01. Analysis
Star counts tell you where a project has been. Star-growth velocity tells you where it's going.
We've been tracking the same cohort of five open-source AI productivity tools since early May 2026 — five projects that together account for 120,651 GitHub stars as of today, June 10, 2026. Over the past four weeks, something interesting happened: two of the five accelerated sharply while the other three either held steady or slowed. The tools gaining momentum aren't the ones dominating the editorial conversation. The tools losing it include one that's been declared the default choice so many times it's started to feel inevitable.
This is what the velocity data actually says, and what it means for which tool deserves your setup time right now.
How we researched this
Our automated research pipeline queries GitHub for AI productivity repositories and logs star counts, descriptions, and last-commit timestamps. We've now run this pipeline five times between May 8 and June 10, 2026, which gives us longitudinal data on the same cohort across a five-week window — something our earlier articles couldn't offer because they each captured a single point in time.
For growth calculations, I used the earliest published star count from our archives for each tool as the baseline (May 8–22, depending on when each tool first appeared) and the June 10, 2026 figures from today's research run as the endpoint. I then divided the net gain by the number of days elapsed to get a daily average. This is a rough measure — GitHub star counts are noisy, viral posts skew single-day figures, and a tool with a stable user base accretes stars slowly but durably. Take the velocity numbers as directional signals, not precise metrics.
Our pipeline also attempted to pull community data from Reddit, Hacker News, and official pricing pages for commercial alternatives (Reclaim AI, Motion, Otter.ai, Superhuman). Reddit and HN returned empty on this run, as they have on all five of our research dates — the rate-limiting issue we've noted before. Commercial tool pricing was also unavailable. This analysis rests entirely on GitHub data and official repository documentation.
All star counts below are as of June 10, 2026.
The Five Tools, Ranked by Growth Velocity
| Tool | Stars (June 10) | Stars (first observed) | Days tracked | Daily avg gain |
|---|---|---|---|---|
| next-ai-draw-io | 31,741 | 30,630 (May 22) | 19 | +58.5/day |
| Personal AI Infrastructure | 15,679 | 14,561 (May 22) | 19 | +58.8/day |
| Rowboat | 14,940 | 14,830 (May 15) | 26 | +4.2/day |
| Khoj | 35,046 | 34,795 (May 15) | 26 | +9.7/day |
| Plotly Dash | 24,245 | 24,225 (May 22) | 19 | +1.1/day |
The story in that table is blunt: next-ai-draw-io and Personal AI Infrastructure are growing at roughly the same pace — about 58–59 stars per day — while the three tools that have received the most editorial coverage here and elsewhere (Khoj, Rowboat, Plotly Dash) are adding stars at a fraction of that rate.
Khoj, at 9.7 stars per day, is the best-performing of the three slow growers. It has almost 2.5x the total star count of next-ai-draw-io's competitors but is growing at one-sixth the velocity. That pattern — large installed base, decelerating new interest — is characteristic of a tool that's become the safe default choice rather than a live discovery.
Rowboat is the most cautionary number in the table. It has received as much direct editorial attention as Khoj over the last six weeks and has roughly one-third the total stars — but is growing at less than half Khoj's already-modest daily rate. That either means Rowboat's actual user conversion rate is lower than the press coverage implies, or its audience has already been reached and it's not attracting new users outside that early cohort.
The two accelerating tools
next-ai-draw-io — 31,741 stars
github.com/DayuanJiang/next-ai-draw-io
The description is accurate and narrow: "A Next.js web application that integrates AI capabilities with draw.io diagrams. This app allows you to create, modify, and enhance diagrams through natural language commands and AI-assisted visualization."
This is not a general AI productivity platform. It is a single-purpose tool that fixes one specific friction point: the gap between "I know what I want to diagram" and "I know how to use draw.io." Natural language commands replace manual shape-dragging; you describe a flow or architecture and the AI builds or modifies the diagram.
The 58.5 stars per day says this is landing. The reason, I think, is that the productivity gain is obvious and verifiable in under ten minutes. You either type a description and get a reasonable diagram, or you don't. There's no multi-week onboarding, no LLM selection, no RAG configuration. The value is immediate or it isn't, and the star velocity suggests it mostly is.
The limitation is that scope: next-ai-draw-io is a diagram tool. It doesn't store your notes, manage your calendar, handle your email, or build agents. If diagramming is a recurring bottleneck for you — architecture reviews, process documentation, technical writing — it's worth installing. If you diagram once a quarter, the ROI isn't there.
Personal AI Infrastructure — 15,679 stars
github.com/danielmiessler/Personal_AI_Infrastructure
The description: "Agentic AI Infrastructure for magnifying HUMAN capabilities."
danielmiessler's project is the most conceptually ambitious tool in this cohort and the hardest to evaluate as a single "tool," because it's not really one. Personal AI Infrastructure is a framework and opinionated architecture for how a person's entire AI stack should be structured — which tools connect where, how agents interoperate, what data flows between them, and how to avoid the common failure mode of having five AI tools that don't know about each other.
The 58.8 stars per day — essentially identical to next-ai-draw-io's growth rate — reflects a different kind of adoption. This isn't people who've used it and starred it. This is people who've read it and starred it as a reference. The repository's primary artifact is a structured thinking document about AI infrastructure design, not a running application. You're not downloading it to run a server; you're studying it to build your own system.
That distinction matters enormously for how you use it. Personal AI Infrastructure is planning software. The payoff is clarity about how your tools fit together, not a capability you can exercise immediately. If you've been accumulating AI subscriptions and feeling like they're pulling in different directions, this repository is worth reading before adding another tool. If you're looking for something you can deploy this weekend, this isn't it.
The three decelerating tools
Khoj — 35,046 stars
Khoj remains the most-starred project in this cohort by a significant margin and is described accurately as "Your AI second brain. Self-hostable. Get answers from the web or your docs. Build custom agents, schedule automations, do deep research." We've covered Khoj in depth twice on this site; I won't reprise the full analysis.
The velocity number — 9.7 stars per day on a base of 35,046 — reflects something real about where Khoj sits in the market. It's the acknowledged incumbent. Technical users who want a self-hosted AI personal assistant have heard of Khoj, have either tried it or decided to try it, and the remaining new stars are coming from people farther down the discovery funnel. That's not a failure; it's what dominance looks like. But it also means Khoj is no longer where the momentum is.
For productivity users who haven't tried it yet: Khoj is still the most fully-featured self-hostable AI assistant in the open-source cohort. The question is whether document retrieval is your primary bottleneck.
Rowboat — 14,940 stars
github.com/rowboatlabs/rowboat
Rowboat is described as "Open-source AI coworker, with memory." The persistent-memory angle is genuinely differentiated from Khoj: Rowboat doesn't primarily retrieve documents, it builds and maintains a memory layer across conversations so agents accumulate working context the way a human colleague does.
The 4.2 stars per day is the number I'd want Rowboat's team to be thinking hard about. The persistent-memory use case is well-articulated and genuinely addresses a real pain point — the stateless AI problem that makes chatbots frustrating for ongoing work. But the velocity suggests the message isn't breaking through to new users at the rate it was in early May. Possible explanations: the use case is harder to demonstrate in under ten minutes than next-ai-draw-io's; the setup requirement (self-hosted, requiring infrastructure decisions) filters out casual evaluators; or the editorial coverage has already reached most of Rowboat's addressable audience and new discovery is slowing.
None of those are fatal problems. They are reasons Rowboat needs a sharper onboarding experience before the velocity gap widens further.
Plotly Dash — 24,245 stars
Dash's 1.1 stars per day is not a story about stagnation — it's a story about maturity. Plotly Dash is a Python data dashboarding framework with 24,000+ stars and years of production deployments behind it. It grows slowly because nearly everyone who would use it already knows about it. The "AI productivity" label fits Dash primarily in the context of building AI-assisted data tools and internal apps, not as an AI assistant.
Dash deserves its place in this cohort as a reminder that not every "AI productivity tool" is a chatbot or an agent framework. If your productivity bottleneck is building internal data applications for your team — dashboards, reporting tools, lightweight data apps — Dash remains the most mature and most capable open-source option. The slow star growth is noise for anyone evaluating it seriously. What matters for Dash is the size and health of its ecosystem, which remains strong.
What we'd use, and why
I'm going to give three different answers here because the right choice depends on the problem you're actually trying to solve.
If you want something working this week: next-ai-draw-io. The feedback loop is the shortest, the setup is a Next.js deployment, and the use case is narrow enough that you'll know within a day whether it earns a permanent place in your workflow. The 58.5 stars per day isn't hype; it's people finding genuine value.
If you want the most complete AI second brain: Khoj. The star count and feature set are both earned. Set aside an afternoon to configure it properly — point it at your actual document store, pick a model you trust, tune the retrieval settings — and you'll have something that rewards the investment over months. Don't treat it as a quick install; it's a platform decision.
If you're building or rebuilding your AI stack from scratch: read Personal AI Infrastructure first, before adding any other tool. The framework doesn't run, but the conceptual clarity it provides will save you from adding three subscriptions that conflict with each other. Think of it as the architecture document you'd write before starting an engineering project. The 58.8 stars per day suggests a lot of people have found this reframe valuable; I agree with that assessment.
Limitations
Three gaps in this analysis are worth naming explicitly.
No community data. Reddit and Hacker News returned zero results on every research run since May 8. This matters because star counts don't capture active users — a repository can have 30,000 stars and a nearly-dead community, or 2,000 stars and a thriving one. We're missing the qualitative signal that would let us distinguish genuine momentum from viral attention followed by abandonment.
No commercial tool data. Reclaim AI, Motion, Otter.ai, and Superhuman are the most prominent commercial players in AI productivity, and none of them appear in our research because their official pages have been consistently unavailable to our pipeline. A complete picture of the AI productivity market requires commercial tool analysis that we simply cannot provide right now.
Velocity is 3–4 weeks of data. Five weeks is a short window for drawing firm conclusions about trajectory. A single viral Reddit post or Product Hunt feature can generate 1,000 stars in a day and completely distort a daily-average calculation. The velocity numbers here are directional, not definitive.
Bottom line
The tools people are talking about and the tools people are actually discovering are not the same list in mid-2026. next-ai-draw-io and danielmiessler/Personal_AI_Infrastructure are growing at more than six times the rate of Khoj, despite receiving a fraction of the editorial coverage. Khoj and Rowboat are consolidating existing audiences rather than expanding into new ones. Plotly Dash is a mature tool that doesn't need velocity to justify itself.
If I were advising someone coming to AI productivity tools for the first time right now, I'd tell them to start with next-ai-draw-io for an immediate win, read Personal AI Infrastructure to get their architecture thinking straight, and then evaluate Khoj seriously if document retrieval is a real bottleneck in their actual work. Don't start with the most-covered tool just because it's the most-covered tool. Start with the one that solves your specific problem fastest.
The velocity data suggests other people are figuring this out.