Andrew Warner and Adam's GitHub roundup moves from media tools to agent infrastructure, company-data mapping and robotics. Its three bonus projects, ten main picks and four community submissions are worth exploring, but they solve very different jobs. This guide keeps the episode list separate from JQ AI SYSTEMS' 14 November 2026 snapshot of GitHub's weekly repository and developer rankings.
Need a personal agent you can inspect? Test OpenMausBot with a disposable account and narrow permissions. Need to study editing, not reproduce someone else's protected footage? Use REA Video for a shot-by-shot production brief. Need a coding workflow? Start with one skill from Matt Pocock or Addy Osmani. Stars show attention, not suitability or safety.
Watch the Full GitHub Roundup
The show-notes page has the original project links and related demos. The episode is presented by Zapier and includes a Hyperagent sponsor segment near 09:49. Neither the sponsor claims nor the creator's short tests are an independent security audit.
Three Bonus Projects: Media and Team Chat
| Project | What it does | Important limit |
|---|---|---|
| OpenGenjutsu | Recasts a short source clip with a supplied character image, then restores source audio. | Local interface, paid external generation API. Use only footage and likenesses you can process and publish. |
| REA Video | Breaks a reference video into shots, visual methods, edit rhythm, audio and a reusable VIDEO.md plan. | A production analysis, not proof that reference media is licensed for reuse. |
| Ember Cloudflare | A self-hosted, Slack/Campfire-style team chat designed for Cloudflare. | Check reliability, retention, integrations and migration before replacing team chat. |
OpenGenjutsu's README requires Python, FFmpeg, an API key and generation credits. It binds the local interface to loopback because the app is not authenticated for public hosting. "Local-first" describes the interface, not the entire processing path: uploaded media is subject to the API provider's terms. The video shows Andrew trying the tool with his own likeness; reuse of another person's face, voice, clip or music needs permission.
REA Video classifies shots and creates an editable breakdown of filming, animation, overlays, subtitles, sound and timing. That is useful for learning a format or briefing an original production. It does not automatically grant rights to remake a video shot-for-shot. Ember Cloudflare is a different sort of tool: a chat alternative. Evaluate history migration, uptime and access controls before moving a real team.
The Episode's Top 10 Repositories
This order follows Andrew and Adam's show, not the changing GitHub Trending list below. Each row links to the maintainer's repository rather than a truncated video-description URL.
| Episode pick | Useful job | What to check |
|---|---|---|
| 1. Sherlock | Looks for a username across supported social platforms. | A matching username does not prove the same person owns every account. Use for authorized research and privacy checks. |
| 2. Anthropic Skills | Public examples of reusable agent tasks and supporting files. | Inspect each skill's license, scripts and tool permissions; a public repo is not a blanket safety approval. |
| 3. OpenMausBot | Independent personal-agent app with per-bot computer options and model/provider choice. | Test with a disposable account and dedicated machine or VM; never give broad credentials on day one. |
| 4. Agent Browser | Browser-automation CLI built for coding agents. | Use scoped sessions and verify page state before any external or account-changing action. |
| 5. Daily Stock Analysis | Combines market data, news and LLM analysis into reports and notifications. | Research aid, not investment advice. Verify data dates, assumptions and sources before acting. |
| 6. NVIDIA AutoOntology | Builds a semantic, graph-based reasoning layer over organization data. | Start with a small schema and check generated relationships against source records. |
| 7. MuJoCo | Physics simulator for robotics and other contact-rich dynamics. | Simulation is not physical validation; measure sim-to-real gaps and safety constraints. |
| 8. Matt Pocock Skills | Practical coding-agent workflows for specs, implementation, PRs and retros. | Adopt one skill at a time and measure whether review quality improves. |
| 9. Agent Skills | Engineering skills and process templates for coding agents. | Read the full instruction file before installing it into a privileged agent. |
| 10. Cloudflare Security Audit Skill | Source-first defensive security review with structured findings. | A security skill is guidance, not a completed audit; execute untrusted code only in an appropriate sandbox. |
Three distinctions matter here. Sherlock searches usernames, not legal identities; similar handles can belong to different people. AutoOntology builds a semantic layer so agents can reason over data relationships, not magically make messy records correct. And Cloudflare's security skill documents a source-first audit method, but installing a skill does not itself validate your code. Its instructions distinguish limited guidance from a full, explicitly requested audit.
Andrew's stock-analysis demo produced a bearish NVIDIA view. That output is a model-generated research artifact, not a verified recommendation. Check data timestamps and the cited evidence before using a tool like this for any financial decision.
First Stars: Four Community Projects
The community segment is a useful reminder that a small repository may solve a sharper problem than a viral one. These descriptions come from the linked project pages, not a claim that we tested the apps.
| Project | Why it is interesting |
|---|---|
| Atlas | Links agent sessions to their code changes and commits, giving teams a trail to review. |
| Bomby | Mac voice interface for several coding-agent CLIs, with completion notifications. |
| Sprint Wake Projects | Native macOS client for GitHub Projects with optional Slack context. |
| Rightsize | Claude Code plugin that selects model and effort per prompt to manage usage. |
In the final section, Adam demonstrates creating a small game through a voice conversation, while Andrew shows a custom video-editing feedback tool built with Codex. These are personal workflows, not part of the ten-repo ranking. The recurring lesson is to make the agent's output reviewable: visible game behavior, a linked commit or time-coded editor feedback beats a vague "done" message.
JQ AI SYSTEMS: 10 Repositories Trending This Week
Snapshot: 14 November 2026, approximately 03:30 UTC. These were the first ten entries on GitHub's live weekly repository page. This is a dated attention snapshot selected for JQ AI SYSTEMS readers, not a security or quality ranking. Its fifth entry, Google's Elephant Goldfish Model, illustrates why the live list should not be silently substituted for the episode's list.
| Weekly rank | What it offers / first check |
|---|---|
| 1. Sherlock | Username discovery; treat matches as leads to verify, not identity proof. |
| 2. Anthropic Skills | Reusable agent-skill examples; inspect each skill before importing. |
| 3. OpenMausBot | Personal agent with computers; check execution and credential boundaries. |
| 4. Agent Browser | Agent browser CLI; test scoped navigation on a non-sensitive site. |
| 5. Elephant Goldfish Model | Google plugin for designing and reviewing software with agents. |
| 6. Daily Stock Analysis | Market research workflow; scrutinize sources and risk language. |
| 7. AutoOntology | Semantic data layer; validate the ontology on known records. |
| 8. MuJoCo | Robotics physics simulator; test models against physical constraints. |
| 9. Matt Pocock Skills | Agent engineering workflows; try one repeatable task. |
| 10. Agent Skills | Agent process templates; check the authority each instruction requests. |
JQ AI SYSTEMS: Ten Trending Developers
These were the first ten profiles on GitHub's weekly developer page at the same check. The linked popular repository gives context. GitHub's ordering is dynamic and is not a ranking of each person's skill or trustworthiness.
| Rank and developer | Popular project | Focus |
|---|---|---|
| 1. l0ng-ai | openidea | Rust IDE experiment |
| 2. deepcoldy | botmux | Bridge from Lark/Feishu chats to coding CLIs |
| 3. ZhuLinsen | daily_stock_analysis | Multi-market stock-analysis workflow |
| 4. derekross | omarchy-mobile | Mobile Omarchy project, work in progress |
| 5. ranxianglei | billion-context | Agent context-compression plugin |
| 6. yetone | magpie | Model switching across coding agents |
| 7. SamuelSchlesinger | lean.vim | Vim tooling for Lean |
| 8. andrewnakas | gtav-recomp | GTA V static recompilation project |
| 9. zts212653 | clowder-ai | Agent-team coordination |
| 10. serenakeyitan | agent-org-chart | Portable agent organization templates |
Original Demos and X Posts
The episode resource page links many more examples. These three are useful companion videos because each shows a different class of tool in action:
The show's resource page also points to Kevin Rose's original Ember post and follow-up. The embedded posts are context, not proof that a new chat system matches Slack's reliability or compliance features.
Ember Cloudflare announcement
Ember follow-up
One Safe Pilot Beats Installing Ten Repos
- Name one bottleneck. For example, an agent loses context between PRs, or your team spends too long making a shot list.
- Read the repository first. Check the README, license, recent changes, issues, install steps and requested permissions.
- Use disposable inputs. Keep client data, credentials, private browser sessions and real media out of the first test.
- Define a measurable result. Compare elapsed time, output accuracy and review effort against your current process.
- Keep a human approval point. Approve publishing, account changes, money movement and outreach yourself.
- Decide keep or remove. A trending tool earns a place in your workflow only if it improves the measured job without creating a larger security or maintenance burden.
Video Chapters
| Time | Topic |
|---|---|
| 00:00 | Viral media tools and weekly overview |
| 00:21 | Three bonus projects |
| 00:27 | OpenGenjutsu |
| 02:10 | REA Video |
| 02:57 | Ember Cloudflare |
| 04:15 | Sherlock |
| 05:43 | Anthropic Skills |
| 07:09 | OpenMausBot |
| 09:49 | Hyperagent sponsor discussion |
| 12:08 | Agent Browser |
| 15:07 | Daily Stock Analysis |
| 17:00 | NVIDIA AutoOntology |
| 18:59 | MCP, CLI and API |
| 22:51 | Zapier and automation costs |
| 23:54 | MuJoCo |
| 26:34 | Matt Pocock skills |
| 28:39 | Addy Osmani skills |
| 31:57 | Cloudflare Security Audit Skill |
| 33:56 | First Stars |
| 36:51 | Voice-built game |
| 38:23 | Andrew's video-editing feedback tool |
Sources and Scope
Checked 14 November 2026 using the supplied episode transcript, the show's source page, linked GitHub repositories and live repository and developer pages. Repository descriptions come from their maintainers; none of the projects was independently security-audited or installed for this article. Trending order can change. The episode includes Zapier and Hyperagent sponsorship. External video and X embeds may depend on browser privacy settings; direct source links remain available.