v1.2 - World 2, Rewired

The Toggling Switch Gate chip has been redesigned to be faithful to the circuit it packages: ports now sit where the w2l1 Gatekeeper construction puts them (control rail straight across the top, inputs west, outputs east, in the Dup/rDup form factor) — and every World 2 level was rebuilt around it, then play-tested end to end.

Highlights

• The Duplicator (W2·L4), redesigned. The #1 play-tester friction point is gone: the pre-placed switch gates spread into a centered column, timing margins are generous (a first-try natural wiring certifies), and the layout is honest — gate A is mounted upside-down so the copy loop runs dead straight, leaving exactly ONE forced wire crossing, bridged by the one Crossover you're handed.
• World 2 planarity. The Sorting Office, Merge Lanes, AND Finally, and Putting It Together all route with zero crossings now; Merge Lanes visibly IS the demultiplexer mirror-imaged ("run it in reverse", literally — and the post-level text now invites you to watch it happen with the ◀ Reverse button).
• Looser schedules, measured. Pulse-injection times on W2·L4/L6 were re-spaced using a new timing-margin instrument in the engine; the capstone now tolerates several extra cells of wire on every path.
• Fairness & polish guards in the test suite: minimum I/O spacing (readable labels), no labels pushed off the board edge, references must place-and-use every supplied component, and a hard timing-margin floor for every reference across all 100 certification seeds — plus the planarity counter now catches a wire crossing itself.
• Attribution & pedagogy: Andrew Ressler is credited alongside Feynman where switch gates are introduced (his 1981 MIT thesis proved them universal; the two inventions were independent); the TCB is named in full (Toggling Controlled Barrier); and several level texts were tightened — the W2·L6 garbage accounting is now exact, and its universality claim is properly scoped to the {Rotary, TCB} element set.

Full change detail in CLAUDE.md's changelog and TODO.md. Test suite: 586 + 196 checks green.

v1.1 - Three Ports Are Enough

A substantive-if-minor upgrade over v1.0: a new World-4 capstone level, two new (aspirational) reversible devices, and a handful of editor/UX fixes.

New

World 4 · Level 7 — "Three Ports Are Enough"

A self-resetting reversible switch gate built from only reversible ≤3-port devices — a constructive proof that unipolar BARCS is universal without the 2017 paper's machinery (no Duplicators, constant-1 streams, or garbage rails). Zero heat; the entire scratch space is one control token on a round trip, holding a logic window open just long enough for the data to slip through.

Two aspirational reversible 3-ports (LPS '23)

• Self-Toggling Switch (STS) — a unipolar 3-port with a one-bit memory; every interaction toggles it. The symbol shows its state at a glance: the live S-path solid, the reflecting port dashed. Also gains the PS-style bent toggle (S/U/D).
• Unconditional Toggle Rotary (UTR) — a rotary that reverses its sense each pulse; the symbol shows the current rotation bold and the reverse it will flip to faint/dashed.

Notebook & references

• New "Three ports are enough" lab-notebook page — the universality argument, with an honest "the device is the open problem" caveat.
• The [LPS 2023] talk added to the in-game references (with a link to the public PDF).
• Ships with both a planar STS reference solution and a certified UTR-based alternate construction.

Fixes & polish

• Selecting an element then starting a run no longer leaves it selected — the arrow keys can't nudge a device mid-run and skew the timing.
• The finale's success dialog now offers an Enter the Sandbox → button as a reward for finishing the sequence.
• Level-picker bonus levels show e.g. 6★ instead of a bare ★.
• The inspector flags aspirational elements ("no known implementation yet as a fully ballistic, reversible JJ circuit").

Live at https://fluxonaut.netlify.app

Version 1.0

This is the first "polished" release of the Fluxonaut puzzle game, presenting the BARCS model of computation.

The initial rough (but playable) draft of the game was synthesized "one-shot" by Claude Fable 5 (Extra High thinking, in the Cowork harness) on Wednesday June 10th, 2026, over a single ~1-2 hour run, after just one prompt, with just a folder of research PDFs as input. Pretty amazing.

After the Fable takedown, subsequent polishing was done more interactively using Claude Opus (mostly 4.8, mostly in Max thinking or Ultracode agent workflow mode, mostly under the Claude Code harness) over the next couple of weeks.

This was a "vibe coding" exercise, in that I (M. Frank) didn't touch a single line of code personally -- I just directed the AI.