TESTDRIVE 1.2.0-beta.1 — Clearer inspections

TESTDRIVE 1.2.0-beta.1

This release simplifies the home screen, test library, guided checks and report. The automatic scan now ends with a readable device overview. Battery charge and website storage allowance appear only when the browser exposes valid readings; missing data stays unknown. Website quota is explicitly distinguished from disk capacity and free disk space.

Touchscreen checks can fill the screen, with a live 60-area progress indicator, movable controls, Escape/Exit and retained results. Browsers that decline native fullscreen use the visible page viewport. Display tours also enter fullscreen with a viewport fallback. Camera and microphone steps have clearer permission, capture, stopped and assessment states. Mobile, tablet and desktop layouts follow a lighter green visual direction; reports separate attention items from unknowns and keep technical observations expandable.

This update makes an inspection easier to start and completes more checks from observed input. One click starts the automatic scan; saving a result takes the user to the next pending check.

Beta 4 makes an unknown keyboard easier to finish. It counts received key responses, shows extra keys such as a number pad, and lets the user choose All my keys responded after leaving capture. This is recorded as explicit user confirmation. A generic reference never invents a required physical key count or automatic full-keyboard coverage, and starting another capture clears the confirmation. Known laptop layouts keep their existing automatic coverage checks.

Beta 3 corrects the ASUS AI/Copilot key: a received Win + Shift + F23 sequence lights its dedicated key, while shortcut modifiers do not falsely count as independently tested keys. Three visible red Escape dots show exit progress, reset after a pause or another key, and confirm the third press. The browser must receive the shortcut to identify it; Win + Shift alone is insufficient.

• Four manufacturer-sourced US keyboard profiles: ASUS Zenbook DUO UX8406CA, Apple MacBook Air 13-inch M2 (2022), Dell XPS 13 9315, and Lenovo ThinkPad T14 Gen 4. A supported confirmed Windows import selects a matching layout; manual choices are remembered locally and cleared with local data.
• Explicit keyboard capture with visible active state, per-key press/release progress, triple-Escape, mouse Exit, 60-second inactivity and focus-loss exits.
• An eight-step pointer challenge automatically checks target movement, clicks, both scroll directions and dragging.
• Four-second camera and five-second microphone samples automatically check received media and stop capture. Optional installed on-device speech recognition can match the displayed phrase. No cloud transcription fallback or language-pack download is used.
• Battery/charger readings where available, a pausable display color tour, clearer instructions, optional human feedback and next-check navigation.

The laptop catalog is a reference database, not proof of current hardware specifications. Exact model selection needs a supplied/confirmed model or a manual choice. Operating-system reserved keys may still open system actions because a website cannot intercept keys the OS withholds. Sound detection does not claim word recognition; picture/sound quality and display defects still need human judgement.

Camera/microphone access starts only after the relevant action. Temporary media and normalized reports stay in the browser unless the user exports a report. Firebase serves the static app; no account, inspection backend or API secret is required.

Verification

The release candidate passed TypeScript, ESLint, all 40 unit tests and the production build. Browser, design and public deployment verification are recorded in verification.md and design-qa.md. Full Chromium is used for browser verification. Earlier beta results retain their own source identities.

Software input, synthetic camera/audio streams, and mobile viewport emulation do not establish physical hardware quality. Physical iPhone/Safari and actual camera/microphone/input hardware still require device testing.

Live app · Public source · Update details · Catalog sources

The earlier beta 3 walkthrough remains historical evidence of that release. It does not demonstrate the 1.1 changes. Earlier source tags are unchanged.

OpenAI Codex and specialist coding agents assisted development and verification. They are AI tools, not human team participants. Third-party packages and fonts retain their licenses in THIRD_PARTY_NOTICES.md. A public release does not establish event submission.

TESTDRIVE 1.1.0-beta.4 — easier unknown-keyboard checks

TESTDRIVE 1.1.0-beta.4

This update makes an inspection easier to start and completes more checks from observed input. One click starts the automatic scan; saving a result takes the user to the next pending check.

Beta 4 makes an unknown keyboard easier to finish. It counts received key responses, shows extra keys such as a number pad, and lets the user choose All my keys responded after leaving capture. This is recorded as explicit user confirmation. A generic reference never invents a required physical key count or automatic full-keyboard coverage, and starting another capture clears the confirmation. Known laptop layouts keep their existing automatic coverage checks.

Beta 3 corrects the ASUS AI/Copilot key: a received Win + Shift + F23 sequence lights its dedicated key, while shortcut modifiers do not falsely count as independently tested keys. Three visible red Escape dots show exit progress, reset after a pause or another key, and confirm the third press. The browser must receive the shortcut to identify it; Win + Shift alone is insufficient.

• Four manufacturer-sourced US keyboard profiles: ASUS Zenbook DUO UX8406CA, Apple MacBook Air 13-inch M2 (2022), Dell XPS 13 9315, and Lenovo ThinkPad T14 Gen 4. A supported confirmed Windows import selects a matching layout; manual choices are remembered locally and cleared with local data.
• Explicit keyboard capture with visible active state, per-key press/release progress, triple-Escape, mouse Exit, 60-second inactivity and focus-loss exits.
• An eight-step pointer challenge automatically checks target movement, clicks, both scroll directions and dragging.
• Four-second camera and five-second microphone samples automatically check received media and stop capture. Optional installed on-device speech recognition can match the displayed phrase. No cloud transcription fallback or language-pack download is used.
• Battery/charger readings where available, a pausable display color tour, clearer instructions, optional human feedback and next-check navigation.

The laptop catalog is a reference database, not proof of current hardware specifications. Exact model selection needs a supplied/confirmed model or a manual choice. Operating-system reserved keys may still open system actions because a website cannot intercept keys the OS withholds. Sound detection does not claim word recognition; picture/sound quality and display defects still need human judgement.

Camera/microphone access starts only after the relevant action. Temporary media and normalized reports stay in the browser unless the user exports a report. Firebase serves the static app; no account, inspection backend or API secret is required.

Verification

Beta 4 passed TypeScript, ESLint, all 37 unit tests and the production build. All 19 browser workflows passed in CI. The six affected keyboard/mobile workflows and the complete buyer/report lifecycle also passed against the live HTTPS app. Exact deployed asset hashes, routes, headers and browser security checks passed. Source identities and prior release results are recorded in verification.md. Full Chromium is used for media tests: the stripped Playwright headless shell crashes in its experimental speech availability API even without this app, while the full browser returns the supported local-language status safely.

Software input, synthetic camera/audio streams, and mobile viewport emulation do not establish physical hardware quality. Physical iPhone/Safari and actual camera/microphone/input hardware still require device testing.

Live app · Public source · Update details · Catalog sources

The earlier beta 3 walkthrough remains historical evidence of that release. It does not demonstrate the 1.1 changes. Earlier source tags are unchanged.

OpenAI Codex and specialist coding agents assisted development and verification. They are AI tools, not human team participants. Third-party packages and fonts retain their licenses in THIRD_PARTY_NOTICES.md. A public release does not establish event submission.

TESTDRIVE 1.1.0-beta.3 — Copilot key and Escape progress

TESTDRIVE 1.1.0-beta.3

This update makes an inspection easier to start and completes more checks from observed input. One click starts the automatic scan; saving a result takes the user to the next pending check.

Beta 3 corrects the ASUS AI/Copilot key: a received Win + Shift + F23 sequence lights its dedicated key, while shortcut modifiers do not falsely count as independently tested keys. Three visible red Escape dots show exit progress, reset after a pause or another key, and confirm the third press. The browser must receive the shortcut to identify it; Win + Shift alone is insufficient.

• Four manufacturer-sourced US keyboard profiles: ASUS Zenbook DUO UX8406CA, Apple MacBook Air 13-inch M2 (2022), Dell XPS 13 9315, and Lenovo ThinkPad T14 Gen 4. A supported confirmed Windows import selects a matching layout; manual choices are remembered locally and cleared with local data.
• Explicit keyboard capture with visible active state, per-key press/release progress, triple-Escape, mouse Exit, 60-second inactivity and focus-loss exits.
• An eight-step pointer challenge automatically checks target movement, clicks, both scroll directions and dragging.
• Four-second camera and five-second microphone samples automatically check received media and stop capture. Optional installed on-device speech recognition can match the displayed phrase. No cloud transcription fallback or language-pack download is used.
• Battery/charger readings where available, a pausable display color tour, clearer instructions, optional human feedback and next-check navigation.

The laptop catalog is a reference database, not proof of current hardware specifications. Exact model selection needs a supplied/confirmed model or a manual choice. Operating-system reserved keys may still open system actions because a website cannot intercept keys the OS withholds. Sound detection does not claim word recognition; picture/sound quality and display defects still need human judgement.

Camera/microphone access starts only after the relevant action. Temporary media and normalized reports stay in the browser unless the user exports a report. Firebase serves the static app; no account, inspection backend or API secret is required.

Verification

Beta 3 passed TypeScript, ESLint, all 37 unit tests and the production build. All 18 desktop/mobile-viewport scenarios passed in CI; all five affected keyboard/mobile scenarios also passed against the public HTTPS app. Public asset hashes, routes, headers and browser security checks passed. The preceding beta 2's full 17-case public run remains separately recorded in verification.md. Full Chromium is used for media tests: the stripped Playwright headless shell crashes in its experimental speech availability API even without this app, while the full browser returns the supported local-language status safely.

Software input, synthetic camera/audio streams, and mobile viewport emulation do not establish physical hardware quality. Physical iPhone/Safari and actual camera/microphone/input hardware still require device testing.

Live app · Public source · Update details · Catalog sources

The earlier beta 3 walkthrough remains historical evidence of that release. It does not demonstrate the 1.1 changes. Earlier source tags are unchanged.

OpenAI Codex and specialist coding agents assisted development and verification. They are AI tools, not human team participants. Third-party packages and fonts retain their licenses in THIRD_PARTY_NOTICES.md. A public release does not establish event submission.

TESTDRIVE 1.1.0-beta.2 — guided automatic checks

TESTDRIVE 1.1.0-beta.2

This update makes an inspection easier to start and completes more checks from observed input. One click starts the automatic scan; saving a result takes the user to the next pending check.

• Four manufacturer-sourced US keyboard profiles: ASUS Zenbook DUO UX8406CA, Apple MacBook Air 13-inch M2 (2022), Dell XPS 13 9315, and Lenovo ThinkPad T14 Gen 4. A supported confirmed Windows import selects a matching layout; manual choices are remembered locally and cleared with local data.
• Explicit keyboard capture with visible active state, per-key press/release progress, triple-Escape, mouse Exit, 60-second inactivity and focus-loss exits.
• An eight-step pointer challenge automatically checks target movement, clicks, both scroll directions and dragging.
• Four-second camera and five-second microphone samples automatically check received media and stop capture. Optional installed on-device speech recognition can match the displayed phrase. No cloud transcription fallback or language-pack download is used.
• Battery/charger readings where available, a pausable display color tour, clearer instructions, optional human feedback and next-check navigation.

The laptop catalog is a reference database, not proof of current hardware specifications. Exact model selection needs a supplied/confirmed model or a manual choice. Operating-system reserved keys may still open system actions because a website cannot intercept keys the OS withholds. Sound detection does not claim word recognition; picture/sound quality and display defects still need human judgement.

Camera/microphone access starts only after the relevant action. Temporary media and normalized reports stay in the browser unless the user exports a report. Firebase serves the static app; no account, inspection backend or API secret is required.

Verification

TypeScript, ESLint, all 36 unit tests, the production build and all 17 desktop/mobile-viewport browser workflows passed locally and in CI. All 17 workflows also passed against the public HTTPS app. Public security-policy checks, deep routes, missing-file 404 and exact deployed JavaScript/CSS hashes passed. Current results are recorded in verification.md. Full Chromium is used for media tests: the stripped Playwright headless shell crashes in its experimental speech availability API even without this app, while the full browser returns the supported local-language status safely.

Software input, synthetic camera/audio streams, and mobile viewport emulation do not establish physical hardware quality. Physical iPhone/Safari and actual camera/microphone/input hardware still require device testing.

Live app · Public source · Update details · Catalog sources

The earlier beta 3 walkthrough remains historical evidence of that release. It does not demonstrate the 1.1 changes. Earlier source tags are unchanged.

OpenAI Codex and specialist coding agents assisted development and verification. They are AI tools, not human team participants. Third-party packages and fonts retain their licenses in THIRD_PARTY_NOTICES.md. A public release does not establish event submission.

TESTDRIVE 1.0.0-beta.3 — OFFGRID beta

TESTDRIVE 1.0.0-beta.3

Know the device before you buy it.

Open the live app · Public source · Immutable implementation

This beta provides an inspection-to-report workflow for computers, phones, and tablets. It combines qualified browser information, bounded processing and graphics samples, guided input/media/display checks, local English Windows specification import, and cautious seller-claim comparison. Reports retain sources, uncertainty, notes, and attempts, and can be saved locally, moved as validated JSON, or printed.

Watch or download the silent 60-second walkthrough. It shows the actual public app with labeled synthetic specifications, seller claims, and software inputs. It does not claim physical-device verification.

Camera/microphone access starts only from the relevant action. Temporary media is not uploaded to an application backend. Inspections remain in the browser unless exported. Firebase serves static assets, including self-hosted fonts; no account or inspection database is required.

Beta 2 added schema-safe keyboard fallbacks when browsers omit physical key codes. Beta 3 improves printed reports by keeping headings and evidence rows readable across pages. Earlier source tags remain unchanged.

Verification

• Implementation commit: d7eea4973711164ee758c70eacab9e383b921fdc; immutable tag: v1.0.0-beta.3-offgrid.
• Final GitHub CI: reproducible install, typecheck, lint, 21 unit tests, production build, and all 7 desktop/mobile-viewport browser scenarios passed.
• Public HTTPS/deep-route behavior, real missing-file 404, security/permission headers, cache policy, and public asset hashes were verified against the checked build.
• All 7 final public workflows passed in 39.2 seconds. A fresh production browser ran all 3 automatic observations under the content policy with no CSP violations, failed requests, third-party requests, or browser errors.
• The actual synthetic browser PDF spans 7 A4 pages; all page renders passed visual review without clipping, overlap, orphan results heading, or split evidence rows. Physical printing was not tested.
• Screenshots and any walkthrough explicitly label synthetic data, software input, and viewport emulation. Detailed production evidence is in docs/evidence/production and docs/verification.md.

Limits

Physical iPhone/Safari behavior and real camera/microphone/input quality have not been independently verified. Browser hints are not exact hardware specifications. Imported reports are user-supplied and are not authenticated. Timing depends on the environment. Battery wear, disk health, ownership, hidden defects, pairing/QR relay, accounts, and cloud report sync are outside this beta.

TESTDRIVE is an inspection record, not a device certification or purchase guarantee. A public app/release does not establish that an event submission was completed.

OpenAI Codex and specialist coding agents assisted development and verification. They are AI tools, not human team participants. Third-party packages and self-hosted fonts retain their licenses; see THIRD_PARTY_NOTICES.md.