A month after the official panel was archived, a tutorial with 70k views pushed honk to 1500 stars

On Sep 24, daed, the official panel for dae (大鹅), the soft-router proxy core, was archived. Four weeks later, honk, a Rust rewrite of a new core from the same organization, filled exactly that gap. A Chinese circumvention tutorial creator made a step-by-step video for it, and the next day it gained 315 stars. By Nov 6, this repo, which is 109 days old and has had nearly 900 commits from one person, was approaching 1500 stars. Below: what it is, how the stars came, who is competing for the space, and which parts of its path you can learn from and which you can't copy if you want a project that earns stars the same way.

A question nobody answered: "What happens to dae now?"

dae is a transparent proxy core that runs on Linux and OpenWrt soft routers. Its feature is eBPF traffic splitting inside the kernel, so domestic direct traffic never goes through user space, which makes it fast. It has 6,400 stars, but it never had an official built-in panel. Users relied on a separate repo, daed (2,000 stars). On Sep 2026 24, daed's last commit was followed by it being set to read-only and archived, with no explanation from the project.

The community immediately started asking. On NodeSeek, a thread titled "Just found out daed is already archived" had replies posting links to both dae and honk, saying "looks like there's a new project". On Oct 3, someone on X also posted specifically to explain the archive, mentioning honk along the way.

By then honk had been up for over two months (since Jul 20), with a one-line repo description: "Inspired by dae & sing-box". It was gaining only two or three stars a day: Oct 9 152, Oct 18 181.

What it is

honk sits under the same daeuniverse organization as dae, is rewritten in Rust, and keeps dae's config syntax, so long-time users' config files mostly copy over directly. Beyond dae, it adds three things:

• sing-box-style outbound groups: nestable, plus a strategy called Score that rates nodes by actual performance and picks automatically;
• Clash API compatibility: it can plug straight into zashboard, the panel mihomo users already know;
• built-in panel doona: written by another contributor, Zakkaus, bundled into the program, where rule edits in the panel can be written back to the config file.

In other words, dae's kernel splitting, sing-box's grouping, and mihomo's panel experience are each not new, but a finished product that combines them into eBPF transparent proxying did not exist before. The bar is not low either: it needs Linux kernel 6.12 or later, only supports x86_64 and aarch64, and the first line of the README reads "Early alpha, not recommended for production".

The code is almost entirely one person's work: the author Glassyiris made 895 commits, doona's author made 171, and everyone else only a handful. The repo includes a rules directory for AI coding assistants, which suggests heavy use of AI in development (inference).

Which three days the stars took off

(Star counts are GitHub daily snapshots; data is missing between 10-19 and 10-27.)

Step one: the official side paves the road. In Oct 25, daeuniverse's OpenWrt software feed merged its first PR, a signed apk feed plus honk's LuCI interface. From that day, adding a feed on the router and running one apk add command gives you the Chinese interface and panel, with no compiling needed.

Step two: a relevant creator. Qichiyu is a Chinese YouTube creator focused on soft-router proxy tutorials like OpenClash and mihomo. In Oct 24 he had just made a dae tutorial, which gathered the people wondering "what happens to dae now". On Oct 28 he posted the honk feature, titled "honk: new eBPF core, combining the best of mihomo, sing-box and dae | dae's kernel direct connections + sing-box's nested groups + smart node selection + a panel that beats mihomo, with rule edits written back to config | OpenWrt step-by-step tutorial". By Nov 6 this video had 70.8k views and 637 likes. The companion blog post shared only five things: commands to check kernel version, commands to add the software feed, a one-line install command, the panel address, and two honk configs he wrote (one grouped by service, one trimmed-down "domestic direct, everything else through proxy").

Viewers could follow along and get it running, which directly explains why the stars doubled the second day (the timing lines up; inference, but high confidence).

Step three: secondary spread. On Oct 29 someone posted on X "VPS are dropping en masse... hurry up and swap your soft router for Honk". From Nov 4 other creators followed with tutorials (1,816 views), and on Nov 6 one video's description was word-for-word identical to Qichiyu's, which looks like reposting. Aggregator accounts also reposted it on Reddit. English-language media and large accounts have still never written about it.

Finally, the author kept up: from Oct 25 to Nov 3 they shipped 4 beta versions, and a "default config DNS resolution fails" bug reported by new users was fixed the next day. When a crowd flooded in, the project didn't scare people off, and that counts too.

Who's competing for this spot

Soft-router transparent proxying is a long-saturated space, with the leaders all at tens of thousands of stars:

honk has a narrow niche: the old dae users who want a panel and sing-box-style grouping. Its 1500 stars didn't come from mihomo. They came from dae's existing users switching over.

What followers can't get

Breaking down its three drivers:

• Demand created by the daed archival: a one-off event, and only an official org could legitimately "take over." At the same time, NaiLink, a third-party panel for dae, had a 30k-view video but only 4 stars on its repo.
• Backing from the official org and its package feed: honk sits under the dae name (6,400 stars) and ships through the official apk feed. A solo project can't offer that.
• Qichiyu's video: this part is copyable. He needed a new core that works as a headline and a config he could follow step by step, and honk provided both.

Without the first two, it would probably still be gaining about two stars a day. That's what the early October data shows.

If it were you

C: Worth learning from, not worth chasing. These stars came from "official panel archived, official org takes over," an identity outsiders can't get. The category is already crowded with projects that have tens of thousands of stars, and honk has filled the remaining gap. What's worth taking is its release cadence, not this topic.

Who it suits: developers who have spent time in an open-source ecosystem and own an alternative or enhancement to an upstream component; or people working in areas like soft routers, NAS, and self-hosting that have established tutorial creator communities.

Who it doesn't suit: anyone wanting to write yet another proxy core or build a panel for dae (NaiLink is an example). Without Linux kernel and networking fundamentals, eBPF transparent proxying has a high barrier to entry and high support costs.

How to do it: copy three steps. ① Watch upstream projects in your ecosystem for archival, abandonment, or protocol changes. When that news breaks, the community spends a while asking "what do we use now?" Offer something that migrates smoothly then (config compatibility, a one-command switch). ② Before launch, compress installation to one line: publish to distro or official package feeds, with a Chinese-language interface. ③ Describe your product in one sentence of the form "A's X + B's Y + C's Z." That sentence should work directly as a tutorial video title, with a ready-to-use config attached.

How to get users: find vertical tutorial creators in your niche. For soft routers, that means people like Qichiyu who make step-by-step videos. They constantly need new cores and tools as topics, and the more copy-paste-ready your product is, the more willing they are to cover it. Show HN and English Reddit help little for tools in demand in the Chinese-speaking community.

When it was before: honk's window is already closed. The "take over after upstream archival" play isn't time-bound, but each window is measured in weeks: honk was picked up by tutorial creators less than five weeks after daed was archived.

What signals you've misread it: ① A month later, stars stall below 2000 and issues go quiet, meaning this was a one-time crowd and real demand for a core swap is smaller than it looks; ② A personal project unrelated to the dae org reaches a thousand-plus stars in the same window, meaning backing mattered less than assumed here; ③ English-speaking developers start writing about it and stars surge again, meaning the product's differentiation is more compelling than judged here.