STEM
OverrideForge
A confident AI assistant is right most of the time — and sometimes fluently, plausibly wrong. Solve a reasoning task, lock in YOUR answer first, then Glib gives its very-sure suggestion. Accept it, Verify it, or Override it — and see your reliance profile: where you leaned well, where you surrendered your own correct answer to a wrong one, where you distrusted a tool that was right. Calibrated reliance — knowing when to trust and when to check — is the skill nobody teaches yet. No timer, no score to chase; on-device, no account. Ages 15–18.
#2f6f8f Distributed-narrative cast
Meet the cast
OverrideForge's cast is the two ways you can react to a confident machine, made into characters (not brain parts): Whim is the fast reaction that either takes the fluent answer at face value or reflexively rejects it; Muse is the slow, careful move that does the one check the tool skipped, then decides. Glib is the assistant itself — it answers everything smoothly and confidently whether or not it is right (a deterministic, on-device tool, never a real chatbot). Mentor Solon holds the line: calibrated reliance is neither blind trust nor blind rejection — know which tasks a tool is solid at, and check the ones it is not. Every trial carries a science-meter: how reliable is AI on this kind of task, really — solid, or still argued? (This is the text-forward 15–18 case-narrative form — no illustrated portraits or chapters, by design.)
Whim
The fast, miscalibrated reliance reaction (System 1) — takes Glib's confident answer at face value, or reflexively rejects a reliable tool, without ever doing the check
Muse
The slow, calibrated move (System 2) — does the one check Glib counts on you to skip, works out whether this is a task the tool is reliable at, and only then decides
Glib
The fluent on-device assistant — answers everything smoothly and confidently whether or not it is right; a deterministic pre-authored tool, never a real cloud AI
Solon
(Mentor) — calibrated reliance is neither blind trust nor blind rejection; know which tasks a tool is solid at, check the ones it is not, and never surrender an answer you already had
What's inside
Learning goal
A confident AI assistant is right most of the time — and sometimes fluently, plausibly wrong. Solve a reasoning task, lock in YOUR answer first, then Glib gives its very-sure suggestion. Accept it, Verify it, or Override it — and see your reliance profile: where you leaned well, where you surrendered your own correct answer to a wrong one, where you distrusted a tool that was right. Calibrated reliance — knowing when to trust and when to check — is the skill nobody teaches yet. No timer, no score to chase; on-device, no account. Ages 15–18.
Question kits
16 curriculum-aligned kits × 25 questions = 400 questions per app, mapped to recognized standards.
On-device AI mentor
FoundationModels-powered hints, feedback, and adaptive difficulty — all running locally.
Mentored by Solon — on-device AI, no data leaves the device.
How OverrideForge handles your kid's data
- ✅ All progress, settings, and AI-generated content stays on the device
- ✅ No analytics, no tracking, no third-party SDKs
- ✅ No ads, no in-app purchases — you pay once
- ✅ COPPA compliant under the 2026 FTC amendments
- ✅ Parental controls + session limits + content filters built in
OverrideForge runs on ForgeKit — the open-source Swift Package Manager framework that powers every Spark & Anvil app. ForgeKit ensures consistent accessibility, COPPA compliance, and design language across the portfolio, so your kid's progress and preferences feel coherent across every app they touch.
Coming to the App Store
OverrideForge is in active development. Email us to hear when it ships — no marketing, no spam, just a one-shot launch announcement.
Email me at launch