's markitdown picked up nearly 2,800 stars today on its way to over 136,000 total, making it one of the most-starred Python tools of the year. It converts virtually any file format — PDFs, Word documents, Excel spreadsheets, PowerPoints, images — into clean Markdown, which makes it a natural pre-processing layer for RAG pipelines and LLM context windows. For teams building document-heavy AI workflows, it consolidates what used to be a dozen one-off parsing scripts into a single dependency.
github.com/microsoft/markitdown
harry0703/MoneyPrinterTurbo: One-Click AI Short Video
MoneyPrinterTurbo added nearly 2,000 stars today, pushing its total past 76,000. The project automates the entire short-video production pipeline using large language models: script writing, voiceover synthesis, asset sourcing, and final video assembly all flow from a single prompt. It supports both Chinese and English, and the sustained star velocity suggests the demand for accessible, model-powered video tooling is far from saturated.
github.com/harry0703/MoneyPrinterTurbo
supermemoryai/supermemory: Persistent Memory as an API
Supermemory is a fast, scalable memory engine built as a drop-in API layer for AI applications that need state across sessions. It gained 264 stars today toward a total of over 10,000. The core insight is simple: stateless LLMs can't behave as persistent assistants without an external memory layer, and supermemory provides that with an emphasis on retrieval speed and clean developer ergonomics. As agent frameworks multiply, durable memory is shaping up to be one of the foundational infrastructure pieces that every production AI app will eventually need.
github.com/supermemoryai/supermemory
EveryInc/compound-engineering-plugin: One Plugin for Every AI Coding Tool
The compound-engineering-plugin gained 251 stars today as a TypeScript package that works across Claude Code, OpenAI Codex, Cursor, and other AI coding environments. It brings a consistent interface — structured workflows, code review tooling, and planning utilities — to whichever AI coding assistant a developer happens to use. As teams increasingly mix models and editors, cross-platform plugins that abstract over the differences are starting to matter.
github.com/EveryInc/compound-engineering-plugin
Crosstalk-Solutions/project-nomad: AI That Works Offline
Project Nomad picked up 374 stars today for a premise that runs against the cloud-first grain: a self-contained, offline survival computer packed with AI tools, knowledge bases, and utilities that require no internet access. Built in TypeScript, the project targets genuinely low-connectivity environments — field work, emergencies, remote deployments — where today's cloud-dependent AI assistants are useless. It's early-stage, but the direction highlights a real gap in the current AI landscape: almost everything useful requires a reliable connection.