A responsive rebuild, a seventh game, and six new standards
Random AI Prompt goes from 2.38.1 to 2.41.0, rebuilding its interface to work across phone, tablet, and desktop, flattening the repository to a single project, and hardening its release path. Fairy Fox Games reaches 0.11.2 with a seventh game, Loft, a growth architecture rolled across the whole collection, and Netlify retired for a single host. On the hub, six new and updated standards land.
A day about reach and structure. Random AI Prompt moved
from 2.38.1 to 2.41.0, making the interface work on small screens and simplifying the
repository underneath it. Fairy Fox Games went from 0.6.0
to 0.11.2 with a seventh game and a depth pass across the whole collection. And the hub
folded a batch of the projects’ own proposals into six shared standards.
An interface that fits a phone
The headline for Random AI Prompt was a responsive rebuild of the web app. It landed
in phases (2.40.1 onward): a fluid token foundation, a dedicated layout cascade layer,
and then the actual small-screen behaviour — the top-bar control pile collapses behind an
overflow menu, the building-block palette and the Manage editor become phone drawers and
master/detail views, the single-image view stacks its image over its metadata, and every
interactive control grows to a comfortable touch target. Header submenus turn into
full-width bottom sheets, and the composer keeps its Generate button anchored bottom-right
even as the field bar wraps. A closing pass (2.41.0) added a tablet tier — split-screen
layouts and compact chrome between 769 and 1024px — so the three form factors each get a
layout designed for them rather than one stretched to fit.
One project at the root
Underneath the UI work, the repository was flattened. The earlier revival had kept the
codebase split into an active engine-v3/ and a frozen engine-v1-2/ snapshot of the
pre-revival CommonJS CLI and classic web UI. That split is now gone: the active project
lives at the repository root — engine under src/, prompt content under data/, the React
- Vite app under
gui/, tests undertests/, all commands run from one place — and the pre-revival system was removed from the tree, preserved in git history and as a read-only reference clone. The architecture deep-dives were rewritten to the current shape and a stale-reference sweep followed.
That change rippled into the hub’s own data: the downloads page and the
Random AI Prompt node both still told people to cd into
engine-v3/, so both were corrected to run from the repository root.
A firmer release path
The day also hardened how releases ship. The main branch is now branch-protected —
releases run through a pull request rather than a local push to main, with strict status
checks and force-push and deletion blocked — and the release workflow keyless-signs its
assets so each published build carries a verifiable signature. Repo-hygiene guardrails (a
broken-doc-link check, a tidy check, auto-deletion of merged branches) and dependency bumps
rounded it out.
A seventh game, and a growth architecture for all of them
Fairy Fox Games had its busiest day yet, climbing from 0.6.0
to 0.11.2. A seventh game joined the collection — Loft, a keepy-uppy juggle where you
tap each glowing orb as it falls to bat it back up; it rewards reading a cluster and
catching several in a row rather than mashing.
The larger move was a growth architecture rolled across every game. Each one now shares the same three layers on top of its own core hook: a readable stage arc (a HUD chip and a field tint that shift as you climb), persistent meta-progression (a per-game lifetime record with skill-safe badges and an end-of-run report), and feel-and-feedback depth — while Polarity was rebuilt as the reference implementation, a precision-combo runner whose last-instant flips grow a multiplier. Every layer is pure, tested logic; the collection’s test count reached 217. Housekeeping came with it: game icons across the collection, a public-copy refresh that leads with the AI-experiment story, and Netlify retired so GitHub Pages is the single host.
Six standards folded into the hub
The site’s own work was on the shared standards. Several of the projects’ process reports
had proposed conventions worth making mesh-wide, and the hub adopted six of them (0.12.0):
a git-workflow fix (after any release, dev must contain main, closing a drift the
projects had actually hit) plus five new standards — supply-chain hardening (least-privilege
workflow permissions, pinned Actions, a security policy, signed releases, branch protection),
dependencies (upgrade aggressively behind a real test gate), legal-docs (self-hosted,
code-accurate Privacy/Terms/Cookies), agent-tooling, and badges. Each ships with a
Verify check wired into the compliance audit, alongside six new templates. A follow-up
(0.12.1) made adopting these the default posture for a node running the check flow, without
loosening any of the anti-recursion or reconciliation safeguards.