What cities spend on strays, rats and pigeons, what it buys, and what a real end costs.

Where the problem is loudest: 500,000 stray cats in Miami, and the cities that have spent a decade losing to rats.

Miami-Dade County

Miami-Dade is the honest case: it names the number (~500,000 stray cats, "holding steady") [15], houses over 20,000 cats a year, returns more than 55 % of them through TNVR, and did 33,548 spay/neuter surgeries in FY23-24 while keeping a 90 %+ save rate for ten straight years [14] [13]. The bill went from $28.9M to $39.7M in three years (general-fund share $16.5M → $27.0M) [13]. That is what "doing it right" costs without tracking: $40M a year to hold 500,000 steady. Steady is the ceiling of untracked TNR because nobody knows which colonies are already done and which are re-breeding.

The rat capitals

Chicago was Orkin's rattiest city ten years running (2015-2024) and only lost the title in 2025 because Los Angeles overtook it [16] [17]. Its Bureau of Rodent Control got $14.85M in 2024 and $12.5M in 2026, and the city's own audit says the current plan is "unlikely" to fix it [18] [19]. Meanwhile Tree House Humane Society's Cats at Work program has placed more than 1,000 sterilized feral cats at businesses and homes as rodent deterrents and runs a waitlist [20]. That is the whole Cat City rodent thesis, already proven in the rattiest city in America, just without tracking, without a map, and without the city paying for it.

Ten years of Chicago rodent budgets at today's rate is roughly $130-150M (est.) for a problem that got worse every year of the decade. NYC put $35.5M into two rat initiatives on top of routine sanitation and hired a czar. Both cities pay for poison, inspections and fines; neither pays for the one thing that keeps working after the contract ends: a sterilized, fed, tracked cat colony that stays.

Damage nobody budgets for

Nationally 3.0 million cats entered shelters in 2025 and 277,000 were euthanized [22]. Free-ranging cats kill an estimated 1.3-4.0 billion birds and 6.3-22.3 billion mammals a year in the US, mostly by unowned cats [23]. Feral pigeons cost about $1.1 billion a year in building damage and cleanup [21]. None of that appears in a city budget line, which is why the shelter line looks like the whole cost. It is not.

The animals are how the mesh gets built. What runs on it afterward is the bigger business.

Who maintains it: jobs and service hours

The playbook already staffs Cat City through city job programs. The v2 addition: people who show up and do well in the community meetings and the pilot get first call on the paid roles (tagging, gateway install, cable runs, battery swaps, data entry), and court-ordered community-service hours can be served on infrastructure upkeep and pigeon tagging under a supervised partner. A city that already administers service hours gets a workforce that maintains the network; the network gets people who know the blocks. Rule: no animal handling without the partner's trained lead present.

Voice over the mesh: send the words, not the sound

LoRa moves 1-20 kbit/s, so we never send audio. We send a binary-encoded description of the speech (text plus a few bytes of tone/pace tags, ~1 byte per character) and a small device on the far end, a phone, a $15 node with a speaker, or a PC, speaks it with AI text-to-speech. Today that is a generic voice; a stored voice profile makes it your voice later (a one-time enrolment, a few KB, cached on the receiving device, nothing per message). A 20-word message is ~150 bytes: one LoRa packet, delivered in seconds, spoken instantly. Real audio clips (Codec2, ~700 bytes per 5 s) stay point-to-point and off the shared channel.

Nothing is lost when the link drops. Every message is chunked with sequence numbers; the receiver acknowledges what it has and asks only for the missing chunks; if the sender walks out of range, both sides hold the pieces and resume when a node is back in range, or when a passing cat or pigeon carries the chunks across (delay-tolerant store-and-forward). Play-when-complete is the default; a partial message is never spoken half-way.

Is this internet? No. LoRa is a messaging layer: text, positions, sensor readings, the voice descriptions above. Internet needs Wi-Fi, cellular or satellite. The two combine: a gateway with Wi-Fi or a Starlink dish bridges the mesh to the internet (Meshtastic already does this over MQTT), so one connected roof can serve a whole district's text traffic. That is the model for cities where cell service is expensive or absent, in Africa and anywhere else: residents and small businesses host the $15-$60 nodes and gateways, a local operator runs the map and the message service, and the animals carry the mesh into the blocks no one wired. Text and alerts for the price of a battery, not a phone plan.

The pigeon science (why the tags carry a compass)

Pigeons navigate with a sun compass, a magnetic sense and smell; the magnetic part is still not fully understood. A 6-gram pigeon tag can carry a 3-axis magnetometer and an IMU for about a dollar. Thousands of tagged city pigeons logging heading, magnetic field and turns, for years, is a dataset no lab has. We will not "use" magnetism as a radio (it is not one), but we can publish the first city-scale map of how they actually route, which is the kind of open science that gets a university to co-sign the permits.