Repos

Three picks from GitHub Trending’s weekly list, captured October 11, 2026; repositories covered in recent roundups are excluded. These are tools to investigate, not recommendations backed by hands-on testing. Agent tooling tester-army/e2e — End-to-end tests for web and mobile apps combine natural-language agent actions with explicit assertions; the project is still pre-1.0. danyuchn/asd-ste100-skill — Applies Simplified Technical English principles to clarify agent instructions and tool descriptions; its structural linter does not prove that rewrites preserve meaning. Building apps flutter/flutter — Dart-based tooling for building mobile, web, and desktop interfaces from a shared codebase, useful when maintaining separate app stacks would be too much work.

Andrew Warner and Adam on The Next New Thing review this week’s ten repositories in a 24-minute roundup. These are discoveries to investigate, not tools tested for this capture. Games and visual creation AnyPS5 (00:28): Ports PS5 executables to native Linux/Windows formats; not a universal emulator, and decrypted game files remain a prerequisite. HyperFrames (03:31): Renders HTML-based animation into video, making revisions code edits rather than repeated generative-video runs. RAD Debugger (05:08): A native graphical debugger that complements agent-written game code with human inspection; the hosts note its alpha status. Text to CAD (13:00): Turns descriptions of parts into 3D models; the practical attraction is custom printable objects, not just downloaded designs. Agent teams, memory, and workflow OpenRig (02:06): Organizes coding agents into persistent teams with specialist roles, shared context, and coordinated work. Cursor Plugins (06:30): Shares connectors and concrete developer workflows that can be inspected and adapted rather than copied from a demo by hand. Claude Mem (07:55): Captures work across sessions and supplies relevant persistent context to future agent runs. Matt Pocock’s Skills (09:09): Simple reusable workflows, especially “grill me,” which challenges an idea before implementation; avoid turning discovery into endless questioning. Research and owning your data Agent Reach (11:24): Collects structured material from sites such as Reddit and YouTube, with alternative backends when an extractor breaks. OpenGym (14:40): Self-hosted workout tracking focused on owning and controlling your records; the hosts flag the inconvenience of hosting and limited iOS options. What to take away The most useful theme is control: editable video sources, inspectable workflows, persistent context, and portable personal data. Treat the PS5 headline cautiously: the repository describes executable relinking and replacement libraries, not an AI rewrite of arbitrary games. The closing discussion covers agent testing, voice interfaces, policy checks, and subscription proxies; claimed protections against account bans are not guarantees.

Andrew Warner of The Next New Thing tours ten open-source projects that can replace parts of paid AI workflows. Running software yourself can mean more control and privacy, but it does not automatically provide free models, equivalent quality, or commercial-use rights. Ten alternatives VoiceStudio (00:27) — local voice cloning, narration, and dubbing instead of ElevenLabs; check the chosen voice model’s license before commercial use. Odysseus (02:42) — private chat, research, memory, and agents instead of some ChatGPT workflows; the local model and your hardware determine capability. FreeLLMAPI (05:51) — routes requests across providers’ free model allowances; not a free Claude subscription, and provider limits can change. Vane (09:27) — self-hosted, cited web research in the Perplexity mold; local models need adequate hardware, while cloud keys may cost money. InvokeAI (11:51) — an image-generation and editing studio with a layered canvas and repeatable workflows; fast local use calls for a capable GPU. ACE-Step (13:57) — generate songs and vocals locally instead of Suno; Warner says the output needs more work and may not match Suno’s quality. OpenShorts (19:30) — cut long videos into captioned vertical clips; self-hosting avoids the hosted trial’s limits but takes compute and manual reframing. Dyad (23:51) — build apps from plain-language prompts on your own machine instead of paying a builder’s credits; the underlying model can still cost money. OpenManus (26:15) — plan and execute browser, research, coding, and file tasks with a model you supply; API usage is not inherently free. Upscayl (29:06) — batch-upscale photos locally rather than subscribing to a hosted image enhancer; a compatible GPU matters, and cloud face repair is separate. The useful question is not “Which subscription can I cancel?” but “What quality, compute, licensing, and setup costs am I taking on in exchange for control?” ...

Seven picks from the GitHub trending board for the week of October 4, 2026. Agent operations lead the list, followed by useful developer infrastructure and two practical learning and media tools. Four high-ranking candidates already appeared in recent repo roundups and were left out. Running agents NVIDIA/OpenShell — a private runtime that puts autonomous agents in policy-controlled sandboxes, checks filesystem, system-call, and network access at the kernel level, and injects credentials only for approved endpoints. mvschwarz/openrig — turns Claude Code, Codex, and Pi sessions into persistent YAML-defined teams with roles, shared context, owned work, and a tmux-based control surface. pingdotgg/t3code — an open agent-harness control surface for operating local Claude Code, Codex, Cursor, OpenCode, and other agents from desktop, web, iOS, or Android. Developer infrastructure tile-ai/tilelang — a Pythonic language built on TVM for writing high-performance GPU, CPU, and NPU kernels without dropping all the way into architecture-specific low-level code. Effect-TS/effect — a TypeScript foundation for typed errors, dependency injection, structured concurrency, scheduling, tracing, schemas, and production integrations; its 4.x line is now an LTS release. Useful tools and references pablostanley/yoinks — a friendly terminal interface around yt-dlp and ffmpeg for downloading video or audio from more than 1,800 supported sites without ad-riddled download pages. cs341-illinois/coursebook — the University of Illinois’s open introductory systems-programming textbook, covering Linux and C with generated PDF, HTML, EPUB, and Markdown editions. Already covered and skipped: Paperclip and Tencent’s Octop appeared in the September 25 roundup; Impeccable and the Claude Skills collection appeared in the October 2 roundup. MoneyPrinterTurbo was left out as automated AI-video churn, and GhostTrack as a phone/location tracking tool with weak fit for this feed.

Andrew Warner and Adam Brakhane of The Next New Thing review ten trending GitHub projects in 24 minutes. This week’s strongest thread is the infrastructure around agents: organize them, give them durable memory and specialist playbooks, improve what they design, and expose conventional software through interfaces they can operate reliably. Organize agents and preserve context paperclipai/paperclip — gives agents an org chart, project-based task queue and multiplayer permissions; creator Dota says it works with different models and harnesses, can use local subscriptions, and is built around discrete assignments rather than one endless chat session. vectorize-io/hindsight — persistent memory across agent sessions, with an open-source core and a company-backed hosted path for teams that value maintenance over operating everything themselves. Voice, learning and professional workflows debpalash/VoiceStudio — a desktop interface for local voice cloning on macOS, Windows and Linux, positioned as a way to explore voices without spending hosted-service credits on every iteration. rohitg00/ai-engineering-from-scratch — a free, staged AI-engineering curriculum whose roadmap makes an intimidating field easier to enter, though the hosts warn that AI-generated course material can be verbose and uneven. anthropics/financial-services — specialist skills, tools and instructions for finance work such as comparable-company analysis, including spreadsheet conventions and source notes that make generated models easier to audit. Safer applications and better interfaces vercel/next.js — appears because a burst of security fixes drove attention to the framework; the practical takeaway is to ask an agent periodically whether any project dependencies have known vulnerabilities that need an update. pbakaus/impeccable — gives coding agents design guidance and named visual styles, adding the human act of choosing an aesthetic before the agent remakes a page or presentation. HKUDS/CLI-Anything — turns supported applications into command-line interfaces, giving agents a faster and more dependable control surface than clicking through a graphical UI. Skills and Claude Code setup alirezarezvani/claude-skills — bundles job-specific playbooks and personas; Adam prefers the persona layer because it combines relevant skills into one focused teammate instead of dumping hundreds of unrelated instructions into a general agent. davila7/claude-code-templates — a searchable collection of Claude Code skills, commands and settings with a quick installer, useful as a curated starting point but still requiring scrutiny because plugins execute code with broad local permissions. What connects the list Task-oriented coordination beats juggling long agent tabs: Paperclip makes ownership, completion state and delegation visible. Local tools lower the cost of experimentation: VoiceStudio lets people iterate before paying for a premium hosted model. Agents work better when software exposes a structured interface: a CLI is usually more reliable and token-efficient than visual clicking. Skills become more useful when composed around a role and workflow rather than accumulated as an enormous undifferentiated library. Popularity is not a security review; inspect plugin provenance and behavior before granting it access to a development machine. “What I actually want is a partner. I want an agent. I want something that understands one specific workflow.” — Adam Brakhane ...

Andrew Warner of The Next New Thing runs through nine GitHub projects that can replace paid business software. The useful frame is not merely “free”: because the tools are self-hostable and customizable, an agent can adapt them to a specific workflow — or they can become the foundation of a consulting service. Find and convert customers PostHog — product analytics, funnels, retention, session replay and experiments; use it to find where customers abandon a site, or sell conversion-leak audits to businesses paying for Mixpanel. OpenSEO — self-hosted keyword research and rank tracking that can feed custom dashboards or a CRM; Warner suggests letting an agent operate it or using it as the backbone of an SEO service. Typebot — conversational forms with branching, media blocks, WhatsApp and Zapier integrations; turn flat email-capture pages into product finders, qualification flows and segmented follow-up. Charge and get paid Invoice Ninja — quotes, invoices, subscriptions and automated reminders without per-seat pricing; useful for turning retainers into recurring billing or modernizing invoicing for local service businesses. Cal.diy — a self-hosted booking page that checks multiple calendars, adds meeting links and can collect payment or a refundable deposit before someone schedules. Autumn — open-source pricing and billing for subscriptions, trials and metered plans, including per-request or per-token charging for software with usage-driven costs. Run sales, marketing and support Frappe CRM — a customizable lead-and-deal pipeline with email, calls, tasks, Twilio and social lead capture; deploy it for a sales team without adding another per-seat bill. Mautic — self-hosted email marketing, landing pages, lead scoring, SMS and visual campaigns; Warner positions it as an owned alternative to expensive marketing-automation suites. Chatwoot — an omnichannel support inbox with unlimited internal seats and bring-your-own AI; build order-status bots, route harder cases to humans and avoid paying for every occasional support user. The recurring business pattern Use the software directly to replace a growing SaaS bill while keeping control of the data and workflow. Let an agent inspect the system, customize it and surface next actions instead of treating the repo as a static clone of a commercial product. As a consultant, find businesses using the expensive incumbent, migrate them to the open tool, then charge for setup, customization and ongoing maintenance. The tradeoff is real: “free” software still needs hosting, upgrades and operational care, so the consulting value is not imaginary — it moves spending from licenses to implementation and support. “This is about more than just free. It’s about customization.” — Andrew Warner ...

Seven picks from the GitHub trending board for the week of September 27, 2026. The top of the board is agent memory and agent-native tooling again, with a security-release tail underneath it. Four of the week’s biggest movers — Anthropic’s financial-services skills, Paperclip, Claude Code Templates, and Tencent’s Octop — were already covered in the September 25 video roundup and are skipped here. Grouped by theme: Agent memory vectorize-io/hindsight — the week’s biggest mover at ~7.3k stars gained (35.6k total, MIT, Python). The pitch is memory that learns rather than just remembers: three operations (retain, recall, reflect) over memory banks, with world facts, the agent’s own experiences, consolidated observations and mental models kept as separate structures instead of one flat vector index. Recall runs four retrievers in parallel — vector similarity, BM25 keyword, entity/temporal/causal graph, and time-range filtering — then merges with reciprocal rank fusion and a cross-encoder reranker. Self-hostable as Docker/pip/Helm or embedded with no server at all, 25+ LLM providers including local Ollama and LM Studio, an MCP endpoint per bank, and 60+ integrations — including a one-command install that wires per-repo long-term memory into Claude Code, Codex, Cursor, Copilot and friends. It claims state-of-the-art on LongMemEval, independently reproduced by researchers at Virginia Tech and The Washington Post. Agent-native software & skills HKUDS/CLI-Anything — ~1.1k stars this week (50.7k total, Apache-2.0, Python). The thesis is in the tagline: today’s software serves humans, tomorrow’s users are agents. Two halves — CLI-Hub, a package manager for ready-made agent CLIs you install with pip install cli-anything-hub, and a plugin that auto-generates one for any codebase through a seven-phase pipeline (analysis, architecture, implementation, tests, docs, SKILL.md, PyPI packaging). The catalog is the interesting part: Blender, FreeCAD, LibreOffice, GIMP, Draw.io, Zotero, Obsidian, Kdenlive, QGIS, MuseScore — real applications called directly, structured JSON out, no screenshots or pixel-clicking. It reports 2,461 passing tests and ships demos of agents driving CAD, 3D scenes and game automation. cloudflare/security-audit-skill — ~6.5k stars this week (22.2k total, MIT, JavaScript). A coding-agent skill that runs a six-phase security audit: reconnaissance into a coverage ledger, isolated hunters assigned from that ledger, adversarial validation where the verifier is never the agent that found the finding, schema-validated findings.json, independent re-verification of every surviving claim, and target-neutral reports. Findings arrive in three verdicts — confirmed, needs-validation, rejected — and severity only counts when there is demonstrated impact. It is the single-repo starting point behind Cloudflare’s fleet-wide vulnerability discovery harness, and the honest headline from their own runs is that one pass finds roughly half of what repeated passes find. Infrastructure cloudflare/quiche — ~740 stars this week (12.6k total, BSD-2-Clause, Rust). The QUIC transport and HTTP/3 implementation with a low-level packet API; it powers HTTP/3 on Cloudflare’s edge, Android’s DNS-over-HTTP/3 resolver, and curl’s HTTP/3 support. Release 0.30.0 (Sept 17) is a breaking one: PathEvent became non-exhaustive with a path-MTU variant, the supported boring range moved to include 5.x — which enables post-quantum key exchange groups by default, large enough that a ClientHello can split across Initial packets and needs set_curves_list() to rein in — and stream_readable_len() landed for contiguous readable bytes. vercel/next.js — ~280 stars this week (142.8k total, MIT). On the board for a security release rather than a feature: 16.3.6 (Sept 22) patches CVE-2026-94545, a critical (CVSS 9.5) remote code execution in the Node.js next/og ImageResponse implementation, triggered when attacker-controlled values reach SVG content, attributes or styles during image generation. Affected range is 16.2.0 through 16.3.5; the Edge implementation is not affected. If anything self-hosted runs Next 16.2+, this is the patch. pytorch/pytorch — ~300 stars this week (103.4k total, Python). Steady traffic, but 2.14.0 (Sept 2) is worth skimming: NVGEMM brings CuTeDSL-generated CUTLASS kernels to Inductor with epilogue fusion, autotuned alongside Triton and ATen; torch.switch generalizes torch.cond to multi-way branching; torch.while_loop can now be captured in a CUDA graph; and dynamic shapes get a declarative @dynamic_spec shared across torch.compile, torch.export and make_fx. Self-hosting odds and ends FxEmbed/FxEmbed — ~420 stars this week (5.5k total, MIT, TypeScript). The service behind fxtwitter.com, fixupx.com and fxbluesky: put fx or fixup in front of a Twitter/X or Bluesky link and clients like Discord and Telegram render a proper embed — multiple images, video, polls, translations — instead of collapsing it. It is a Cloudflare Worker, but there is a documented self-hosting path (Docker running the local Workers runtime through Wrangler) with your own domains and branding, which is the reason to care if you would rather not route links through someone else’s domain. Registry note: all seven repo URLs are recorded in the roundup registry, so next week’s script won’t repeat them.

Andrew Warner and Adam Brakhane run through the week’s trending GitHub repos in 35 minutes, and this list has an unusual shape: three of the ten are ways of giving your agent a job instead of a tool, and the rest are scaffolding around coding agents — shared instruction files, a marketplace, and a console for running several of them side by side. The three getting extra attention NewsJack (1.4k stars) — turns your agent into a full PR team: it finds reporters covering topics you should be part of, finds the stories worth wedging into, and helps you get in front of them. It has been out for a while; what set it off this week is Jev, the small purpose-built model the demo runs on. The hosts’ point is the one they keep making about small models: Jev is not replacing a frontier model, it is making a decision loop (is this story right for this client?) fast enough and cheap enough to run constantly. Adam says his firm swapped it in for the flash-tier models they used for profile evaluation and the difference was night and day — right down to running ELO-style tournaments between candidates, which used to be too expensive to bother with. OpenSEO (21.1k stars) — an open-source Semrush and Ahrefs alternative. You run it yourself and pay only for the data source, or pay roughly $10/month for hosting. It got attention for a different reason: the creator announced he was in Y Combinator as a 100% open-source company, and people piled on that he had “vibe coded a wrapper on SEO data.” Adam is a paying customer and calls it a wrapper of a wrapper — and then makes the business argument: raising $500M and making $5M a year is a failed company, raising $1M and making $5M a year is a great one, so a YC round for a mid-tier SaaS that is simple and clear is not obviously wrong. Paperclip (85k stars) — organizes multiple agents into something like a company: an org chart for agents, a task list, and one surface where a team can see who is doing what. It blew up after a sponsored YouTube video (roughly half a million views) making the case that it manages agents like employees. The hosts flag both sides: the video was paid promotion, which at least means the people behind it are spending money to grow it, and Paperclip does not yet work with the consumer-grade agent apps — it is aimed at more advanced setups, and Adam’s read is that corralling agents from different providers into one place is where the opportunity is. The top 10 Anthropic Financial Services (37.6k stars) — ten finance agents that do the grunt work: know-your-customer screening, stock analysis, the workflow a finance team would otherwise staff. The hosts’ framing is floor versus ceiling: a novice takes the skill off the shelf as-is, a professional who knows the domain customizes it and gets superpowers. Andrew admits he blew these off last time they came up, until he met a lawyer running Anthropic’s legal skills to shave hours off his week — and who used the reclaimed time to start studying to be a sommelier. Alibaba Open Code Review (41.3k stars) — the code reviewer that started inside Alibaba and got open-sourced. Hybrid architecture: deterministic pipelines plus an LLM agent, line-level comments, built-in rules for the classic bug classes. It keeps growing because people keep trying to sneak bugs past it and it catches them. The week’s change is a VS Code extension, which Adam argues is the point: you can do everything in a terminal, but extensions bring the buttons, panes and automations that make a tool comfortable. Claude Code (148.1k stars) — not open source, on GitHub for issues and feature requests, and it just started reading agents.md instead of only claude.md. See the section below; this was the change of the week for anyone juggling several agents. ECC (267.5k stars) — the biggest number on the list again: one engineer’s decade-plus of practice turned into an agent harness. What the hosts single out is that the standard loop (plan, test, implement, review, verify) is bookended by remember and improve — steps that write back to agents.md, save memories and save scripts, so the next pass is easier. Most skill packs skip that, which is why you hit the same blockers in week ten that you hit in week one. Tencent WeKnora (30.1k stars) — point it at the files on your computer and it builds a queryable RAG, a self-maintaining wiki and a knowledge graph. Adam’s one-line description is the cleanest on the episode: Obsidian plus RAG plus a graph database plus integrations. Instead of hand-organizing everything, it reads your files and works out that the receipt in your downloads folder relates to that project — then links them. Addy Osmani Agent Skills (99.1k stars) — 25 production-grade engineering skills from an engineer who spent 14 years at Google and then Anthropic. The hosts point Claude at it to compare against ECC: both are permissively licensed, both work with all models, ECC has 292 skills to Agent Skills’ 25. Adam’s analogy is the useful part — ECC is a full workshop (great if you know what each tool is for, overwhelming if you do not), Agent Skills is a toolbox with one of everything. Anthropic Knowledge Work Plugins (25.6k stars) — plugins for finance, HR, legal, marketing and operations. The hosts’ audience for this is the everyman: you started a business because you are good at one thing, and now there is a marketing plugin and a legal plugin you can actually talk to. Adam’s plan is to hand the set to their own hyper agent and ask which of these skills the team should adopt based on how they already work. Orca (78.3k stars) — run five coding agents at once and keep the best result. One window for Claude Code, Codex and others, plus a mobile connection, and worktrees so parallel agents stop butting heads over the same files. Adam’s framing: VS Code is built around the code, and Orca is built around the agents — he expects these agent-first interfaces to replace the IDEs people have used for two decades. Tencent Cloud Octop (5k stars) — a self-hosted, multi-user, multi-agent assistant aimed at teams and families on one machine. The hosts’ critique is its integrations: Telegram and QQ instead of iMessage and WhatsApp, which makes it a tool with a real audience that is mostly not in the United States. Andrew’s aside is that there is still an opening for someone to build a genuinely simple family agent. Claude Code Templates (31.8k stars) — a marketplace for Claude Code add-ons: skills, agents, commands, plugins and MCP connectors in one browsable place, with download counts, stars and search. Roughly 900 skills and 34 plugins, and Adam argues the curation is the feature — GitHub is a supermarket nobody can navigate, so a smaller catalog you trust is worth having bookmarked. ECC sits at the top of it. The agents.md truce Claude Code now reading agents.md is a small change with a big effect. Codex, OpenCode and the rest already look at agents.md; Claude only ever read claude.md, so anyone switching between agents had to maintain both files and forget one of them. There were hacks around it — symlinks, cross-references — and Adam’s summary is that the mess just goes away: rename claude.md to agents.md and every agent sees the same instructions. Skills and MCP connectors are still not portable, and that is the next sore spot. Adam’s example: a Google Calendar automation worked in one agent and died in another because the MCP connector was installed on one side only. Where the expertise lives The week’s theme, in the hosts’ words, is packaging: skill packs take one person’s years of practice and hand them to everyone else, plugin marketplaces make them findable, and consoles like Orca make several agents workable at once. Andrew’s offer to viewers: he is talking to Mustafa Afaan, who built ECC, about a full walkthrough — ask in the comments what you would want covered. What they are building Andrew has an agent watching his calls and Slack for passion signals (an F-bomb counts) that reports back the next day with the three things he was most worked up about, so he can turn them into posts and guest invitations. Adam is 85-90% through rebuilding a call recorder on a Mac mini: Granola’s on-device recording with the speaker tagging Fireflies gets right, minus the conference-room problem where everyone is labeled as one person. It will cost slightly more per seat than the two tools it replaces, and he is giving it to the people in their residency anyway. “If you are using VS Code today, there is a really good chance that there is no code on your screen — it is an agent, and a folder, and a plugin over here. VS Code is still designed around the code.” — Adam on why Orca is interesting ...

Eight picks from the GitHub trending board for the week of September 20, 2026. The top of the board is agent orchestration — an app for running a fleet of coding agents, and a CLI that gives an agent access to the platforms that normally block it — over the usual steady infrastructure. Grouped by theme: Parallel agent orchestration stablyai/orca — the week’s biggest mover at ~5.4k stars gained (73k total, MIT). An ADE for running a fleet of coding agents: fan one prompt across five agents, each in its own isolated git worktree, then compare the diffs and merge the winner. Runs Codex, Claude Code, OpenCode, or Pi on your own subscription rather than an API key. Desktop plus a mobile companion app for steering sessions from a phone, SSH worktrees for running agents on a bigger remote box, Ghostty-class terminals, and a “design mode” that clicks a UI element in an embedded Chromium window and drops its HTML, CSS, and a screenshot into the prompt. Panniantong/Agent-Reach — ~3.9k stars this week (84k total, MIT). One CLI that gives an agent read/search access to the platforms with the highest information density and the worst APIs: Twitter/X, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, plus generic web pages to clean Markdown. The pitch is that it picks and installs a backend per platform, keeps a primary plus fallback route so a blocked path gets swapped without your involvement, and keeps cookies local. agent-reach doctor reports what works and what does not. Free backends only, no paid API keys. Note the main README is Chinese; the English version is at docs/README_en.md. The repo also carries an explicit disclaimer that it has no token, coin, or wallet connection, which is worth repeating given the name has been borrowed by unrelated projects. Coding agents & chat cline/cline — ~1.1k stars this week (69k total, Apache-2.0). The open-source coding agent as an SDK, IDE extension, CLI, and native macOS/Windows desktop app. The CLI runs interactive or fully headless for CI/CD and scripting; the desktop build adds scheduled routines and model/plugin/MCP management. danny-avila/LibreChat — ~1.6k stars this week (44k total, MIT). The mature self-hosted multi-provider chat UI: agents, MCP, skills, artifacts, code interpreter, message search, and per-user auth, with DeepSeek, Anthropic, OpenAI, Azure, AWS Bedrock, Groq, Mistral, Vertex, and OpenRouter all wired in from the same interface. The long-running answer to “I want the ChatGPT UI without the account.” anthropics/knowledge-work-plugins — ~1.0k stars this week (25k total, Apache-2.0). Eleven role-shaped plugins for Claude Cowork (also compatible with Claude Code): productivity, sales, customer support, product management, marketing, legal, finance, data, enterprise search, bio-research, and one for building more plugins. Each bundles skills, slash commands, sub-agents, and MCP connectors for a specific job function. Everything is markdown and JSON — no code, no build step — so the intended move is to fork them and swap in your own tools and terminology. A useful read even if you never install one, as a worked example of what a well-organized agent skill set looks like. Infrastructure supabase/supabase — ~1.4k stars this week (110k total, Apache-2.0). The Postgres development platform: auth, realtime, storage, edge functions, and pgvector embeddings on top of a plain Postgres database you can also self-host. Still the most common way to stand up a backend for a small product without assembling five services. home-assistant/core — ~420 stars this week (91k total, Apache-2.0). Local-first home automation with an enormous integration surface. No cloud dependency in the default path, which is the entire reason it keeps growing. cilium/cilium — ~360 stars this week (25k total, Apache-2.0, CNCF). eBPF-based networking, security, and observability for Kubernetes — the layer that replaces kube-proxy and adds policy and flow visibility. Heavier than most solo builders need, but the reference implementation for what eBPF can do to a data path. Registry note: all eight repo URLs are recorded in the roundup registry, so next week’s script won’t repeat them.

Andrew Warner and Adam Brakhane run through the week’s ten trending GitHub repos in 26 minutes, and the list is mostly guardrails: something that reviews the code your agent wrote, something that makes its answers short enough to read, something that keeps its context clean, and two skill packs that hand it an engineering process instead of hoping it has one. The top 10 Open Code Review (36.6k stars) — Alibaba’s code-review tool, number one this week. Deterministic pipelines plus an LLM agent, line-level comments, and built-in rules for the classic bug classes (null pointers, thread safety, XSS, SQL injection). It ships a CLI and agent skills, and Andrew’s pitch is wiring it into a Git hook so every commit gets checked before it lands. The demo he cites: someone deliberately planted a plaintext password, and the tool found it, then produced a full HTML report from one command. I Have ADHD (48k stars) — still growing, and the hosts keep coming back to it. One skill file that tells your coding agent to keep answers short and scannable. Andrew shows a before-and-after from the Claude web app where a slash command turns a wall of text into something readable — and Adam’s reason it keeps mattering is that the models themselves are getting more verbose. God’s Eye View (37.8k stars) — open data on a photorealistic 3D globe: TomTom traffic, public CCTV cameras you can click into live, aircraft overhead. Also distributed as a plain installable app, so you don’t have to assemble it from source. ECC (262k stars) — the biggest number on the list. A workflow pack that encodes planning, test-driven development and review as commands you invoke rather than advice you remember. Adam’s framing: it injects decades of engineering habits into a session, which matters most to people who never picked those habits up. Context Mode (23.5k stars) — a sandbox for tool output. Instead of your main session loading fifteen files to answer one question, a side session does the searching and hands back the answer, and that noisy conversation gets discarded. The advertised figure is 98% less context clutter; the caveat the hosts add is the useful part — compress something the model actually needed and it re-requests the full context, which is expensive on a long conversation. WeKnora (27k stars) — Tencent’s knowledge platform. Point it at your files and it builds a queryable RAG, a self-maintaining wiki and a knowledge graph. Adam’s description is the clearest one on the episode: Obsidian plus RAG plus AI plus a graph database, where the links between people, meetings and projects get built for you. OpenAI Plugins (7k stars) — 62 plugins written for Codex, each bundling an integration together with the skills and scripts that make the integration useful. Claude will either convert them or bridge them through a connector, and Adam’s advice is to install from the platform you actually use. Worktrunk (8k stars) — a CLI for Git worktrees, the feature that gives you several checked-out copies of the same project side by side on disk. Creating, listing, merging and deleting a worktree each become one short command, which is what makes running several agents on the same repo practical. Agent Skills (96.4k stars) — Addy Osmani’s pack of 25 skills that the hosts single out because it is one ordered workflow (spec, plan, build, test, review, ship) rather than a pile of unrelated files. Adam’s point: most people skip straight to build, and test and review are the two steps that get dropped. Claude Code (146.3k stars) — the repo exists, the code does not. Adam’s read: it gives people a normal place to file bugs and lets Anthropic track issues internally, so it is a repository minus the source. The skills vote Last week the hosts asked viewers whether to keep covering skills at all. The result was 32 keep to 9 kill, with strong feelings on both sides, and Warner says he spent a Saturday trying to answer the comments individually. The argument that moved the conversation: skills matter most for the less capable models. A frontier model already knows how to run test-driven development; a local or open-weight model may need the steps spelled out. Warner’s middle position: even a skill he never installs shows him the creator’s mental model. He also cut a separate video on the most-installed skills according to Vercel, whose CLI is how many of them get installed. The verdict for the format: skills stay in the roundup while they keep trending. Plugin security Adam’s distinction: a skill is text and the model stays in charge. A plugin can ship scripts and code, and installing one means handing over your files, your source and your keys. His rule: install plugins from the vendors — OpenAI or Anthropic — and treat third-party plugin marketplaces the way you treat installing unfamiliar software. Viewer submissions Logic Loop (12 stars) — one workspace for several concurrent agent sessions. The premise on screen: with five agents running, the thing that breaks is your memory, not theirs. Claude Dashboard (5 stars) — a local board for every Claude Code session on a machine: live sessions, transcripts, costs, git status, and which one is stuck. Deck Gauge (4 stars) — one board over Jira, GitHub, GitLab and Azure DevOps with DORA metrics. Warner pushed back on pulling tasks out of Jira, then made the case for it anyway. Campus AIOS (10 stars) — an AI staff in one folder: a chief of staff plus departments that remember your business. Reckoner (47 stars) — a credits and billing dashboard across AI providers, still climbing since it got its first star on the show last week. “Why not just give each of those agents its own little playground where it can have its own copy of the code, test it itself, edit it, and then once the feature is done, merge all of them back into your one main branch?” — Adam on Git worktrees ...

The Next New Thing’s Andrew Warner and Robbie Abed (founder of NoMiddle) go through the ten most-installed agent skills on GitHub, ranked by installs through Vercel’s skills CLI. It is a short episode — twelve minutes — and four of the ten entries are Matt Pocock’s, which is its own finding: a handful of small text packages are becoming the default instruction layer for coding agents. The top 10 find-skills — the most-installed skill is a skill for finding skills. You ask your agent for something (“find a skill for writing tests”) and it comes back with candidates to install. Warner’s framing: the community needed an app store and built one as its first app. grill-me — a few lines that invoke the grilling skill. Warner ran it to prep for the episode and it opened by asking what he was trying to convince the listeners of, which he calls the endgame question, asked before any work started. frontend-design — Anthropic’s skill for stopping the signature AI look, starting with that warm cream background. The episode shows the same site built before and after: same request, different color logic and more deliberate layout. agent-browser — headless browser control for an agent, no visible mouse. The examples on screen are the practical ones: log into the dashboard and screenshot the billing page; open staging, sign up as a test user, and tell me if it works. setup-matt-pocock-skills — a one-time scaffold that wires his engineering skills into a repo and configures the things those skills assume: issue tracker, triage labels for the five canonical roles, where domain docs and ADRs live, and how consumers should read them. Robbie’s read: it is the first 60 days of a consultant, delivered as setup. vercel-react-best-practices — roughly 70 React and Next.js performance rules handed to the agent as rules. You do not have to understand all of them for the skill to be worth installing, since the point is the mistakes the agent now avoids. grilling — the engine behind grill-me, split out so it can be reused anywhere the agent needs to interrogate you: how to sequence the questions, how to read the answers, what to push on. Both hosts treat it as the same idea in a reusable container. lark-doc — ByteDance’s workplace platform, Feishu outside China and Lark inside: contacts, messenger, the spreadsheet equivalent, calendar, wiki, video conferencing. Warner is candid that almost nobody watching will use it — it is on the list because the user base is enormous. teach — turns the agent into a tutor that accumulates lessons across sessions as numbered HTML files in a lessons folder. Matt’s demo teaches him to solve a Rubik’s cube; the interesting part is that it remembers where he stopped and what he stumbled on, and builds the next lesson on both. web-design-guidelines — Vercel’s web standards as an agent checklist: icon-only buttons need an aria-label, form controls need labels and autocomplete, plus typography and animation rules. The demo has the agent review a checkout page and report that the icon button has no accessible name. Where the list is really coming from Vercel shows up four times (find-skills, agent-browser, React best practices, web design guidelines) and Matt Pocock four times (grill-me, setup, grilling, teach). Two authors account for eight of the ten slots. The grill-me / grilling split is the pattern to notice: one skill is the entry point (“grill me before I build”), the other is the reusable engine underneath it. Install counts measure the install path as much as the content — Vercel’s CLI is the thing people install skills through, so its own skills are the easiest to install. Skills are readable. The hosts keep pointing out that you can open the file and read plain English instructions, which is what makes them inspectable rather than magic. Where two skills disagree frontend-design and web-design-guidelines can pull in opposite directions: one optimizes for a distinctive, animated interface, the other for accessibility compliance. Robbie’s example is a healthcare clinic — the animation that reads as tasteful on a portfolio site fails an accessibility review, and the two installed skills will argue about it through your agent. The practical upshot: skills stack, and stacking them means owning the conflicts rather than assuming they compose. The takeaway Ranked by installs, the top of the list is not about making the model smarter. It is infrastructure around the model: discovering other skills, interrogating a plan before it is built, keeping the codebase’s rules in front of the agent, and controlling a browser or a workspace. The two that lean hardest on taste — frontend-design and web-design-guidelines — are also the two most likely to contradict each other, which is a fair description of where agent tooling is right now. ...

Eleven picks from the GitHub trending board for the week of September 13, 2026. The week is heavy on agent skills — Claude Code and Codex are now the distribution channel for everything from diagrams to CAD — and on the plumbing that makes running several agents at once bearable. Grouped by theme: Agent skills & methodology cathrynlavery/diagram-design — 39 editorial diagram types as agent skills for Claude Code, Codex, Factory Droid, and Pi; self-contained HTML + SVG output, with layout grammars for Sankey, fishbone, Wardley maps, kanban, user journeys, dependency graphs, UML class, story maps, and database schemas. The week’s biggest mover at ~7.4k stars gained, and the explicit pitch is “no Mermaid slop.” obra/superpowers — an agentic skills framework plus a software-development methodology, installable across the major coding agents. The largest repo here by total stars (~286k), and the clearest statement of the “skills as the unit of reuse” thesis. github/spec-kit — GitHub’s own toolkit for spec-driven development: write the spec first, then let the agent plan and implement against it. Worth a look if your workflow already starts from a written spec rather than a prompt. coreyhaines31/marketingskills — marketing skills for Claude Code and AI agents: conversion-rate optimization, copywriting, SEO, analytics, and growth engineering. Turns the agent into a marketing hire rather than a codegen tool. petergyang/no-ai-slop — a skill that strips 20+ recognizable AI-writing tics out of any draft. Useful as a final pass on agent-authored prose before it ships anywhere public. earthtojake/text-to-cad — a library of agent skills for CAD, CAE, and CAM work, i.e. driving mechanical-design and simulation tooling from natural language. Actively updated and the most concrete example this week of skills reaching past software. Running several agents at once kunchenguid/firstmate — you talk to one agent; it runs the crew. Spawns autonomous agents in a visible session backend, gives each its own git worktree, supervises them to completion, and hands back finished PRs, approved local merges, or investigation reports. Aimed squarely at the tab-juggling problem. max-sixty/worktrunk — a Rust CLI for git worktree management built for parallel agent workflows: create, switch, and clean up per-agent worktrees without remembering paths. Small pick (~257 stars this week) but the most boring-necessary tool on the list. Documents, knowledge & media microsoft/markitdown — Microsoft’s Python tool for converting files and office documents into Markdown, presumably for feeding them to models. The mature, high-traffic option at ~183k total stars. Tencent/WeKnora — an open-source LLM knowledge platform in Go: raw documents in, queryable RAG + an autonomous reasoning agent + a self-maintaining wiki out. The self-hosting-friendly answer to the hosted document-QA products. heygen-com/hyperframes — “Write HTML. Render video. Built for agents.” Frames the render pipeline as an agent target, with a component catalog and playground, so programmatic video stops being a bespoke FFmpeg job. Registry note: all eleven repo URLs are recorded in the roundup registry, so next week’s script won’t repeat them.

The Next New Thing’s Andrew Warner and Adam Brakhane run through this week’s GitHub roundup, and it lands heavily on skills: small packages that tell a coding agent how to think before it writes anything. Andrew opens by calling it a skills-heavy week, and Adam spends the episode arguing that most of these skills package something you could just say out loud. Then OpenAI deprecates its skills repo mid-roundup, which kind of makes Adam’s point for him. ...

The Next New Thing’s Andrew Warner and Mat Nolen test a collection of GitHub projects aimed at one of the biggest costs in agentic coding: context and token usage. Ten repos get demoed on camera (plus one more flagged in the episode’s resource list), and for most of them Mat reports his own measurement next to the project’s claim. Routing around the bill Free LLM API — stacks the free tiers from multiple providers behind one interface and automatically switches to the next one when a tier runs out. Warner interviewed the founder, who says providers now pitch him for inclusion because they want to be the entry point where users experiment. Ships Mac, Windows, and Android apps. OmniRoute — one endpoint, hundreds of providers, free tier first. Unlike Free LLM API it also routes across your paid providers (Anthropic to OpenAI and on down the list) when a quota runs dry, using the RTK compression engine on the way through. The hosts saved it for the end because it folds most of the other tricks into a single gateway. Compressing what comes back Headroom — a compression layer that sits between your agent and the API and squeezes tool output on the way back: it shortens what you read, not what you write. Mat’s test compressed a file to roughly 37% of its size and he had no complaints — at times it beat the README’s numbers. Magic Compact — a lossless context compressor that swaps a large tool output for a one-line note and moves the payload to disk on the logic that context is expensive and disk is cheap, so move it rather than delete it. Mat ran it eight times and got identical results every time. Low profile at 147 stars, and the developer has paused it to build a successor around agent memory. PX Pipe — the viral experiment: render the context as images and have the model read the picture instead of the text. It worked, but accuracy slipped, especially on older or local models that read images poorly — and the failure mode is a hallucinated context. 7,000+ stars off the back of the X attention alone. LeanCTX — change one import line and compress prompts by dropping low-value filler words. The project claims 40–60% savings; Mat measured 52%. His caveat: you do not get to see exactly which words were thrown away. Keeping context out of the window Context Mode — keeps raw tool output out of the context window entirely, storing it in a local index the model queries when it needs the data. The README claims a 98% reduction; Mat measured around 95%, which he called a hit. Token Optimizer MCP — an MCP server that blocks expensive calls the agent has already made and points it back at what it already read, acting as a short-term memory layer. Watch this one: installing it turns off Claude Code’s trust prompt, which the hosts flag as a production no-no. Graphify — parses a codebase into a knowledge graph so the agent queries the map rather than opening files, and it builds that map with zero model tokens, so you can try it without spending anything. 100,000+ stars and a Y Combinator batch behind it. codebase-memory-mcp — listed in the episode’s resources as a local memory server for codebases; not demoed on camera. Writing less instead of cheaper Ponytail — the jaded senior developer sitting inside your agent: where an overeager junior installs a library and wraps it, this pushes the agent toward what the browser or the codebase already has. Mat measured the cheapest token usage of the batch, and it stayed on par with its claims. The takeaway The pattern across the roundup: most of these savings come from moving work out of the context window — to disk, to a local index, to a graph, or to another provider — rather than from making the model smarter. The two caveats worth remembering are the ones the hosts surfaced themselves: tools that strip context you cannot inspect, and tools that loosen your agent’s permission settings on install. ...

Ten picks from the GitHub trending board for the week of September 6, 2026. Agent tooling still dominates — but the week skews less toward plugin ecosystems and more toward agent temperament (a “laziest senior dev” harness, a research-first skill set) plus local-first AI and networking utilities. Grouped by theme: Agent tooling & skills DietrichGebert/ponytail — makes your AI agent think like the laziest senior dev in the room; the week’s biggest mover at ~11.7k stars, built on the thesis that the best code is the code you never wrote. affaan-m/ECC — an agent-harness performance optimization system: skills, instincts, memory, security, and research-first development layered over Claude Code, Codex, Opencode, Cursor, and beyond. Imbad0202/academic-research-skills — a research → write → review → revise → finalize skill set that walks Claude Code through a full academic workflow. ChromeDevTools/chrome-devtools-mcp — Google’s official MCP server giving coding agents the Chrome DevTools surface: DOM inspection, network, console, and performance traces. Local-first AI & media debpalash/VoiceStudio — a fully local, open-source ElevenLabs alternative: voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation across 646 languages. magnitudedev/magnitude — an open-source inference server that runs the best local models your hardware can handle and plugs into the agent you already use — Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, and more. pollen-robotics/microduck_rl — RL training environments for Pollen Robotics’ Microduck platform (mjlab), so robot policies get trained and iterated in simulation. Builder utilities tailscale/tailcat — netcat over Tailscale’s data plane without its control plane: pipe bytes between machines on your tailnet with minimal moving parts. majd/ipatool — a command-line tool for searching and downloading .ipa packages from the App Store across iOS, iPadOS, tvOS, and visionOS — handy for pulling store builds down for inspection or sideload testing. Reference punkpeye/awesome-mcp-servers — the community directory of MCP servers; the fastest route from “I want my agent to drive tool X” to a working server for it.

The Next New Thing’s Andrew Warner and Adam Brakhane look at this week’s GitHub projects, with a recurring theme: open-source tools are turning expensive or specialized AI workflows into things people can run, inspect, and customize themselves. The roundup includes ten featured repos and seven audience submissions. The top 10 Archify — turns a system description into an interactive, clickable architecture diagram. The diagram is editable, so it can serve as both documentation and a living spec rather than a disposable generated image. God’s Eye View — a browser console for real-time planes, ships, satellites, wildfires, and public cameras. The useful core is consolidating scattered public tracking data into one navigable interface. OpenMAIC — Open Massive AI Empowerment Course: generates an immersive learning environment with slides and simulated classmates instead of returning another chat transcript. Go Modern Guidelines — gives coding agents current Go package, feature, and best-practice guidance so they are less likely to write code based on stale training data. OpenClaude — a Claude Code-style workflow rewired for OpenAI, Gemini, DeepSeek, or local models. The hosts could not identify a clear advantage over existing open tools, and the project itself warns that its legal position is unsettled. Scientific Agent Skills — more than 160 specialized skills for research and lab work, from querying scientific databases to genomics and chemistry calculations. Its specificity is the point: scientists can use the one or two skills relevant to their niche. MiniMind — an approachable recipe for training a small language model from scratch, including pre-training, fine-tuning, and preference optimization in readable PyTorch. OpenSEO — an open, customizable alternative to expensive SEO suites. The software is free; the data source costs money, or the hosted version costs $10/month. TimesFM — Google Research’s model for forecasting time-series data such as signups, impressions, or temperatures. Hindsight — reviews coding-agent sessions so mistakes and useful lessons can feed back into future work instead of being rediscovered every time. Viewer submissions Model Deck — tracks remaining usage across AI subscriptions. Cortex Suite — persistent memory and codebase analysis for longer agent sessions, with Graphite integration. RTA Smriti Brain — local-first, evidence-aware project memory for coding agents. YouTube Subscriptions Ingest — turns subscribed videos into searchable Markdown; its first ingest processed 11,500 videos from 41 channels in 10 minutes. Reimagine It — creates a reusable visual design system that an agent can apply directly from an HTML style file instead of guessing at a vibe. Radiant — a native Mac app where multiple agents share a conversation, consult one another, and can use subscriptions or local models. The stronger pattern across the roundup is practical rather than magical: make system context, design taste, model training, and agent memory explicit artifacts that tools can reuse. ...

Andrew Warner’s (The Next New Thing) weekly repo roundup, this time themed: ten open-source projects that replace paid subscription software. The pitch is bigger than free — every pick is source you can own, run yourself, and build on. Doodle Note — Granola alternative: a local meeting note-taker that captures audio from your computer (no bot joins the meeting), transcribes on-device, and never lets the audio leave your machine. Mac/Apple silicon now, Windows in beta — new and low-starred, so expect rough edges. Cap — Loom alternative: screen + camera + mic recording with instant share links and a studio edit mode. 21k stars; free cloud for videos up to 5 minutes, or self-host. Note: the official app wants a commercial license for client work. Vaultwarden — 1Password alternative: an unofficial Bitwarden-compatible server (66k stars, since 2018) that works with the official Bitwarden apps on iPhone, Android, and browser. Organizations, collections, sharing, admin page included. Your box, your backups — if the host dies, so do the passwords. Ghost — Substack alternative: non-profit, MIT-licensed publishing since 2013 (55k stars). Built-in paid memberships with 0% revenue cut, on your own domain and design. Self-hosting is work, but a hosted plan exists for writers who don’t want the job. Twenty — HubSpot alternative: self-hosted CRM with contacts, companies, pipelines, and workflows on a box you own (a cloud option exists). Caveat: the GitHub license is “other” — some files are commercial. LanguageTool — Grammarly alternative: grammar, spelling, and punctuation checks across 30+ languages and dialects. Point the browser add-on at a local server and your sentences never leave your system; the free tier beats Grammarly’s without the constant upsell. Umami — Google Analytics alternative: privacy-focused, no personal tracking, no cookie banners (GDPR-compliant), fast, and it bypasses most ad blockers for accurate counts. Missing: Facebook ads connection and advanced funnels. Nextcloud — Google Drive alternative: self-hosted files, sync to the computers and phones you already own, share with access control, plus contacts and calendars in one system. Setup is a process (Andrew’s involved a NAS and a support call), but you pay once instead of renting storage. Chatwoot — Intercom alternative: live chat, email, WhatsApp, Instagram, Telegram, and social messages in one inbox, with a help center, private notes, labels, and canned replies. Disappointment flagged: the AI agent features sit behind the paid plan — you can’t bring your own agent. SlowBooks Pro — QuickBooks alternative: books live on your computer — invoices, bills, US payroll, inventory, bank feeds — and the data is queryable by AI (connect Claude, Codex, Grok) for accurate reports. Source-available license, mostly one fed-up dev; backups are on you.

Ten picks from the GitHub trending board for the week of August 30, 2026. The week skews agent skills and plugin ecosystems — diagram skills, a science-skills library, the official Claude plugin directory, Cursor’s plugin spec — with a few solo-builder gems mixed in (an AI job-search framework, a screenshot-to-code classic). Grouped by theme: Agent tooling & skills archify (tt-a1i) — an agent skill for architecture, workflow, sequence, data-flow, and lifecycle diagrams: self-contained HTML with motion, crisp export, and verifiable structure. The week’s biggest mover at ~14.9k stars. scientific-agent-skills (K-Dense-AI) — “turn any AI agent into an AI Scientist”: 165 validated agent skills plus 100+ scientific databases covering biology and beyond; claims 190k+ scientist users. garden-skills (ConardLi) — an open-source collection of reusable agent skills covering web design, knowledge retrieval, and image generation. claude-plugins-official (anthropics) — Anthropic’s own managed directory of high-quality Claude Code plugins — the official counterpart to the community marketplace that topped last week’s roundup. cursor/plugins — Cursor’s plugin specification and official plugins; more evidence that plugin ecosystems are becoming the organizing unit of agent tooling. Local-first AI & learning openhuman (tinyhumansai) — a Rust “personal AI superintelligence”: a brain that builds a local-first memory of your life, orchestrates agent fleets and workflows, and does deep research. ai-engineering-from-scratch (rohitg00) — “Learn it. Build it. Ship it.” — a from-scratch path into AI engineering, trending at ~3.5k stars. OpenMAIC (THU-MAIC) — open multi-agent interactive classroom: an immersive multi-agent learning experience in one click. Solo-builder utilities ai-job-search (MadsLorentzen) — an AI job application framework that runs on your machine via Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it — no SaaS. screenshot-to-code (abi) — the classic that keeps trending: drop in a screenshot, get clean HTML/Tailwind/React/Vue code.

The Next New Thing’s weekly roundup of the 10 hottest GitHub repos. This week’s crop leans AI-heavy (a security-scanned Claude plugin registry, OpenAI Codex going open source, Apache incubating its own agent workspace) with a few pure-software gems (an open-source Logitech driver replacement, DHH’s Mac-seamless Linux, a self-hosted net-worth tracker). The picks: Claude Plugins community repo — 2,282 Anthropic-approved plugins, security-scanned, and formatted machine-readable so Claude can search it directly instead of scrolling the website. The hosts’ caveat: plugins run real code on your machine — vet them like Mac apps. GPT Image 2 prompt library — a searchable gallery of image prompts with their outputs; copy one and edit the sections you want different instead of writing prompts from scratch. Omari (DHH) — “Linux that’s as seamless as a Mac,” aimed at reviving old machines macOS no longer supports; $10M raised, Touch ID support in the works. OpenLogi — open-source replacement for Logitech’s bloated driver software: no login, no birthday questions, no telemetry phoning home. OpenAI Codex is open source — the CLI is free, editable, and redistributable: swap in cheaper models, rewrite the prompts, or fork it into your own tool. (Claude Code, by contrast, is closed.) Apache AI agent workspace (incubating) — a foundation-backed agent runtime with an append-only log of every model call, tool call, and permission. The durability bet: “I wouldn’t build a business on vibecoded agent tools; I would on Apache.” Mojo — Python-like syntax with C/C++ performance, aimed at the AI-era workload of LLM-generated code. Money Printer Turbo — open-source CapCut: all-AI or AI-composited video from existing footage and audio, all the knobs, no upsell. 34 free AI tiers stack — a rotation of free API tiers to hop between when you burn through one; $19/yr for the maintained list. TOS-gray, but handy for agent experimentation. Securo — self-hosted net-worth tracking wired to 2,000+ banks via SimpleFIN ($1.50/mo); the local, no-upsell alternative to the Personal Capital/QuickBooks path.

Github Awesome’s weekly scan of the GitHub trending board — episode #46, 35 projects (14:54). Full text version at githubawesome.com/github-trending-weekly-46/. The week skews agent infrastructure and local-first tooling. Grouped by theme: Data & databases latticedb — embedded graph database in a single file, written in Zig with zero dependencies: relationship traversal, vector similarity, and full-text ranking in one query language over the same data. A RAG query can hop from embedding → document → author in one statement instead of gluing three systems together. selfdb — a program that IS a SQLite database instead of an ELF file; the interpreter rebuilds/maps the data at launch, and SQL can inspect dependencies or update embedded content. Experimental, NixOS-heavy format — by fzakaria, whose “your executable is a SQLite database” essay is already in the /software/ feed. kern — starts a real OCI container in ~3.5ms from a 1.52MB static binary: rootless, no daemon, nothing running at rest, whole Rust dependency tree is libc. Agent tooling & memory Gradient — trains research agents with reinforcement learning (GRPO): the agent searches a synthetic company workspace, cites evidence, records every tool call, and gets scored on correctness, citation quality, and efficiency. proliferate — runs Claude Code through its own native harnesses: each task gets its own git work tree, agents delegate to subagents, and MCPs/skills/custom tools are configured once and shared. factory (addyosmani) — installs a repeatable delivery loop into a GitHub repo: issues become the cue, fresh Claude Code/Codex sessions triage bounded work, run project gates, and hand the diff to an independent verifier before a draft PR. Humans own every merge. ownmem — coding-agent memory as reviewed markdown inside the repo: Claude Code/Codex/Cursor/Grok recall the same decisions, ranked by a deterministic BM25F index — no model call, no network. backpass — improves your agents.md from its own mistakes: finds repeated failures across session transcripts, proposes small instruction changes with quoted evidence (rules need proof from 2+ sessions; nothing writes until you approve). agenttrail — shows what a coding agent is doing NOW: watches plans, tool calls, and file changes, mapped onto repo components with working/blocked/completed states. Claude Code gets rich run cards via hooks; other agents work via file watching. FrontierAgent — packages long-running research and file work into a terminal agent you can inspect: one stateful worker, a coordinator with a live taskboard, read-only inputs, diffs for every write, resumable/revertible sessions. rome — the bet that the environment matters more than the model: the unit isn’t a chat but an app — manifest + typed actions, app-owned agents, skills, hooks, a web UI, and its own database that survives the conversation. Describe something missing and it scaffolds it in as ordinary git-tracked source. sentio — gives every AI agent a real mailbox (structured authenticated events, replies in-thread) without stitching IMAP/SMTP/webhooks yourself; multi-tenant out of the box. halofy — one governance boundary between company knowledge and the agents using it: credentials resolve identity/role/namespace server-side so agents can’t claim broader access in an MCP request; reads/writes carry provenance, deletion emits a signed eraser certificate. spec-ptc — speculative programmatic tool calling: launches the slow sub-LLM calls inside generated code as futures WHILE the code streams, instead of after the harness finishes running it. doop — multiplayer design canvas where humans and AI agents edit the same work live: sandboxed iframes, cursor/comment streaming, built-in MCP server, self-hostable. Desktop & developer experience buzz — self-hosted team workspace where humans and AI agents join the same rooms with separate identities and audit trails; messages, approvals, and git activity become signed Nostr events in one searchable log. walgit — a git server in one binary backed by S3 or GCS, no database or elected leader: the bucket IS the repository, each server’s disk is just a disposable cache, compare-and-swap decides which push wins. monocode — several coding-agent CLIs (Claude Code, Codex, Cursor, FX) behind one desktop interface without becoming another model provider; each tool keeps its own account. rataflow — React Flow-style node editors in a Rust TUI: automatic layout, dragging, pan/zoom, themes, savable graph snapshots, semantic events instead of hidden state. system-atlas — architecture discussions become two synchronized views from one data file: an interactive isometric map for people and a generated system.md for the repo. vault-graph — turns an Obsidian vault into a deterministic disc layout instead of a forced-directed hairball; works as a plugin or exports an offline HTML file. On-device & local AI CarWatch — a Raspberry Pi 5 in the car running a 35B-parameter model locally (~3.5 tok/s) that answers from the car’s own 745-page owner’s manual with page citations, voice on Pi, no wake word; includes an honest status table (OBD reader marked unverified against the real vehicle). better-voice — macOS dictation that takes the screen with it: hold option and talk, circle anything on screen, the transcript lands in your text field with those screenshots after it. Runs locally on Parakeet with an opt-in grammar model. hayamimi — realtime speech-to-text on CPU, nothing sent to the cloud: detects each utterance’s language in the first seconds and routes to a specialist model (Japanese/Chinese/English), with a 1,600-language fallback behind them, all quantized via sherpa-onnx. kimodo.cpp — NVIDIA’s Komodo text-to-motion model in a native C++/GGML runtime: SNPLX rotations and root movement locally on CPU or Vulkan, with a browser demo. ambient-context — reads the text of whatever window you’re focused on (macOS accessibility tree) into a per-day markdown file: no screenshots, no network; password managers/private browsing excluded, capture folder excludes itself. ocr-it — browser extension that pins one screen region and OCRs that exact rectangle on every hotkey, appending to a running transcript; Tesseract bundled, fully offline, minimal permissions. Media, learning & misc workout-guide — 302 exercise illustrations for apps that need consistent movement visuals: three matching frames each, searchable metadata, typed lookup helpers, framework-agnostic SVG. learn — the author’s own Pi configuration for learning with AI, published as-is: a teach skill holds the philosophy, a visualize skill adds diagrams only when clearer, subagents split off research. AI-Engineering-Lab — free self-paced 24-week course from Python basics to deployed AI systems; every week extends the same fictional freight company, and from week three every artifact needs a metric. open-sheet — lets coding agents build spreadsheets as React models instead of juggling cell addresses: name the fields, the framework resolves final Excel coordinates; live formulas export to XLSX. LinearAbilityExtThreeJS — a 3JS VFX workbench for designing game abilities while they’re still moving: seven casts share five targeting shapes, effects built from procedural geometry and GLSL instead of sprite sheets; pause mid-eruption and every slider stays live.

The Next New Thing’s weekly Top 10 GitHub roundup (hosts Adam and Andrew, 27:22). The crop skews agent-tooling: diagram design systems, memory layers, model routers, and one genuinely huge star count. The top 10 Flow-chart design system for Claude (Catherine) — #1 this week, 24k stars. A process (with seven “dials”) that turns Claude’s default samey-looking diagrams into clear, well-designed flow charts — sketch/terminal/paper styles, built-in templates, and it can match your own site’s design. OpenViking — agent memory management that looks like a filesystem: a “phantom drive” Claude or Codex reads and writes directly, instead of a database layer. The hosts debate whether this is a step backward from vector DBs or the best of both worlds. (This week’s hosts use Hermes Agent — and yes, that’s the same OpenViking this blog runs on.) Omari v4 (DHH) — the opinionated, Mac-like Linux distro hit version four: Wi-Fi QR-code sharing, keyboard-shortcut comfort for Mac refugees, and a pitch for the $200-300 mini boxes replacing doubled-price Mac minis as always-on AI machines. Needle — an AI model so small it fits on a device: Raspberry Pi, smart watch, Home Assistant, mobile apps. On-device, free, fast — the “poke at one thing and do it really well” class of model. Semantica — “open-source Palantir for AI agents”: takes messy company data, builds a knowledge graph your agent can reason over. The hosts’ caveat: big systems need real tuning/structuring work to beat plain search. NVIDIA Switchyard — a model router with escalation: every request starts at the cheapest model (even a free local one), gets evaluated, and escalates up the ladder (Haiku → Sonnet → Opus) only when needed. The token-bill insurance for product builders. Mojo compiler, fully open source — 3 days before the video. Python-like syntax with C/C++ performance, built for the AI era; includes benchmarks for how well AIs understand Mojo. Money Printer Turbo — type a topic, get a finished video: pulls in existing footage or stock, adds ElevenLabs voices, works with AI-generated content too. The hosts joke the name undersells it. public-apis — 467k stars: a collection of free APIs (weather, financial data, everything). The point isn’t that Claude can’t find APIs — it’s that reading the pros/cons list lets you guide it better. Holehe — checks whether an email is registered on 120+ sites via password-recovery probing. Legit uses (username squatting, customer validation) but the hosts find it mostly sketchy. Community builds The hosts feature listener projects, including: Shockwave (local markdown notes your agent can work on), Lumina (local-first desktop agent with multi-tier memory — “a local private OpenClaw”), Imagine CLI (20 images in ~a minute from the terminal), a TLDR-newsletter-to-audio tool for walks, Oracle (a coding-agent framework, OpenCode-like), a Minto pyramid skill (write like a consultant: answer first, then reasons, then evidence), and a persona-consultation tool (Hermosi, Jobs, Buffett, Bezos, and your future self on big decisions). ...

Sean Kochel’s roundup of five relatively new GitHub repos for vibe coding — half of them under 10,000 stars. He demos each one on his own AI food-logging app and shows how they slot into a workflow. 27:30. draw.io skill — architecture diagrams on demand (~0:26) A Claude Code skill that uses the draw.io CLI to generate editable architecture diagrams of your repo from a natural-language request. Say “visualize the layers of the services in my repo” — it explores the codebase and draws the presentation layer, front-end state, service layer, database layer, and edge functions (PostHog, OpenAI, etc.). The point isn’t the diagram — it’s understanding what you built. When something breaks, you can point the model at the resolver layer instead of burning tokens on a blind “go fix this.” His framing: you can only push these models as far as your expertise goes, so use the tools as learning loops. Ponytail — the anti-over-engineering plugin (~7:59) Makes your AI agent think like the long-ponytail senior dev who replaces your 50 lines with one. ...