AI × ENGINEERING × FREE EDA
Less noise.
More working circuits.
Compare AI costs and verify engineering ideas with tools. Software, boards, embedded systems and RTL — with an open process and reproducible checks.
01 / TOKEN PLANS
Seven ecosystems. Terms first.
An agent, a subscription and an API are different products. Compare prices, constraints and the cost of a clearly defined workload.
02 / PRACTICE
From an idea to a verified result.
Concise guides to free engineering tools. An editorial selection; personal project reports will follow later.
KiCad: from requirements to a checkable PCB
Use AI for interface tables and review preparation, with ERC, DRC and fabrication review built into the project workflow.
FreeCAD: treat an enclosure as testable parameters
A practical enclosure workflow: dimensions, dependencies, boundary variants and export checks that go beyond an attractive rendering.
Embedded: ask AI for a testable patch
Using Zephyr as an example: record the environment, scope one small task and verify behavior from build to hardware.
Yosys + Verilator: two checks for one RTL design
Use lint, simulation and synthesis as distinct evidence when reviewing AI-proposed RTL.
SpinalHDL: check the generator and its RTL
Parameterization needs checks across configurations. AI can propose the matrix; elaboration, simulation and synthesis check the result.
cocotb: turn a protocol into executable checks
Test against an independent behavior model, control when signals are sampled and retain reproducible failures.
03 / REPRODUCIBILITY
Not the best AI in general.
The cost of a specific task.
The same prompt does not guarantee the same token usage. Compare real usage, one task revision and identical acceptance criteria.
Method and measurement table ↗