Browse · Organizations
Organization · Latent Space wiki

The Fourth Law

A Kyiv-based robotics company developing AI autonomy modules and drones for Ukraine’s defense, beginning with terminal guidance and expanding into interception and broader autonomous capabilities.

The Fourth Law (TFL) is a Kyiv-based robotics autonomy company founded by Yaroslav Azhnyuk. Its initial work centered on AI guidance for first-person-view (FPV) drones: allowing a drone to complete its approach to an operator-selected target after losing its connection to the pilot. Azhnyuk described its goal as “massively scalable on-drone autonomy” and its mission as strengthening the defense of Ukraine and the free world.[1][3][4][2]

From terminal guidance to interception

TFL-1, the company’s autonomy module, performs terminal guidance during approximately the last 500 meters of flight. The pilot flies toward the target and locks onto it; the drone then guides itself through the final approach. This addresses both electronic jamming and the radio horizon: descending near the ground can put a drone behind terrain or beyond direct radio visibility. Azhnyuk identified terminal guidance as the company’s first launched product.[1][3]

The company offers autonomy as both a module for other manufacturers’ aircraft and an integrated drone, Lupynis-10-TFL-1. A July 2025 report described Lupynis as having TFL-1 built in and a reported range of 30 kilometers with a 1-kilogram payload. By February 2026, TFL’s AI modules had reportedly been integrated with dozens of third-party UAV manufacturers, supporting external airframes, different ground stations, and multiple connectivity architectures.[3][4]

On September 1, 2025, TFL announced that Lupynis-10-TFL-1 and the TFL-1 module had been codified by Ukraine’s National Guard and Ministry of Defense. On September 15, it and drone manufacturer Vyriy announced the start of mass production of drones equipped with terminal guidance. The announcement supplied no production volume.[2]

By February 2026, the product range included TFL-AntiShahed, an interceptor-drone module using AI on the device to detect and identify Shahed and Geran strike drones. It was described as identifying them faster than manual observation. Detection assistance is a distinct capability from autonomous engagement.[4]

Tech.eu’s February 16, 2026 report stated that TFL-1 and Lupynis-10-TFL-1 were used by more than 50 Ukrainian military units across multiple frontline areas. The company claimed that first-level autonomy increased FPV mission success rates two to four times while adding around 10 percent to unit costs; the report supplied no controlled-test methodology or representative baseline.[4]

The company’s drone work followed Azhnyuk’s earlier career in consumer hardware. His experience building Petcube supplies context for the transition to defense technology.[1][3]

Earlier history: Petcube and the 2023 decision

Azhnyuk recalled living in San Francisco from 2014 to 2020 while building Petcube, a pet-camera company. In his Latent Space interview, recorded in April 2026 and published May 18, he said that by 2023 he expected a prolonged war and a new arms race centered on drones. He pointed to software-defined platforms, supply-chain risks, and AI as reasons to start a defense technology company. This was his account of the decision, rather than a precise incorporation date.[1]

In parallel with The Fourth Law, Azhnyuk and his Petcube co-founders, partners, and friends started Odd Systems to make thermal cameras. In the April 2026 interview, he described the companies as moving closer together and probably going to merge. He described three business units—cameras, drone autonomy, and drones—and said the group’s camera and autonomy operations sold to over 200 Ukrainian drone manufacturers, while its UAV operation sold drones to Ukraine’s armed forces and government. That figure covered the described group, rather than standalone TFL customers; the merger remained an intention in his account.[1]

Financing and research priorities

Resilience Media reported on July 15, 2025 that TFL was announcing its first outside funding, from investors in the United States, Canada, Switzerland, Finland, and Spain. The amount and investor identities were undisclosed. The funding was intended to support development and sales of existing products and exploration of dual-use enterprise applications. Azhnyuk also said the company had offices in the United States, United Kingdom, and European Union in anticipation of export expansion.[3]

On February 16, 2026, Tech.eu reported that TFL had secured a new funding round backed by Axon, the US public safety technology group. Azhnyuk said the funding would support research and development of autonomy capabilities to protect cities and critical infrastructure against Shahed-type attacks. The report disclosed neither an amount nor a valuation.[4]

TFL’s autonomy software stack includes simulation, analytical tools, autonomous applications, and fleet management systems. Azhnyuk also described the group as operating its own simulation and a training school for warfighters. His five-level autonomy taxonomy progressed from terminal guidance through bombing, autonomous target detection and engagement decisions, navigation, and autonomous takeoff and landing. Those levels described capabilities, rather than establishing that every generation had been commercially deployed.[4][1]

Azhnyuk emphasized the engineering gap between a prototype, a system ready to scale to millions of units, and one actually deployed at that scale. He linked that gap to the long tail of edge cases encountered as autonomous behavior becomes more complex. The distinction was central to his account of developing successive autonomy capabilities; it was not a claim that TFL had delivered millions of systems.[1]

Financing

First outside funding
Resilience Media reported on July 15, 2025 that TFL was announcing its first outside funding, from investors in the United States, Canada, Switzerland, Finland, and Spain. The amount and investor identities were undisclosed.[3]

From the Latent Space archive

Original posts and separately reviewed explanations.

Sources

  1. 1
  2. 2
  3. 3
  4. 4

Model source and teaching reviews passed. Human editorial review pending. This profile describes the sources as of 2026-09-17.

Latent Space

Join Latent Space.

Sign in with Google, or get the newsletter by email.

Loading Google sign-in…

or subscribe by email

Free. Unsubscribe any time.