October 7, 2025
The First AI Hardware Engineer
Share
With this release, Flux can take your requirements, generate a complete plan, and execute multi-step workflows right inside the editor. It researches components, builds schematics, places and routes parts, and runs checks along the way — pausing for your feedback when it needs direction.
Think of it as your AI hardware engineer: fast, explainable, and always improving — but still guided by someone who knows the craft. Flux works transparently, explains its reasoning, and remembers how you like to work.
It’s the biggest step yet toward the first true AI hardware engineer.
This new functionality is available now. Log in to Flux today to take it for a spin. Full workflow capabilities will roll out gradually over the coming days.
Start by telling Flux what you need to build. Flux now understands design requirements—the goals, constraints, and specs that define your project. Describe the functionality, power targets, interfaces, layer count, or components you want to use, and Flux will turn that into a complete, step-by-step plan.
You’ll see a clear outline of the plan: parts research, schematic creation, layout, checks, and milestones for review. From there, simply tell Flux about any desired changes—add details, reorder tasks, or lock decisions—and it will refine the plan for you. It’s up to you how in the weeds you get.
Next, click “Start” and Flux will begin get to work, sharing progress and decisions along the way, and checking in with you at key points to get your feedback.
“Design a sub-25 × 25 mm wearable PCB with Bluetooth, an accelerometer, and on-board battery charging.
It must include a BLE SoC (OTA-capable), a low-power accelerometer with interrupt/wake, power-path + charging for a 1-cell Li-ion/LiPo, and headers/pads for programming and test.
Power: 1-cell Li-ion/LiPo with on-board charger (5 V USB input) optimized for low quiescent current.”
“Design a compact field-oriented control (FOC) BLDC motor driver board.
It must include Bluetooth Low Energy for wireless control and data-logging.
The key subsystems are: power stage and gate drive, sensing, MCU selection, comms, and protection to thermal/mechanical stress.
Power: USB-C PD at 12 V (with local regulation as required).”
“Design a low-noise electret microphone preamplifier for a 24-bit ADC, integrated into a consumer household device.
It must have switchable 20–40 dB gain, correctly sized coupling capacitors with a ~20 Hz high-pass, an output anti-alias RC for ~20 kHz bandwidth, and thorough decoupling plus pop-suppression.
Follow the op-amp, microphone, and ADC datasheets and industry best practices; use the 3.3 V analog rail and make cost-effective component choices without asking for spec confirmation.
Power: USB-C 5 V input (with local regulation as required).”
When you approve a plan, Flux doesn’t just hand you suggestions—it gets to work.It now executes full workflows inside the editor, acting like an extension of your team that can solve real problems while keeping you in the loop.
Flux handles the structured parts of the process—researching components, wiring schematics, placing and routing parts, and reviewing its own work for correctness—while you focus on the decisions that need human judgment.
You can think of it like having an AI engineer on your team who works fast, communicates clearly, and never forgets a detail. Feel free to close your browser or go for a walk, Flux will keep working in the background, and drop you a line when it’s time to check in.
Flux is built for collaboration. Every plan, action, and decision it makes is visible and explainable so you can review, guide, and adjust as it goes.
You can pause execution, modify the plan mid-flow, or roll back using version history. Lock regions, nets, or components to prevent changes, or ask Flux to revisit a specific step. And because Flux runs inside a full browser-based ECAD, you can jump in and edit anytime—make manual tweaks, move parts, or add your own changes without breaking its flow.
Flux doesn’t just follow instructions—it learns through your conversations and feedback. When you correct something or clarify how you like to work, Flux can ask if you want to remember it. You choose whether that learning should apply just to the project you’re in or across your entire account.
Over time, Flux picks up the same kind of tribal knowledge your team already shares—naming conventions, layout habits, design rules—and starts applying them automatically. You can refine what it remembers, edit entries, or forget things entirely through the Knowledge Base.
It’s how you teach Flux to work the way you do—so it keeps getting smarter, faster, and more aligned with your standards. Learn more.
The new planning and execution architecture inside Flux is designed to scale—so the agent you’re working with today will keep getting smarter and more capable over time.
This is just the beginning. You can already fork your projects and have Flux explore multiple directions in parallel. Soon you’ll be able to delegate even broader, more complex, assignments to Flux, and have it build even more advanced boards.
We envision a future where Flux is not just one AI engineer, but a coordinated group of AI engineers, each with their own specialization, that seamlessly integrate with your team. The endgame is a world where hardware teams are infinitely scalable: totally parallel, deeply collaborative, and still human-led.
Hardware is entering a new era—where AI becomes part of the team, instead of part of the toolkit.
It starts here. Give Flux a job, review the plan, and help define how engineers and AI build hardware together.
ERC checks schematic-level electrical issues while DRC checks PCB layout rules -- engineers run ERC before layout and DRC during or after routing.
The complete hardware design workflow covers requirements, schematic capture, PCB layout, validation, prototyping, and manufacturing.
PCB constraint management defines routing, spacing, impedance, and manufacturing rules and validates them throughout PCB layout.
PCB revision management tracks what changed, who changed it, and why -- keeping schematics, layout, BOM, and manufacturing files aligned.
PCB symbol design directly impacts how fast a team can interpret a circuit, spot errors, and generate an accurate netlist.
Real-time PCB collaboration allows multiple engineers to review, edit, and manage PCB designs in a shared environment, reducing version confusion and speeding reviews.
Common schematic design mistakes include unclear labeling, missing power connections, inconsistent symbols, unconnected pins, poor organization, and skipping ERC.
How hardware teams apply Git-style version control to PCB design — tracking revisions, collaborating safely, and maintaining a complete design history.
