GitHub Trending Daily — 2026-05-16: 런타임, 스킬, 음성, 생성형 도구가 함께 뜬 날

Daily snapshot of GitHub Trending, captured at KST 2026-05-16.

Snapshot

GitHub Trending today is less about one breakout model and more about a broader stack: runtime, reusable skills, on-device speech, perception, local code knowledge, and self-hosted generative media.

★ today	Repo	Total ★	Signal
+1,601	tinyhumansai/openhuman	10,647	Personal AI runtime keeps the consumer “AI OS” angle alive
+1,281	obra/superpowers	193,958	Skills methodology stays one of the clearest agent abstractions
+990	ruvnet/RuView	58,308	WiFi sensing and spatial intelligence pull perception into the stack
+745	supertone-inc/supertonic	6,826	On-device multilingual TTS shows speech is a first-class layer
+669	K-Dense-AI/scientific-agent-skills	23,102	Domain-specific agent skills are now normal infrastructure
+414	oven-sh/bun	91,170	Fast runtime/tooling still attracts broad developer attention
+397	colbymchenry/codegraph	2,483	Local code knowledge graphs are now part of the Claude Code conversation
+393	Anil-matcha/Open-Generative-AI	14,399	Self-hosted image/video generation widens the trending surface beyond agents

What changed today

1. Runtime and skills are still the front door. openhuman and superpowers remain the clearest pair in the chart, with scientific-agent-skills reinforcing the idea that skills are now a deployable artifact.
2. Speech and perception are now part of the same stack. supertone-inc/supertonic pulls voice into the foreground, while RuView keeps a non-text perception layer on the board.
3. The chart widened beyond agents. Open-Generative-AI is the most visible new outlier: a self-hosted image/video generation studio, not an agent tool, but still part of the same practical workflow ecosystem.
4. Local code knowledge is becoming a product surface. codegraph is a strong signal that pre-indexed, local-first knowledge graphs for Claude Code are a real demand signal, not just an experiment.
5. Runtime tooling remains sticky. bun is the reminder that low-level runtime speed still attracts broad attention even when the rest of the chart is AI-heavy.

Why this matters

The chart is broadening. A few days ago the story was mostly about memory and browser control; today the signal is more layered:

• Can the system run as a personal runtime?
• Can it reuse skills across tasks and domains?
• Can it speak, sense, and generate media?
• Can it keep local code knowledge close to the workflow?
• Can it plug into the systems that already exist?

That is a more durable market signal than a single breakout app. It implies the center of gravity has moved from “model demo” to “operational stack” plus a few adjacent capabilities that make the stack useful in practice.

Compared with yesterday

Yesterday’s snapshot leaned harder on memory and browser control. Today keeps the agent theme, but the balance shifts toward runtime, skills, speech, perception, local knowledge graphs, and self-hosted generation.

• openhuman and superpowers stay at the core.
• scientific-agent-skills reinforces the skills layer.
• supertonic and RuView add modality breadth.
• codegraph adds local code-knowledge structure.
• Open-Generative-AI widens the chart into creative generation.
• bun shows the surrounding runtime stack still matters.

See also: 2026-05-15-github-trending-daily and moc-ai-agents.

Related pages

• moc-ai-agents
• moc-ai-agents-orchestration
• moc-ai-agents-memory
• 2026-05-11-openhuman
• 2026-05-15-ruview

Raw source

• raw/web-2026-05-16-github-trending-daily.md