The printed-circuit-board (PCB) design workflow in the open source electronic design automation (EDA) suite KiCad is still mostly manual, involving interactions with the schematic editor and PCB editor. Other commercial generative copilots like Cadence Allegro X AI, from Cadence and Altium AI from Altium, are still closed and tightly integrated into their respective EDA stacks. We introduce Electrode 1.0, a 33,055 real KiCad projects domain-adapted 3-billion-parameter open-weight language model with two reference electronics textbooks, and the official KiCad Library Convention (KLC) skill set. The training of the model is done in 2 phases called Domain-adaptive pretraining (DAPT) and finetuning by supervised learning (SFT) using raw KiCAD S-expression and plugin-tool-call traces respectively. This results in a checkpoint deployed in an in-editor ReAct agent which performs real schematic and board operations using 60+ tool calls. Electrode 1.0 gets 0.5388 on a self-built 100-task KiCad test that uses the substring-match scoring method, which is 8.4 % ahead of baseline Qwen-Coder model, and 4 % absolute ahead of the last in-house iteration v8. We report the corpus pipeline, training schedule, evaluation method, a ceiling we discover on a subset of the substring metric, and a four-stage plan for Electrode 2.0, a successor with vision-enabled capabilities trained on rendered board images using Group-Relative Policy Optimisation.