honeybees is the second vertical in our "taking AI vision seriously for a small hobbyist community" series, after cheesemaking. The pattern is the same: a domain where humans already take detailed notes, where a second opinion on a visual artifact would be useful, and where nobody has built a decent app because the market is too narrow for venture scale.

Backyard beekeeping. About 125,000 hobbyists in the United States, depending on how you count. People keeping 1-4 hives in their yard, their friend's property, or a local community garden. They inspect their hives every 7-14 days during the season. They write down what they see. They worry about their queen, their mite counts, their honey stores going into winter.

We built honeybees for them.

What's In the App

Four core models, built on our standard Flask + SQLite template:

Hives. Per-hive record with name, variety (Italian, Carniolan, Russian, Saskatraz, "local mutt" — yes, that's a legitimate option), location, installed date. Status transitions: active / queenless / absconded / overwintered. Full hive timeline.

Queens. Each queen has a source (packaged, nuc, raised-from-cell, caught-swarm), install date, marker color (international convention: white for years ending 1/6, yellow for 2/7, red for 3/8, green for 4/9, blue for 5/0), notes. Installing a new queen auto-retires the previous one so you don't end up with orphan queen records.

Inspections. The workhorse. Per-visit: temperature, weather, frames drawn, brood pattern assessment, queen-seen checkbox, varroa count, free-form observations. Builds a searchable timeline so you can answer "when did we last see eggs?" without scrolling through a paper notebook.

AI frame reader. This is the feature people actually ask about. Upload a photo of a pulled frame. Claude classifies the frame type (brood / honey / pollen / mixed / empty), grades the brood pattern, estimates comb coverage by percent, flags mite signs, and suggests a next step.

Why the Frame Reader Matters

Frame reading is a genuinely hard skill for a new beekeeper. You're looking for a dozen different things: eggs, larvae, capped brood, stores, pollen, queen cells, mite signs, any abnormality. Even with books and YouTube, new beekeepers misread frames all the time. A misread frame can mean missing a queenless hive for two weeks, or not catching a mite problem until it's crashed.

The Claude-powered frame reader doesn't replace the beekeeper's judgment — it gives them a second set of eyes. "This frame shows capped brood with a solid pattern, queen coverage appears strong. Comb coverage ~85%. No visible mite signs. Next step: standard inspection cycle." Or: "This frame shows spotty brood, some dead larvae visible, possible signs of European Foulbrood or Varroa damage. Recommend a mite count and consultation with your local bee association."

That second read is useful in a way that no static book can be. It's looking at your frame, this week, in your yard. It's fast. And crucially, when it's not sure, it says so. The dcch wrapper handles the "API is down" case by falling back to a template — the app keeps working, you just don't get the vision analysis that one time.

The Build

Same template as every other vertical. Flask + SQLite, single-password auth, deterministic port assignment, Stripe wiring. About 1,400 lines of custom Python on top.

Quickstart:

git clone https://github.com/Dangercorn-Enterprises/honeybees.git
cd honeybees
pip install -r requirements.txt
echo "AUTH_PASSWORD=pickyourown" >> .env
echo "ANTHROPIC_API_KEY=sk-ant-..." >> .env   # optional
python app.py

Your yard's records live in honeybees.db on your disk. You can back them up by copying the file. If we go out of business tomorrow, your hive log survives.

Pricing

Like every vertical, the self-host tier is the same code as hosted. If you don't want the operational burden of running it yourself, $29/mo is a fair trade. If you do, clone the repo and go.

Who It's For

The hobbyist with 1-4 hives who already keeps notes. The new beekeeper in year 1 or 2 who's still learning to read frames and wants the second opinion. The local beekeeping club that wants a shared inspection calendar and a way to log collective yard visits.

Not for: commercial operations with 100+ hives, serious queen-rearing operations, migratory beekeepers doing California almond pollination. Those folks need industry-grade software (and there are a couple of expensive packages that serve them). honeybees is calibrated for people in the range of "5 hives and fifty weekend hours a season."

The addressable revenue on honeybees is maybe $14M/year at 100% saturation across North America. That's not a venture-scale number. It's a great small-business number, especially alongside 219 other verticals on the same template.

What's In the Backlog

Features I've heard requests for, from the handful of beekeepers who've been beta testing:

The last one is interesting because it's network-effect-y, and I generally don't build network-effect features (they're hard to ship as a self-host-friendly app). But swarm calls are local, so the network can be "beekeepers in my county" rather than "beekeepers globally," which is more tractable.

If You Keep Bees

Pull the repo. Try it. Tell me what's missing. tim@dangercorn.net lands in my inbox. This is still early enough that a single well-reasoned feature request can change the roadmap.

And if you don't keep bees but know someone who does: send them the link. The best marketing we have right now is "tell a beekeeper who takes notes."