Skip to main content
Dogeartech

Lattice Prompt 1.1 Extends AI Assistance Across FPGA Design and Timing Closure

Lattice Prompt 1.1 connects AI assistants to FPGA tools, with company examples spanning motion control, OCT code porting and 200 MHz timing closure.

Lattice Prompt 1.1 Extends AI Assistance Across FPGA Design and Timing Closure

FPGA design work can stretch across tool setup, implementation, verification and repeated timing fixes. Lattice Semiconductor’s Prompt 1.1 aims to shorten that cycle by letting engineers describe tasks in everyday language while an AI assistant works with Lattice’s development tools. The company reports substantial time savings in selected examples, including a timing closure task completed in four hours. Those results are notable, though they remain company examples rather than independent benchmarks.

An AI assistant connected to FPGA tools

Writing RTL in Verilog, SystemVerilog or VHDL is only one part of an FPGA project. Designers also handle synthesis, placement, routing, timing analysis, simulation and configuration bitstream generation. Prompt provides a way to ask for work in natural language, then helps translate the request into tool operations, inspect results and suggest next steps.

Prompt doesn’t include its own large language model. Lattice designed it to work with compatible AI assistants, including Claude Code and Cursor, so engineers can bring a preferred model and interface. The AI agent supplies reasoning and makes design decisions; Prompt contributes Lattice-specific knowledge and access to the company’s tools.

That connection uses the Model Context Protocol (MCP), an open standard for linking AI applications with external tools, alongside “Skills” that provide task instructions and specialized knowledge. Through these mechanisms, Prompt can connect an agent with Lattice development tools, device information, IP catalogs, reference designs and engineering knowledge. Deterministic validators check results against design rules and constraints, adding a defined check to a workflow that otherwise relies on an AI agent’s interpretation.

Lattice says that Prompt keeps design files, build logs and license settings on the user’s machine. Engineers still need to consider how their selected AI assistant handles data. That distinction matters for teams weighing the convenience of a connected workflow against their own data handling requirements.

Lattice Prompt lets designers describe tasks in natural language. (Source: Lattice)

What changed in version 1.1

Prompt 1.0, released in September 2026, focused on automating work in Lattice Radiant, the company’s FPGA design software. Its scope included synthesis, placement, routing, timing analysis, simulation and bitstream generation. Version 1.1, released on 30 October 2026, adds design consultant capabilities for system-on-chip construction with Lattice Propel, RISC-V firmware development, timing closure, RTL simulation and hardware debugging with Lattice Reveal.

Propel supports configurable RISC-V soft-core processors in four principal variants. Nano targets simple control and monitoring; SM is intended for small bare-metal systems; MC suits more capable microcontroller applications; and RX addresses higher-performance uses, including systems with real-time operating systems. Prompt 1.1 can help assemble processor-based systems, configure IP blocks and peripherals, and produce associated firmware, drivers and linker scripts from natural-language requests.

Two MCP servers divide the work behind the interface. The Lattice Prompt Server supplies design knowledge, validation and execution capabilities, including Propel integration. The Lattice Radiant MCP Server handles the Radiant implementation flow, including timing analysis and closure. The AI agent calls on the relevant server for a task, while the designer can continue using a chosen assistant.

Lattice Prompt 1.1 uses two servers to connect design tasks with implementation tools. (Source: Lattice)

Three examples, three different engineering tasks

Lattice’s first example applies Prompt to an industrial motion-control drive using Time-Sensitive Networking (TSN). TSN combines Ethernet enhancements intended to support deterministic communication, a relevant requirement when motors, sensors and actuators need to exchange data predictably. The sample request asks for “a TSN Ethernet endpoint for a motion-control drive, deterministic and low power.” Prompt can help identify suitable device information and IP, run development tools and work toward a bitstream, timing and simulation results, and documentation. The example describes a workflow; it doesn’t provide measured latency or power figures.

The second example moves optical coherence tomography (OCT) reconstruction from NVIDIA CUDA code to a Lattice Avant FPGA. OCT uses reflected light to produce cross-sectional images of materials and biological tissue. Lattice’s engineers began with CUDA code and asked Prompt to port it. The described process first produced a Python implementation checked against the CUDA version, then RTL checked against Python, before implementation on the FPGA. This staged comparison gives the team intermediate reference points as the design moves from software to hardware.

The resulting FPGA design ran at 200 MHz. Lattice reports that the engineers had no previous experience with CUDA or OCT and that the work, including verification, took about 30 hours. The company compares that with months of development using conventional techniques. The comparison indicates the potential value of assistance in a specialized porting task, but the reported time depends on this particular design and verification process.

The final example addresses timing closure for Relay-BP, a decoder associated with quantum error correction. Its target was 200 MHz on a Lattice Avant FPGA; the initial implementation reached about 98 MHz. The request asked Prompt to fix timing that was missing the target by 5 ns. The AI-assisted process helped identify timing violations and guide implementation changes, reducing the number of iterations, according to Lattice.

The reported result met 200 MHz with 0.075 ns of positive slack at 85°C, zero hold-time errors and passing simulations. Lattice says the task took four hours, compared with an estimated effort of days using conventional methods. The frequency, temperature and slack figures make this example more concrete than a general productivity claim, although the result is still a company-reported case.

Where the time savings need scrutiny

Lattice claims productivity gains of 10 times or more for certain tasks. Its examples show why the claim may attract FPGA teams: natural-language requests can trigger work across tools and help reduce manual setup or repeated analysis. The practical value will depend on how well an agent interprets a particular design goal, how much expertise is needed to review its output, and whether validation catches errors before hardware testing.

Prompt 1.1 is available as a free download with a one-command installer and user guide. Lattice’s examples point to a workflow that can span specification, code generation and implementation, but independent verification and hardware debugging remain areas for further development. The useful measure for engineering teams will be whether these assisted flows produce repeatable, reviewable results on their own designs, with the timing and verification evidence needed to trust the hardware.

Source: Max Maxfield · www.eejournal.com

Filed alongside