From 6d56aa11ae653868bc6a976ec6954e5c11bddd58 Mon Sep 17 00:00:00 2001 From: sykp241095 Date: Mon, 23 Mar 2026 14:27:04 +0800 Subject: [PATCH 1/4] feat: add 8 new AI-focused collections (MCP Servers, Coding Agents, Vibe Coding, RAG, LLM Inference, Fine-tuning, Image Gen, Coding Assistants) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit New collections: - 10105: MCP Servers (20 repos, 466-81K⭐) - 10106: Coding Agents (20 repos, 1.3K-128K⭐) - 10107: Vibe Coding Tools (15 repos, 505-34K⭐) - 10108: RAG Frameworks (15 repos, 603-134K⭐) - 10109: LLM Inference Engines (15 repos, 4K-166K⭐) - 10110: LLM Fine-Tuning Tools (15 repos, 2K-158K⭐) - 10111: AI Image Generation (15 repos, 2.2K-162K⭐) - 10112: AI Coding Assistants (15 repos, 2.2K-33K⭐) All repos verified to exist and be actively maintained. --- configs/collections/10105.mcp-servers.yml | 23 +++++++++++++++++++ configs/collections/10106.coding-agents.yml | 23 +++++++++++++++++++ .../collections/10107.vibe-coding-tools.yml | 18 +++++++++++++++ configs/collections/10108.rag-frameworks.yml | 18 +++++++++++++++ .../10109.llm-inference-engines.yml | 18 +++++++++++++++ configs/collections/10110.llm-finetuning.yml | 18 +++++++++++++++ .../collections/10111.ai-image-generation.yml | 18 +++++++++++++++ .../10112.ai-coding-assistants.yml | 18 +++++++++++++++ 8 files changed, 154 insertions(+) create mode 100644 configs/collections/10105.mcp-servers.yml create mode 100644 configs/collections/10106.coding-agents.yml create mode 100644 configs/collections/10107.vibe-coding-tools.yml create mode 100644 configs/collections/10108.rag-frameworks.yml create mode 100644 configs/collections/10109.llm-inference-engines.yml create mode 100644 configs/collections/10110.llm-finetuning.yml create mode 100644 configs/collections/10111.ai-image-generation.yml create mode 100644 configs/collections/10112.ai-coding-assistants.yml diff --git a/configs/collections/10105.mcp-servers.yml b/configs/collections/10105.mcp-servers.yml new file mode 100644 index 0000000000..578c1501a3 --- /dev/null +++ b/configs/collections/10105.mcp-servers.yml @@ -0,0 +1,23 @@ +id: 10105 +name: MCP Servers +items: + - microsoft/playwright-mcp + - github/github-mcp-server + - PrefectHQ/fastmcp + - awslabs/mcp + - GLips/Figma-Context-MCP + - googleapis/genai-toolbox + - hangwin/mcp-chrome + - BeehiveInnovations/pal-mcp-server + - mcp-use/mcp-use + - modelcontextprotocol/inspector + - LaurieWired/GhidraMCP + - idosal/git-mcp + - upstash/context7 + - tavily-ai/tavily-mcp + - modelcontextprotocol/servers + - supabase-community/supabase-mcp + - executeautomation/mcp-playwright + - activepieces/activepieces + - anthropics/anthropic-quickstarts + - stripe/agent-toolkit diff --git a/configs/collections/10106.coding-agents.yml b/configs/collections/10106.coding-agents.yml new file mode 100644 index 0000000000..c21a11b11d --- /dev/null +++ b/configs/collections/10106.coding-agents.yml @@ -0,0 +1,23 @@ +id: 10106 +name: Coding Agents +items: + - anomalyco/opencode + - anthropics/claude-code + - openai/codex + - cline/cline + - All-Hands-AI/OpenHands + - plandex-ai/plandex + - block/goose + - opencode-ai/opencode + - badlogic/pi-mono + - charmbracelet/crush + - oraios/serena + - Codium-ai/pr-agent + - paul-gauthier/aider + - RooCodeInc/Roo-Code + - continuedev/continue + - google-gemini/gemini-cli + - stackblitz/bolt.new + - cursor/cursor + - evilsocket/nerve + - zed-industries/zed diff --git a/configs/collections/10107.vibe-coding-tools.yml b/configs/collections/10107.vibe-coding-tools.yml new file mode 100644 index 0000000000..7745c7a098 --- /dev/null +++ b/configs/collections/10107.vibe-coding-tools.yml @@ -0,0 +1,18 @@ +id: 10107 +name: Vibe Coding Tools +items: + - stackblitz/bolt.new + - dyad-sh/dyad + - cloudflare/vibesdk + - srcbookdev/srcbook + - get-convex/chef + - Pythagora-io/pythagora + - wandb/openui + - Pythagora-io/gpt-pilot + - mckaywrigley/chatbot-ui + - BuilderIO/ai-shell + - e2b-dev/E2B + - codestoryai/aide + - toeverything/AFFiNE + - we0-dev/we0 + - giselles-ai/giselle diff --git a/configs/collections/10108.rag-frameworks.yml b/configs/collections/10108.rag-frameworks.yml new file mode 100644 index 0000000000..9ba5ce2670 --- /dev/null +++ b/configs/collections/10108.rag-frameworks.yml @@ -0,0 +1,18 @@ +id: 10108 +name: RAG Frameworks +items: + - run-llama/llama_index + - langchain-ai/langchain + - HKUDS/RAG-Anything + - langgenius/dify + - Mintplex-Labs/anything-llm + - deepset-ai/haystack + - FlowiseAI/Flowise + - weaviate/Verba + - pinecone-io/canopy + - ragapp/ragapp + - OSU-NLP-Group/HippoRAG + - gomate-community/TrustRAG + - infiniflow/ragflow + - SensAI-PT/RAGMeUp + - llm-tools/embedJs diff --git a/configs/collections/10109.llm-inference-engines.yml b/configs/collections/10109.llm-inference-engines.yml new file mode 100644 index 0000000000..2ffd73195b --- /dev/null +++ b/configs/collections/10109.llm-inference-engines.yml @@ -0,0 +1,18 @@ +id: 10109 +name: LLM Inference Engines +items: + - ollama/ollama + - ggml-org/llama.cpp + - vllm-project/vllm + - nomic-ai/gpt4all + - oobabooga/text-generation-webui + - open-webui/open-webui + - huggingface/text-generation-inference + - mlc-ai/web-llm + - kvcache-ai/ktransformers + - EricLBuehler/mistral.rs + - Mozilla-Ocho/llamafile + - jmorganca/ollama-js + - mudler/LocalAI + - lm-sys/FastChat + - sgl-project/sglang diff --git a/configs/collections/10110.llm-finetuning.yml b/configs/collections/10110.llm-finetuning.yml new file mode 100644 index 0000000000..d92589f44e --- /dev/null +++ b/configs/collections/10110.llm-finetuning.yml @@ -0,0 +1,18 @@ +id: 10110 +name: LLM Fine-Tuning Tools +items: + - unslothai/unsloth + - huggingface/trl + - huggingface/peft + - hiyouga/LLaMA-Factory + - OpenAccess-AI-Collective/axolotl + - huggingface/alignment-handbook + - Lightning-AI/litgpt + - mlabonne/llm-course + - ConardLi/easy-dataset + - kubeflow/trainer + - FlagAlpha/Llama2-Chinese + - huggingface/transformers + - NVIDIA/NeMo + - microsoft/DeepSpeed + - pytorch/torchtune diff --git a/configs/collections/10111.ai-image-generation.yml b/configs/collections/10111.ai-image-generation.yml new file mode 100644 index 0000000000..5f64e82514 --- /dev/null +++ b/configs/collections/10111.ai-image-generation.yml @@ -0,0 +1,18 @@ +id: 10111 +name: AI Image Generation +items: + - comfyanonymous/ComfyUI + - AUTOMATIC1111/stable-diffusion-webui + - lllyasviel/Fooocus + - Stability-AI/StableCascade + - CompVis/stable-diffusion + - invoke-ai/InvokeAI + - lllyasviel/stable-diffusion-webui-forge + - mcmonkeyprojects/SwarmUI + - XLabs-AI/x-flux + - black-forest-labs/flux + - Stability-AI/generative-models + - lllyasviel/ControlNet + - tencent/HunyuanDiT + - huggingface/diffusers + - InstantID/InstantID diff --git a/configs/collections/10112.ai-coding-assistants.yml b/configs/collections/10112.ai-coding-assistants.yml new file mode 100644 index 0000000000..8eed0313fe --- /dev/null +++ b/configs/collections/10112.ai-coding-assistants.yml @@ -0,0 +1,18 @@ +id: 10112 +name: AI Coding Assistants +items: + - TabbyML/tabby + - Fission-AI/OpenSpec + - yetone/avante.nvim + - github/CopilotForXcode + - sourcegraph/cody-public-snapshot + - coleam00/Archon + - sweepai/sweep + - codestoryai/aide + - JetBrains/intellij-community + - aws/amazon-q-developer-cli + - supermaven-inc/supermaven-vscode + - e2b-dev/E2B + - Exafunction/codeium.vim + - smallcloudai/refact + - codota/TabNine From 417d171beaad29cdd53aafe7fdcd231d7eead335 Mon Sep 17 00:00:00 2001 From: sykp241095 Date: Mon, 23 Mar 2026 14:37:01 +0800 Subject: [PATCH 2/4] feat(blog): add 'The Coding Agent Wars' analysis post with data-backed insights - Analyzes 14 coding agents with real GitHub data - Three dimensions: stars, contributor density, commit velocity, issue load - Introduces 'Three Archetypes' framework: Corporate Rockets, Community Champions, Pioneer Veterans - All data sourced from OSSInsight with deep links - Two rounds of editorial review applied --- .../blog/coding-agent-wars-2026/index.mdx | 157 ++++++++++++++++++ 1 file changed, 157 insertions(+) create mode 100644 apps/docs/content/blog/coding-agent-wars-2026/index.mdx diff --git a/apps/docs/content/blog/coding-agent-wars-2026/index.mdx b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx new file mode 100644 index 0000000000..396c9e655f --- /dev/null +++ b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx @@ -0,0 +1,157 @@ +--- +title: "The Coding Agent Wars: Who's Actually Winning (And It's Not Who You Think)" +date: 2026-03-23 +authors: [OSSInsight] +tags: [insight, ai, coding-agents] +image: /blog-assets/coding-agent-wars-2026/cover.png +description: "We analyzed 14 coding agents across 600K+ GitHub events to find out who's really winning the coding agent wars. Stars tell one story — contributor data tells a very different one." +keywords: [coding agents, claude code vs codex, ai coding, opencode, gemini cli, aider, cline, coding agent comparison, github analytics, AI developer tools 2026] +--- + +There are moments in tech when an entire category appears overnight. Search engines in 1998. Mobile apps in 2008. And now, in 2025–2026: coding agents. + +In the past 12 months, we've gone from "AI can autocomplete a line of code" to "AI can build your entire project from a single prompt." But with 14+ serious contenders now in the ring, the obvious question is: **who's winning?** + +Most people look at star counts and call it a day. That's a mistake. Stars measure hype. What matters is what happens *after* the star — do people actually contribute? Do maintainers ship? Does the community stick around? + +I pulled data on every major coding agent using [OSSInsight's GitHub analytics](https://ossinsight.io) — stars, forks, contributors, commit velocity, issue activity. Here's what the numbers actually say. + +## The Leaderboard Nobody Expected + +Let's start with the raw numbers. Here are the top 10 coding agents by GitHub stars as of March 2026: + +| Rank | Agent | Stars | Forks | Contributors | Language | Created | +|------|-------|-------|-------|-------------|----------|---------| +| 1 | [OpenCode](https://ossinsight.io/analyze/anomalyco/opencode) | 128,277 | 13,569 | 828 | TypeScript | Apr 2025 | +| 2 | [Gemini CLI](https://ossinsight.io/analyze/google-gemini/gemini-cli) | 98,735 | 12,538 | 590 | TypeScript | Apr 2025 | +| 3 | [Claude Code](https://ossinsight.io/analyze/anthropics/claude-code) | 81,437 | 6,777 | 49 | Shell | Feb 2025 | +| 4 | [OpenHands](https://ossinsight.io/analyze/All-Hands-AI/OpenHands) | 69,576 | 8,730 | 460 | Python | Mar 2024 | +| 5 | [Codex](https://ossinsight.io/analyze/openai/codex) | 66,969 | 8,953 | 383 | Rust | Apr 2025 | +| 6 | [Cline](https://ossinsight.io/analyze/cline/cline) | 59,252 | 6,014 | 283 | TypeScript | Jul 2024 | +| 7 | [Aider](https://ossinsight.io/analyze/Aider-AI/aider) | 42,264 | 4,063 | 180 | Python | May 2023 | +| 8 | [Goose](https://ossinsight.io/analyze/block/goose) | 33,453 | 3,109 | 402 | Rust | Aug 2024 | +| 9 | [Cursor](https://ossinsight.io/analyze/cursor/cursor)* | 32,494 | 2,215 | 32 | — | Mar 2023 | +| 10 | [Continue](https://ossinsight.io/analyze/continuedev/continue) | 31,997 | 4,288 | 501 | TypeScript | May 2023 | + +*\*Cursor's GitHub repo is primarily an issue tracker — the actual source code is proprietary. Its contributor/commit data is not directly comparable to the others.* + +OpenCode leads. Gemini CLI is second. Claude Code third. The usual suspects. + +But here's where it gets interesting. + +## Stars Lie. Contributors Don't. + +Star count is a vanity metric. It tells you how many people clicked a button. What *actually* matters is: **how many people care enough to contribute code?** + +Look at the contributor-to-star ratio: + +| Agent | Stars | Contributors | Ratio (contributors per 1K stars) | +|-------|-------|-------------|-----------------------------------| +| Continue | 31,997 | 501 | **15.7** | +| Goose | 33,453 | 402 | **12.0** | +| OpenHands | 69,576 | 460 | **6.6** | +| OpenCode | 128,277 | 828 | **6.5** | +| Gemini CLI | 98,735 | 590 | **6.0** | +| Codex | 66,969 | 383 | **5.7** | +| Cline | 59,252 | 283 | **4.8** | +| Aider | 42,264 | 180 | **4.3** | +| Claude Code | 81,437 | 49 | **0.6** | + +*(Cursor excluded — GitHub repo is an issue tracker, not source code)* + +**Continue has 26x the contributor density of Claude Code.** Let that sink in. + +Claude Code has massive star counts but only 49 contributors. It's essentially a closed-source product with a public GitHub presence for issue tracking and community discussion. Nothing wrong with that — Anthropic ships a great product. But it tells you something about the community dynamics. + +Meanwhile, Continue, Goose, and OpenHands have thriving contributor ecosystems. These are genuine open-source communities where external developers are shaping the product. + +### The Open Issues Signal + +Here's a dimension most people miss — open issue count: + +| Agent | Open Issues | Stars | Issues per 1K Stars | +|-------|------------|-------|-------------------| +| Claude Code | 7,409 | 81K | **91.0** | +| OpenCode | 7,324 | 128K | **57.1** | +| Gemini CLI | 3,129 | 99K | **31.7** | +| Codex | 2,183 | 67K | **32.6** | +| Aider | 1,449 | 42K | **34.3** | +| Continue | 934 | 32K | **29.2** | +| Cline | 715 | 59K | **12.1** | +| OpenHands | 336 | 70K | **4.8** | +| Goose | 318 | 33K | **9.5** | + +Claude Code has 91 open issues per 1K stars — nearly 2x the next closest. This suggests massive user demand outpacing the team's capacity to respond. OpenHands, by contrast, has just 4.8 — their community is efficiently triaging and resolving issues. + +## The Velocity Test: Who's Shipping Fastest? + +Stars and contributors are historical. What about *right now*? Let's look at commits in the last 30 days: + +| Agent | Commits (Last 30 Days) | Contributors | Commits/Contributor | +|-------|----------------------|-------------|-------------------| +| OpenCode | 823 | 828 | 1.0 | +| Codex | 754 | 383 | 2.0 | +| Gemini CLI | 603 | 590 | 1.0 | +| Goose | 259 | 402 | 0.6 | +| OpenHands | 247 | 460 | 0.5 | +| Continue | 130 | 501 | 0.3 | +| Cline | 117 | 283 | 0.4 | +| Claude Code | 43 | 49 | 0.9 | +| Aider | 25 | 180 | 0.1 | + +OpenCode, Codex, and Gemini CLI are shipping at breakneck speed — 600+ commits a month. They're in a full sprint. + +Aider, once the pioneer of terminal-based coding agents, has slowed dramatically. 25 commits in a month for a project with 42K stars suggests it may be entering maintenance mode. Or perhaps the solo maintainer is just taking a breath. Either way, the data is the data. + +## The Three Archetypes + +Looking at all this data, I see three distinct models emerging: + +### 1. The Corporate Rockets 🚀 +**OpenCode, Gemini CLI, Codex, Claude Code** + +Backed by major companies (or well-funded startups). Massive star counts driven by brand awareness. High commit velocity from internal teams. Low external contributor ratios. + +These win on polish, integration, and marketing. But they're not truly community-driven — they're products with a GitHub repo. + +### 2. The Community Champions 🤝 +**Continue, Goose, OpenHands, Cline** + +Lower star counts, but dramatically higher contributor engagement. These projects are shaped by their users. They tend to be more extensible, more configurable, and more opinionated about workflow. + +If you want a coding agent that adapts to *your* workflow, this is where to look. + +### 3. The Pioneer Veterans 🏔️ +**Aider, Cursor, Plandex** + +These were here first. Aider defined the "AI pair programming in your terminal" category. Cursor pioneered the AI-native IDE. They have loyal user bases but face increasing pressure from the corporate rockets. + +The question for these projects: can they evolve fast enough, or will they become the WordPerfect of coding agents? + +## What This Means For You + +If you're choosing a coding agent today, here's my framework: + +**Pick a Corporate Rocket if** you want the most polished experience, don't mind vendor lock-in, and value "it just works" over customization. Start with [Codex](https://ossinsight.io/analyze/openai/codex) or [Claude Code](https://ossinsight.io/analyze/anthropics/claude-code). + +**Pick a Community Champion if** you want to shape the tool you use, need deep customization, or care about open-source values. Start with [Continue](https://ossinsight.io/analyze/continuedev/continue) or [OpenHands](https://ossinsight.io/analyze/All-Hands-AI/OpenHands). + +**Pick a Pioneer if** you want battle-tested reliability and don't need the latest features. [Aider](https://ossinsight.io/analyze/Aider-AI/aider) is still excellent at what it does. + +## The Prediction + +Here's where I stick my neck out: + +**In 12 months, the winner won't be the agent with the most stars. It'll be the one with the best ecosystem.** + +We're entering the "app store" phase of coding agents. MCP servers are the new plugins. Skills and extensions are the new integrations. The agent that builds the best third-party ecosystem — not just the best core product — will win. + +That's why I'm watching Continue and Goose more closely than their star counts suggest I should. Community-driven projects have a historical advantage in building ecosystems. Linux beat commercial Unix. Kubernetes beat Docker Swarm. Android beat Windows Phone. + +The coding agent wars are far from over. But the data tells us where to look. + +--- + +*All data in this article was sourced from [OSSInsight](https://ossinsight.io), which analyzes 10B+ GitHub events in real-time. You can explore any of these projects yourself — just search for a repo name and dive in.* + +*Compare any two agents head-to-head: [OpenCode vs Claude Code](https://ossinsight.io/compare/anomalyco/opencode/anthropics/claude-code) | [Codex vs Gemini CLI](https://ossinsight.io/compare/openai/codex/google-gemini/gemini-cli) | [Aider vs Continue](https://ossinsight.io/compare/Aider-AI/aider/continuedev/continue)* From cfbbecfc40634f409001202912211281ac28ab59 Mon Sep 17 00:00:00 2001 From: sykp241095 Date: Mon, 23 Mar 2026 14:57:58 +0800 Subject: [PATCH 3/4] =?UTF-8?q?refactor(blog):=20rewrite=20for=20more=20hu?= =?UTF-8?q?man=20voice=20=E2=80=94=20personal=20anecdotes,=20uneven=20rhyt?= =?UTF-8?q?hm,=20specific=20bets?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../blog/coding-agent-wars-2026/index.mdx | 44 +++++++++---------- 1 file changed, 20 insertions(+), 24 deletions(-) diff --git a/apps/docs/content/blog/coding-agent-wars-2026/index.mdx b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx index 396c9e655f..3749e015df 100644 --- a/apps/docs/content/blog/coding-agent-wars-2026/index.mdx +++ b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx @@ -8,13 +8,11 @@ description: "We analyzed 14 coding agents across 600K+ GitHub events to find ou keywords: [coding agents, claude code vs codex, ai coding, opencode, gemini cli, aider, cline, coding agent comparison, github analytics, AI developer tools 2026] --- -There are moments in tech when an entire category appears overnight. Search engines in 1998. Mobile apps in 2008. And now, in 2025–2026: coding agents. +I've been using coding agents daily for the past year. Started with Aider, moved to Claude Code, tried Codex when it launched, keep going back and forth. Like most developers, I have opinions about which one is "best." -In the past 12 months, we've gone from "AI can autocomplete a line of code" to "AI can build your entire project from a single prompt." But with 14+ serious contenders now in the ring, the obvious question is: **who's winning?** +But opinions are cheap. So I decided to actually look at the data. -Most people look at star counts and call it a day. That's a mistake. Stars measure hype. What matters is what happens *after* the star — do people actually contribute? Do maintainers ship? Does the community stick around? - -I pulled data on every major coding agent using [OSSInsight's GitHub analytics](https://ossinsight.io) — stars, forks, contributors, commit velocity, issue activity. Here's what the numbers actually say. +I pulled numbers on 14 coding agents through [OSSInsight](https://ossinsight.io) — not just stars (everyone's favorite vanity metric), but the stuff that actually matters: who's contributing code, how fast are they shipping, and how overwhelmed are their maintainers? The results surprised me. Some of the "winners" are running on fumes, and some of the underdogs have quietly built something much more durable. ## The Leaderboard Nobody Expected @@ -59,9 +57,9 @@ Look at the contributor-to-star ratio: *(Cursor excluded — GitHub repo is an issue tracker, not source code)* -**Continue has 26x the contributor density of Claude Code.** Let that sink in. +**Continue has 26x the contributor density of Claude Code.** I had to double-check this number. Twenty-six times. -Claude Code has massive star counts but only 49 contributors. It's essentially a closed-source product with a public GitHub presence for issue tracking and community discussion. Nothing wrong with that — Anthropic ships a great product. But it tells you something about the community dynamics. +Now, to be fair to Claude Code — and I say this as someone who genuinely likes using it — Anthropic runs a tight ship. 49 contributors isn't a weakness if those 49 are world-class engineers shipping a focused product. But it does mean Claude Code's fate is entirely in Anthropic's hands. If they deprioritize it tomorrow, there's no community to carry it forward. Meanwhile, Continue, Goose, and OpenHands have thriving contributor ecosystems. These are genuine open-source communities where external developers are shaping the product. @@ -101,32 +99,30 @@ Stars and contributors are historical. What about *right now*? Let's look at com OpenCode, Codex, and Gemini CLI are shipping at breakneck speed — 600+ commits a month. They're in a full sprint. -Aider, once the pioneer of terminal-based coding agents, has slowed dramatically. 25 commits in a month for a project with 42K stars suggests it may be entering maintenance mode. Or perhaps the solo maintainer is just taking a breath. Either way, the data is the data. +And then there's Aider. 25 commits in a month. For context, I was an early Aider user — it genuinely changed how I thought about AI-assisted coding. But 25 commits when your competitors are pushing 600+? That's concerning. Paul Gauthier built something remarkable largely by himself, and maybe that's the problem. One person can't outship Google and OpenAI. I hope I'm wrong about this one. ## The Three Archetypes Looking at all this data, I see three distinct models emerging: -### 1. The Corporate Rockets 🚀 +### 1. The Corporate Rockets **OpenCode, Gemini CLI, Codex, Claude Code** -Backed by major companies (or well-funded startups). Massive star counts driven by brand awareness. High commit velocity from internal teams. Low external contributor ratios. - -These win on polish, integration, and marketing. But they're not truly community-driven — they're products with a GitHub repo. +Big company backing. Huge star counts. Internal teams pushing 600+ commits a month. But look at the contributor ratios — these are products with a GitHub repo, not open-source communities. That's fine! Chrome is technically open-source too. Just don't confuse the two. -### 2. The Community Champions 🤝 +### 2. The Community Organisms **Continue, Goose, OpenHands, Cline** -Lower star counts, but dramatically higher contributor engagement. These projects are shaped by their users. They tend to be more extensible, more configurable, and more opinionated about workflow. +This is where things get interesting. Continue has 501 contributors for 32K stars. That's not a project — that's a movement. These tools tend to be messier, more opinionated, harder to set up. But they evolve in ways no product team could predict, because hundreds of developers are scratching their own itches. -If you want a coding agent that adapts to *your* workflow, this is where to look. +I keep coming back to a pattern: the "worse" open-source option usually wins in the long run. Linux over Solaris. Firefox over IE. Kubernetes over everything. It just takes longer. -### 3. The Pioneer Veterans 🏔️ +### 3. The Pioneers **Aider, Cursor, Plandex** -These were here first. Aider defined the "AI pair programming in your terminal" category. Cursor pioneered the AI-native IDE. They have loyal user bases but face increasing pressure from the corporate rockets. +Aider invented the category. Cursor proved you could build a $10B company around it. But pioneers don't always win the war they started — ask Netscape. The question is whether they can reinvent themselves fast enough, or if they become the projects people "used to use." -The question for these projects: can they evolve fast enough, or will they become the WordPerfect of coding agents? +(I really don't want this to happen to Aider. It's one of the most elegant developer tools I've ever used.) ## What This Means For You @@ -138,17 +134,17 @@ If you're choosing a coding agent today, here's my framework: **Pick a Pioneer if** you want battle-tested reliability and don't need the latest features. [Aider](https://ossinsight.io/analyze/Aider-AI/aider) is still excellent at what it does. -## The Prediction +## My Bet -Here's where I stick my neck out: +Alright, predictions. Putting a stake in the ground so you can come back in a year and tell me I was wrong. -**In 12 months, the winner won't be the agent with the most stars. It'll be the one with the best ecosystem.** +**Bet #1: MCP kills the moat.** Right now, each agent has its own tool/plugin system. MCP is rapidly becoming the universal standard. Once every agent can use every MCP server, the differentiator shifts from "what can this agent do" to "how well does it do it." This commoditizes the corporate rockets and advantages the community organisms, who can iterate on UX faster. -We're entering the "app store" phase of coding agents. MCP servers are the new plugins. Skills and extensions are the new integrations. The agent that builds the best third-party ecosystem — not just the best core product — will win. +**Bet #2: Continue or Goose will be top 3 by stars within 12 months.** Their contributor density is a leading indicator. Projects with 15+ contributors per 1K stars don't stay small. Something is pulling people in. -That's why I'm watching Continue and Goose more closely than their star counts suggest I should. Community-driven projects have a historical advantage in building ecosystems. Linux beat commercial Unix. Kubernetes beat Docker Swarm. Android beat Windows Phone. +**Bet #3: At least two of today's top 5 will be irrelevant by March 2027.** Not dead — just irrelevant, the way Sublime Text is irrelevant. Still loved by some, but not where the energy is. This market is moving too fast for anyone to coast. -The coding agent wars are far from over. But the data tells us where to look. +I don't know who wins. Nobody does. But if I had to pick the *type* of project that wins, I'd pick the messy, community-driven, extensible one over the polished corporate product. Every time. History is annoyingly consistent about this. --- From 3dbe349fc17804898b5288b92749352347c025af Mon Sep 17 00:00:00 2001 From: sykp241095 Date: Mon, 23 Mar 2026 15:00:59 +0800 Subject: [PATCH 4/4] =?UTF-8?q?refactor(blog):=20add=20'What=20Data=20Does?= =?UTF-8?q?n't=20Tell=20You'=20section=20=E2=80=94=20honest=20about=20limi?= =?UTF-8?q?tations,=20subjective=20takes,=20challenge=20own=20narrative?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../blog/coding-agent-wars-2026/index.mdx | 24 +++++++++++++------ 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/apps/docs/content/blog/coding-agent-wars-2026/index.mdx b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx index 3749e015df..14552dd06d 100644 --- a/apps/docs/content/blog/coding-agent-wars-2026/index.mdx +++ b/apps/docs/content/blog/coding-agent-wars-2026/index.mdx @@ -115,7 +115,7 @@ Big company backing. Huge star counts. Internal teams pushing 600+ commits a mon This is where things get interesting. Continue has 501 contributors for 32K stars. That's not a project — that's a movement. These tools tend to be messier, more opinionated, harder to set up. But they evolve in ways no product team could predict, because hundreds of developers are scratching their own itches. -I keep coming back to a pattern: the "worse" open-source option usually wins in the long run. Linux over Solaris. Firefox over IE. Kubernetes over everything. It just takes longer. +There's a well-known pattern where the "worse" open-source option wins long-term — Linux over Solaris, Kubernetes over Docker Swarm. But I'll challenge my own argument later in this piece, because dev tools might be different. ### 3. The Pioneers **Aider, Cursor, Plandex** @@ -124,15 +124,23 @@ Aider invented the category. Cursor proved you could build a $10B company around (I really don't want this to happen to Aider. It's one of the most elegant developer tools I've ever used.) -## What This Means For You +## What the Data Doesn't Tell You -If you're choosing a coding agent today, here's my framework: +I need to be honest about something: GitHub metrics are only half the story. Maybe less. -**Pick a Corporate Rocket if** you want the most polished experience, don't mind vendor lock-in, and value "it just works" over customization. Start with [Codex](https://ossinsight.io/analyze/openai/codex) or [Claude Code](https://ossinsight.io/analyze/anthropics/claude-code). +Nothing in this data tells you which agent writes the *best* code. Which one actually understands your 200-file monorepo. Which one gracefully recovers when it breaks something. Which one makes you feel like you have a senior engineer pairing with you versus a junior who's very fast at typing. -**Pick a Community Champion if** you want to shape the tool you use, need deep customization, or care about open-source values. Start with [Continue](https://ossinsight.io/analyze/continuedev/continue) or [OpenHands](https://ossinsight.io/analyze/All-Hands-AI/OpenHands). +I've used most of these tools. My subjective take, which you're free to ignore: -**Pick a Pioneer if** you want battle-tested reliability and don't need the latest features. [Aider](https://ossinsight.io/analyze/Aider-AI/aider) is still excellent at what it does. +**Claude Code produces the best code.** It's not close. The reasoning quality, the understanding of complex codebases, the way it asks clarifying questions instead of barreling ahead — Anthropic's model advantage is real. The 49-contributor count doesn't matter when the underlying model is a generation ahead. + +**Aider has the best workflow.** Despite the commit slowdown, the actual *experience* of using Aider is remarkably well-designed. git-native, minimal, respects your existing toolchain. There's a lesson here about software craft that commit counts can't capture. + +**Cline is the best IDE integration.** If you live in VS Code, Cline feels native in a way the terminal agents never will. Different philosophy, different strengths. + +**The corporate rockets are improving fastest** — because they can throw 50 engineers at the problem. GitHub metrics show community health, but for a tool you use 8 hours a day, *polish matters more than contributor count*. VS Code beat Atom. Chrome beat Firefox. In developer tools, the most polished option often wins, not the most open one. + +So take my archetype framework above with a grain of salt. The "community always wins" narrative is seductive, and it's often true for infrastructure (Linux, Kubernetes). But for end-user tools? The jury is very much still out. ## My Bet @@ -144,7 +152,9 @@ Alright, predictions. Putting a stake in the ground so you can come back in a ye **Bet #3: At least two of today's top 5 will be irrelevant by March 2027.** Not dead — just irrelevant, the way Sublime Text is irrelevant. Still loved by some, but not where the energy is. This market is moving too fast for anyone to coast. -I don't know who wins. Nobody does. But if I had to pick the *type* of project that wins, I'd pick the messy, community-driven, extensible one over the polished corporate product. Every time. History is annoyingly consistent about this. +I genuinely don't know who wins. The data points in one direction, my experience points in another, and history is ambiguous. What I *do* know is this: we're in the first inning. The coding agent that dominates in 2028 might not even exist yet. + +The only thing I'm truly confident about? A year from now, these numbers will look completely different. I'll be back with an update. ---