Your Brain in your own AI tools, honest prices, and night tasks that finish
Product- Brain, Your tools: an owner or admin makes a key and pastes one command in a project folder. It adds your Brain to AGENTS.md, CLAUDE.md, .cursorrules and the other files AI tools read, inside a marked section, and leaves what you wrote there alone. The old install command named a package that is not ours; it is gone.
- Prices and refunds now say the same thing on every page as the Terms do: Solo, Pro and Max, the first 14 days free, no share of your sales, and a 7-day cooling-off right on your first charge in place of a money-back promise the Terms did not give.
- Night tasks no longer stop halfway on a busy night. They run five at a time inside a time limit, with up to three passes a night, and a task that cannot finish is marked failed with the reason instead of staying "running".
- Long S.I.D conversations are now summarised in the background as designed, so S.I.D keeps track of a long chat. Until now that job never ran in production.
- Settings and the docs say which AI models a paid plan runs on our account (Qwen) and which need your own key.
- Usage now shows where your credits went. Every charge is counted against the left-menu section you were working in (Studio, Customers, Build and the rest) and the AI model it ran on, and one row of filters narrows the whole page by date, organization, project, section or model. Owners and admins can see the whole team. Until now the page counted S.I.D chat and Build only; credits spent anywhere else were charged but not shown.
Clear answers when a limit is hit, and an account page that knows your plan
Platform- When S.I.D cannot run a request because of credits, it says which limit was reached (this month’s workspace allowance, the daily limit, your balance or a model your plan does not include) and gives you the button that fixes it, instead of an error code.
- Billing now shows how much of this month’s allowance you have used and when it resets. Credits you top up can be spent beyond the plan allowance.
- The Account page shows your real plan. Pro customers are no longer asked to upgrade, and the email note only mentions Google when you signed in with Google.
- Docs, Help, Changelog and Status read as normal web pages when you are signed out.
- The sign-in and sign-up forms keep what you typed while the page is still loading.
Hands mode on every page, and S.I.D listens everywhere
Product- Hands mode controls any Ikigai ADE page by hand. Point, pinch to click, pinch and move to scroll. Anything that deletes, publishes, sends or pays has to be held until a ring fills, so it cannot be triggered by accident.
- Gestures now has five boards: Brain, Build, Distribute, Market and Executives. Cards open the project, site or page they point to.
- The S.I.D page has a HANDS button that opens Gestures beside the console, and its gesture camera now runs on the same on-device tracker.
- Voice input in the S.I.D panel works on every app page, not only the S.I.D and Model Fleet pages.
- S.I.D no longer mistakes an ordinary sentence for a command. Asking about maintenance, a brain, an agent or the status of a launch now gets an answer instead of a page change or a canned reply.
Gestures: move your brain with your hands
Product- A new Gestures tab lays your brain out over your camera as orbs and cards. Pinch to grab, throw to dismiss, stretch a card with both hands to read it, crush it for the gist.
- S.I.D mode narrates what your hands do in S.I.D’s own voice and sends crushed cards to S.I.D for a summary, so the answer comes from the same brain and lands in the same conversation as everywhere else.
- Hand tracking runs entirely in the browser. Camera frames are never uploaded, and the deck is locked to Ikigai ADE’s own origin by its security policy.
Free tier fixed, and the executives got faster
Platform- A free workspace was restricted to two models that had been retired upstream, so a new account was refused on its first request. Free workspaces now run on live models, including the one behind the single-instruction asset flow.
- Six surfaces each opened a conversation on a retired model and paid for the failure before reaching a working one. Every surface now reads one shared default.
- A provider that returns success and then fails inside the response stream now moves to the next model instead of passing the error on. The wait before trying another model was halved.
- The shared learning layer now quotes only curated knowledge, never internal operating telemetry.
- The free tier was later replaced: since 30 September 2026 every plan is paid (Solo, Pro and Max), with the first 14 days free.
One Build surface and enterprise sign-in
Platform- Build unified into a single surface in the web app.
- Enterprise single sign-on, per-workspace provider identity, and an admin audit log.
- A customer workspace now runs on its own provider account rather than sharing the platform subscription.
Sellable catalog and the ADE command layer
Product- Product catalog and cart with real stock reservation, so a workspace can sell what it builds.
- Command palette, problems panel and click-to-line diagnostics in the development environment.
- One-click meeting attendance from the desktop assistant (the desktop assistant is not offered to customers yet).
Immutable releases and rollback
Security- Publishing now keeps every previous release instead of replacing it, so a rollback is a pointer change.
- Closed a path that could serve workspace files without a valid token.
- Tenant isolation re-proved against production with no secrets required in the pipeline.
Four pillars, one program
Product- Three competing navigations reconciled into the four pillars: Build, Distribute, Market and Sell.
- Distribute surface added, built on real publish history rather than a mock.
- The development environment publishes projects. Opening local repositories and running tasks are not available in customer workspaces yet.
- Studio renders real videos from the motion templates (not yet available in customer workspaces).
Credit top-ups and honest model health
Pricing- Self-serve credit packs for paid plans, priced at or above the plan rate so a subscription is never undercut.
- Model health now performs a real billing probe instead of assuming a configured key means a working model.
- Outreach lead discovery and email sending repaired end to end.
Cockpit pages and the dark command center
Design- Every authenticated page moved to the midnight and gold design language with a single coherent shell.
- New pages: system status, changelog, help, reports, notifications, search, and a profile page per executive.
- Campaign drill-down added under Outreach with live send and reply stats.
Multi-tenant isolation hardening
Security- Row-level security verified across every tenant-scoped table.
- CRM contact and deal writes are now scoped to the owning workspace.
- Founder intelligence endpoints tightened to admin-only access.
Product truth and brand-safety gate
Platform- A single source of truth now drives pricing, plan names, and identity across the product.
- A continuous brand-safety check blocks any drift before it ships.
Pricing simplified to Free and Pro (since replaced)
Pricing- At the time: two tiers, Free and Pro, with Pro adding a share of revenue closed by managed channels.
- Since replaced: on 30 September 2026 pricing became Solo, Pro and Max with no free plan and the first 14 days free, and on 2 October 2026 the share of sales was dropped. You pay one monthly fee.
Personal Brain Sync
Product- Announced here; the working install arrived on 5 October 2026 as Brain, Your tools: a key and a one-line command that adds your Brain to the instruction files your AI tools read.
- Per-tenant memory stays isolated; only generalized, eval-validated agent improvements are shared.
Intelligence OS dashboard
Product- The dashboard became the proof surface for the executive suite: KPIs, signals, and live activity.
- S.I.D routes work to the right executive and every action leaves a trail.