This blog launched on July 3rd with three posts, then went quiet. The product did the opposite. The weeks on either side of that launch were the busiest de:volt has ever had, and almost none of it got written down where users could find it. That is a documentation failure, and this post is the repair.

We have backfilled the changelog with one entry per release, each dated to the day the work actually reached you, reconstructed from the release history. A few entries predate the blog itself, because the story starts in mid-June and pretending otherwise would be sillier than admitting the blog showed up late. Here is the arc.

Mid-June: Learn and collaboration grew up

In one week, de:volt Learn gained its Logic Gates course, honest objective feedback, fairer teacher marking, and in-product data export and deletion. The next day, live collaboration went live: shared projects with roles, real-time co-editing, comments pinned to components, annual pricing, and an AI tutor in Learn that reads your actual circuit before it hints. The same day, the whole simulator learned to work on phones and tablets, and the day after, the canvas learned to talk to screen readers, including keyboard wiring and spoken breadboard hole addresses.

That week also brought jumper wires, rail-less breadboards, and proper tactile and DPDT switches, and closed with the oscilloscope gaining triggered acquisition, node-mapped channels, PNG capture, and on-canvas labels.

Early July: MicroPython arrived

The single biggest addition of the summer: two boards that run real MicroPython in the browser. A BBC micro:bit V1 with a live display, clickable buttons, and wireable edge pads. An RP2040 board with the full GPIO, PWM, and ADC surface, emulated at the silicon level. Classroom code runs unmodified, and its pins drive the same simulated breadboard as every resistor and LED in the catalog. Both boards on every plan, free included.

Mid-July: the truth release

July 11th was one push with three stories in it. The scope stopped lying: acquisition moved inside the solver at up to 10 kSa/s, and measurements, cursors, honest noise, and coupling arrived. Parts started failing like parts: junction temperature, thermal shutdown, energy-based damage, real microcontroller pull-ups, floating-input warnings, and batteries that run down. And the micro:bit got sensors you can drive by hand, a GPIO breakout, an ultrasonic distance sensor, and a rebuilt Inspector across all 99 parts.

Two days later, a small compatibility release made display.clear() behave like real hardware, which fixed Snake, which we consider load-bearing.

Late July: polish, power, and fixes

The interaction polish pass landed on the 16th: decluttered scope controls, menus with visible shortcuts, dialogs that behave, switches that toggle without hijacking your selection, and differential probes plus net renaming made free for everyone. On the 18th came the engine release: a sparse solver for large circuits, instant small-signal Bode plots, free SPICE netlist import, nine new device families from JFETs to a transformer, and the engine itself published as open source.

The month closed the way we want every month to close: a teacher sent a detailed bug report, and it turned into three fixes and then two follow-ups, including guest circuits that survive a closed tab and a micro:bit audio model faithful down to which pin your program plays through.

What happens now

The changelog stays current from here; when something ships, it gets written down. Starting tomorrow there is also a daily post: how-to guides and technique pieces, one circuit or one tool at a time, each one built and measured in the simulator before it is written about. If it does not help you build, teach, or understand something, it does not ship. That was the bar this blog set for itself in its first post, and a month of silence does not lower it.

If any of the above is news you can use, the bench is at sim.devoltapp.com.