GitHub Developers

GitHub Trending Developers: Who to Follow This Week and What to Learn From Them

Direct Answer

The most useful people on GitHub Trending this week are not all building the same kind of product. Luis Novo is assembling a set of composable open AI products around Open Notebook. Cole Murray combines background coding agents with observability. Daniel Oster turns years of battery and vehicle knowledge into working open hardware infrastructure. Charlie Marsh shows what happens when performance work improves an entire developer loop instead of one benchmark.

The daily movers add another layer: Elie Steinbock turns an open-source application into a product and teaching platform; misaki masa makes a focused terminal tool feel like a polished product; Marco Cadetg makes network observability usable; and Matt Van Horn demonstrates how a developer can contribute across many agent ecosystems while maintaining a recognizable research workflow.

JQ AI SYSTEMS take: Follow builders for their working patterns, not their rank. Pick three profiles: one close to your stack, one adjacent to your craft, and one outside your normal lane. Read the code, issues, releases, and maintenance history before adopting a tool.

Source Note

This article uses live snapshots of GitHub Trending Developers weekly and GitHub Trending Developers daily, checked on 15 July 2026. The order changed while the article was being researched. That volatility is why this is a dated people radar, not a permanent leaderboard.

I merged names that appeared in both views and checked the shortlisted developers against their GitHub profiles, highlighted repositories, current READMEs, official websites, release histories, and public creator links. A place on Trending does not endorse every repository, business model, dependency, or security decision associated with a developer.

SignalDeveloperHighlighted workWhy followUseful links
Weekly + dailyLuis NovoOpen NotebookComposable, self-hosted AI products with provider choice and local-model support.Website / X
Weekly + dailyCole Murraybackground-agentsBackground execution, agent observability, and practical cloud integration.Website / X
WeeklybinaricatNetcattyA cross-platform SSH workspace that treats terminals, files, credentials, and AI as one operator surface.X
Weekly + dailyAlan Buscagliagentleman-book-mcpTurns architecture teaching and team standards into context an agent can query.Book / Profile
Weekly + dailyDaniel OsterBattery EmulatorDeep domain expertise translated into maintained hardware and energy infrastructure.Official setup video
WeeklyEmre SokulluWebBrainAn open browser-agent layer with a long history of product and platform experimentation.Website / X
WeeklyCharlie Marshuv / RuffPerformance engineering that changes package, lint, format, and development workflows.Website / X
WeeklyAddy Osmaniagent-skillsProduction-minded patterns for reusable agent instructions and frontend engineering.Website / X
Weekly + dailyShubham Sabooawesome-llm-appsA broad library of runnable agent, RAG, voice, memory, and multi-agent examples.X
WeeklyHenrik RydgardPPSSPPLong-running cross-platform C++ maintenance, compatibility work, and community stewardship.Website
WeeklyMichael WaskomseabornA reminder that durable scientific visualization and documentation still matter beside AI launches.Documentation
DailyElie SteinbockInbox ZeroOpen-source product building, public teaching, and repeatable automation loops.Product / X
Dailymisaki masaYaziA focused, high-performance Rust terminal utility with strong interaction design.Website / X
DailyMarco CadetgrustnetPer-process network monitoring packaged as a usable cross-platform terminal product.Website
DailyYiwei Hoopen-slidePresentation infrastructure designed to be readable and controllable by agents.Website / X
DailyKane WangTakumiDeveloper-facing media infrastructure that renders JSX, HTML, and CSS into SVG or images.Profile
DailyMatt Van HornLast30DaysCurrent-signal research skills and a visible pattern of contributing across agent ecosystems.X / JQ guide

What GitHub Trending Developers Actually Measures

GitHub does not present Trending Developers as a quality score, hiring ranking, security audit, or lifetime influence list. It is a discovery page tied to recent repository activity and attention. The daily view is useful for spotting a launch early. The weekly view is better for seeing whether attention persists beyond one announcement. Neither tells you whether a project is maintained well, licensed for your use, safe with credentials, or a fit for production.

The most useful signal is therefore not position. It is the relationship between a developer's highlighted repository and the rest of their work. A single repository can trend because of a social spike. A profile that repeatedly shows coherent projects, releases, issue handling, documentation, and domain depth is a stronger learning source.

Read the list as a map: repositories show what shipped; profiles show the habits, adjacent experiments, collaborators, and maintenance choices behind the release.

Weekly Developers Worth Following

Luis Novo: build an ecosystem, not an isolated demo

Luis Novo is the clearest example of a portfolio that compounds. Open Notebook is a self-hosted research and notebook application with multiple model providers, local-model options through tools such as Ollama and LM Studio, content ingestion, search, API access, and podcast generation. His wider profile includes projects for content processing, prompting, and podcast creation.

The lesson is architectural. Each project can be useful alone, but the combined set points toward reusable primitives for research, content, and generation. Builders should study the boundaries between those projects, the provider abstraction, and how a visible product can sit above reusable infrastructure.

Cole Murray: background agents need observability

Cole Murray is highlighted for background-agents, but the adjacent Claude Code OpenTelemetry project is just as instructive. Long-running coding agents are not only a prompting problem. They need queues, isolated execution, logs, traces, cost visibility, failure handling, and a clear handoff back to a person.

Follow Cole if you are moving from one interactive coding session to an internal agent platform. The durable pattern is the pairing of autonomy with evidence.

Daniel Oster: domain knowledge is still the moat

Daniel Oster develops Battery Emulator, a translation layer that helps compatible electric-vehicle batteries communicate with stationary energy inverters. This is not a weekend wrapper. It combines embedded software, vehicle protocols, inverter compatibility, testing, documentation, releases, and a real safety envelope.

The project also makes the limit explicit: high-voltage systems can injure or kill. Installation must follow local rules and qualified electrical practice. The lesson for AI builders is useful precisely because it is not an AI lesson. Deep domain knowledge, a real verification loop, and responsibility for physical consequences remain hard to commoditize.

Alan Buscaglia: make expertise agent-readable

Alan Buscaglia turns his software-architecture material into gentleman-book-mcp. The important pattern is broader than this book: a team can expose its architecture principles, examples, vocabulary, and decision rules through an interface an agent can query while it works.

This is more reliable than pasting a giant style guide into every prompt. It also creates a maintenance obligation. Teams should version the knowledge, show citations, mark deprecated advice, and review whether the agent is retrieving the relevant chapter rather than merely producing confident architecture language.

binaricat and Emre Sokullu: agents need useful operating surfaces

Netcatty combines SSH, SFTP, local and remote terminals, snippets, credential management, and built-in AI in one cross-platform workspace. WebBrain approaches the browser as an agent surface. Both point toward the same shift: useful agents need durable interfaces to computers, files, websites, and sessions.

These are powerful surfaces, so they deserve a stricter review than a static library. Inspect credential storage, browser or SSH permissions, telemetry, update channels, and what an agent can execute without approval.

Charlie Marsh, Henrik Rydgard, and Michael Waskom: durable craft still wins

Charlie Marsh is worth following for the way projects such as uv and Ruff improve an entire Python workflow. Henrik Rydgard's PPSSPP demonstrates years of cross-platform C++ compatibility and community maintenance. Michael Waskom's seaborn shows how scientific visualization, documentation, and stable APIs can remain valuable through multiple technology cycles.

Trending is often described as a launch feed. These profiles are a reminder that engineering reputation also compounds through sustained maintenance, careful interfaces, and tools that other developers trust.

Addy Osmani and Shubham Saboo: curate patterns people can run

Addy Osmani's agent-skills and Shubham Saboo's awesome-llm-apps package knowledge differently. One focuses on reusable agent instructions; the other collects runnable application patterns across agents, retrieval, voice, memory, and multi-agent systems.

Curation is useful when it reduces search and makes evaluation easier. Readers should still inspect each skill or example independently. A large collection can contain different licenses, security assumptions, dependencies, model costs, and levels of maintenance.

Daily Movers Worth Watching

Elie Steinbock: connect open source, product, and education

Inbox Zero is a useful study in building an open-source product around a painful, recurring workflow. Elie Steinbock also explains the operating ideas around his work publicly, including repeatable business loops. Follow the connection between repository, product, user problem, and teaching rather than treating the code as an isolated artifact.

misaki masa and Marco Cadetg: terminal tools can feel like products

Yazi is a high-performance terminal file manager written in Rust. rustnet turns per-process network activity into a terminal monitoring experience. Their shared lesson is focus: performance matters, but discoverability, defaults, interaction design, cross-platform behavior, and clear documentation are what make a utility usable.

Yiwei Ho and Kane Wang: give agents structured creative infrastructure

open-slide explores presentation creation in a form that agents can manipulate. Takumi renders JSX, HTML, and CSS into SVG or raster images. Both are more interesting as infrastructure than as one-shot generators: they give agents a deterministic representation, a rendering path, and an output that can be tested.

Matt Van Horn: contribution patterns are part of the profile

Matt Van Horn is associated with Last30Days, but his profile is also useful for its cross-project activity. He contributes fixes and integrations across the agent tools he uses. That is an underrated open-source pattern: the strongest users do not only request compatibility; they help create it.

Video Walkthroughs

These two videos show opposite ends of the developer radar. Daniel Oster's official guide deals with physical hardware, safety, and a maintained installation process. Matt Van Horn's interview shows a software skill that diagnoses source access and synthesizes current public discussion for agents.

Official Battery Emulator installation walkthrough from Dala's EV Repair. High-voltage work requires appropriate qualifications and local compliance.

Andrew Warner interviews Matt Van Horn about recent-source research, Doctor mode, and using Last30Days across agent workflows. See the JQ AI SYSTEMS setup and verification guide for a source-checked operator view.

Six Patterns Worth Learning

  1. Build composable products. Luis Novo's projects show how a visible application can share ingestion, model, content, and generation layers with other tools.
  2. Pair autonomy with observability. Cole Murray's agent and telemetry work belongs together. A long-running agent without traces is difficult to trust or improve.
  3. Turn expertise into interfaces. Alan Buscaglia exposes architecture knowledge through MCP; Daniel Oster encodes battery and inverter knowledge into tested software.
  4. Treat local-first as a product choice. Open Notebook, Inbox Zero, Yazi, rustnet, and other tools show different forms of local control, self-hosting, or inspectable operation.
  5. Optimize the whole loop. Charlie Marsh's work matters because package installation, linting, formatting, and execution compound across every development cycle.
  6. Maintain beyond launch day. PPSSPP and seaborn demonstrate that compatibility, documentation, stable interfaces, and community support create value for years.

Who I Would Follow First by Role

If you are...Follow firstStudy this
Building agent platformsCole Murray, Emre Sokullu, binaricatBackground execution, observability, browser control, SSH workspaces, and permission boundaries.
Building local or self-hosted AILuis Novo, Elie SteinbockProvider abstraction, data ownership, product packaging, deployment, and user-facing setup.
Improving developer experienceCharlie Marsh, misaki masa, Marco CadetgPerformance, terminal interaction, defaults, cross-platform behavior, and clear feedback.
Creating reusable skills or learning systemsAddy Osmani, Alan Buscaglia, Shubham Saboo, Matt Van HornInstruction design, retrieval, examples, source freshness, diagnostics, and maintenance.
Working near hardware or physical systemsDaniel OsterDomain protocols, compatibility matrices, testing, release discipline, and safety documentation.
Maintaining an open-source communityHenrik Rydgard, Michael WaskomLong-term stewardship, documentation, compatibility, contributor pathways, and stable interfaces.
Building creative infrastructureYiwei Ho, Kane WangStructured representations, deterministic rendering, agent control, and testable outputs.

One Cautious Watch: codex-lb

codex-lb appeared prominently in both snapshots. It presents a local load-balancing and proxy layer for multiple Codex or ChatGPT accounts, with tracking and a dashboard. That can be technically interesting, but it is not where I would begin.

A proxy of this kind can sit close to authentication tokens, account limits, model traffic, and provider rules. Before testing, read the source that stores credentials, check whether dashboards bind beyond localhost, understand account-sharing assumptions, rotate test tokens, and review the current terms for every connected provider. Inclusion here is a research signal, not an endorsement of bypassing limits or sharing subscriptions.

How to Evaluate a Trending Developer

  1. Open the profile, not only the trending repo. Look for coherent projects, contribution history, release work, and how the developer responds to issues.
  2. Check maintenance evidence. Read recent commits, release notes, open bugs, stale pull requests, and whether the default branch matches the documentation.
  3. Read the license and install path. Do not assume a public repository is automatically suitable for commercial use or safe to install.
  4. Map the permissions. List every key, account, browser profile, SSH credential, local file, cloud service, and network endpoint the tool can reach.
  5. Separate project claims from measured results. Stars, speed claims, benchmark screenshots, and provider counts need verification in your own environment.
  6. Run the smallest useful test. Use a disposable workspace, synthetic data, test credentials, and a clear acceptance check.
  7. Learn the pattern even if you skip the tool. The architecture, interface, maintenance practice, or documentation method may be more valuable than another dependency.
  8. Follow selectively. Subscribe to releases or a developer's public updates only when their work maps to a skill you are actively building.
CTA: Choose three developers from this map. Read one release, one difficult issue, and one adjacent repository from each. Write down the pattern you can reuse before you install anything.

Bottom Line

The weekly developer list is strongest when it reveals durable engineering habits. Luis Novo builds an ecosystem of composable AI products. Cole Murray treats background agents as observable systems. Daniel Oster applies deep domain knowledge to real hardware. Charlie Marsh optimizes complete developer loops. Alan Buscaglia turns teaching into agent-readable infrastructure. Henrik Rydgard and Michael Waskom show the value of long maintenance horizons.

The daily list adds faster-moving signals: local-first products, Rust terminal tools, agent-controlled presentation and rendering infrastructure, and cross-project open-source contribution. These are useful people to study because their work exposes decisions, not because GitHub placed them in a particular order for a few hours.

Use Trending to discover. Use profiles to understand. Use repositories and issues to verify. Then bring one proven pattern into your own work.

Sources

Common questions

Who are the best GitHub Trending developers to follow this week?
For the 15 July 2026 snapshot, Luis Novo and Cole Murray are strong follows for open AI products and background-agent infrastructure. Daniel Oster is valuable for deep hardware-domain engineering, Charlie Marsh for performance-focused developer tooling, Addy Osmani for reusable agent skills, and Matt Van Horn for current-signal research workflows. The best follow depends on the craft you want to learn.
What is the difference between weekly and daily GitHub Trending Developers?
The daily list reacts quickly to releases, social attention, and short bursts of activity. The weekly list is a slightly stronger momentum signal. Both change continuously, so this article records the date, merges repeated names, and treats position as discovery rather than a permanent ranking.
Does a trending developer mean their repositories are safe to install?
No. Trending is not a security review. Inspect the repository owner, license, releases, install scripts, secrets, telemetry, open issues, permissions, and provider terms. First runs should use a disposable workspace and test credentials, especially for proxies, browser agents, SSH tools, and autonomous coding systems.
How should I choose which open-source developers to follow?
Choose one person close to your stack, one adjacent to your craft, and one outside your normal lane. Follow their releases, issues, design decisions, and maintenance habits. A useful follow should teach you a repeatable engineering pattern, not only point you toward another tool.
Why do some developers appear in both the June and July roundups?
Repeated appearances are useful because they suggest continuing attention rather than a one-day spike. Cole Murray, Luis Novo, Matt Van Horn, and Daniel Oster appeared in earlier coverage and remain relevant in this new dated snapshot, while the repositories and surrounding work have continued to evolve.
Is codex-lb recommended for sharing multiple AI accounts?
It is included as a cautious watch, not a first-install recommendation. A local load balancer or proxy can touch authentication tokens, account limits, network traffic, and provider terms. Review its token storage, dashboard exposure, account-sharing assumptions, and the terms of every connected service before testing it.
Share
X LinkedIn Reddit
Build Yours

Want a system
like this one?

Book a free 30-minute call. We map your situation, identify the highest-impact automation, and figure out if we are a fit.

Book Free 30-min Call