Simantic Wants AI Agents to Simulate Firmware Before It Touches a Board
Simantic, a four-person startup in Y Combinator's Fall 2026 (F26) batch, is building hardware simulation so that AI agents can work on embedded firmware without waiting for physical prototypes. Its Y Combinator profile, titled "Firmware simulation for AI agents," lists the company's status as Active.
The company was founded in 2025. Its listed founders are Seungmin Hong, who attended the Korea Advanced Institute of Science and Technology, and Ahnaf Shahriar. The listed location is Waterloo, and the primary Y Combinator partner is Aaron Epstein. The founder listings also show the company in the F26 batch with a team size of four.
Two descriptions of the product are published. A job posting says Simantic is developing an electronics simulation platform based on SPICE and Xyce for agentic AI. SPICE is the circuit-simulation netlist standard on which much analog and mixed-signal design still runs. Xyce is a simulator in the SPICE tradition. The company's profile says it is modernizing the embedded firmware industry with hardware simulation developed for the adoption of AI in hardware engineering, and that it gives both humans and AI visibility into hardware previously limited by the physical world.
The company is hiring a Lead Simulation Researcher (Electronics / Firmware), according to its jobs page. The posting suggests the simulation work is still being staffed up.
The problem will be familiar to anyone who has brought up a board. Firmware development usually waits on physical prototypes, test rigs and bench time. An agent that writes code quickly still has to wait for that loop, unless hardware behavior can be modeled in software with enough fidelity to be useful. Simantic's framing targets that gap. Whether a SPICE-based approach can cover the timing, peripheral and analog behavior that firmware depends on is a technical question.
In my view, the company's language should be read as a statement of intent. "Modernizing" an industry and making hardware "visible" to AI are ambitious descriptions for a team of four that is still hiring a lead researcher. Simulation fidelity is the measure that will matter to engineers.
The broader context here is a shift in how engineering tools are built. Software agents have improved quickly at writing code, but embedded systems add constraints that software sandboxes do not, such as registers, interrupts, power budgets and analog effects. A simulator that an agent can run thousands of times without touching a board would change the economics of firmware iteration, if it works at the required depth. Simantic is one of several attempts to build that layer, and no results have been published yet.
For now, the facts are narrow: a 2025 founding, a four-person F26 team, an electronics simulation platform described as SPICE- and Xyce-based, and an open research role. Engineers who work on firmware and circuit verification will be watching for benchmarks against real silicon behavior. If the company publishes them, the case for simulation-first firmware development can be tested directly.
