Changes
What shipped, what moved, what we learned
Every entry references its source commit. Evidence claims carry the same stage labels the product uses, and limitations are part of the entry — not a footnote we skip. RSS
Click a gap in the feed, land on that gap
Clicking a gap card in the Activity feed now takes you inside that workflow's review and puts you on the gap itself — scrolled into view and briefly highlighted — instead of leaving you to find it by eye.
To make that possible the review gained a real gap list: every gap for the workflow in one place, worst first, resolved ones kept but quieter. It also works on its own — click any gap to select the step it belongs to.
Cards that summarise a burst ("19 gaps found") open the list at the top rather than pretending to point at one gap, and a link to a gap that no longer exists says so in one quiet line.
Known limitation: The landing works from the feed. Reviews are not yet reflected in the address bar, so refreshing or sharing a link does not reopen the review — that needs its own change and is tracked.
One selection truth on the canvas: the ring follows you everywhere
Wherever a workflow gets selected — clicking its star, clicking an activity card, the empty-panel button, Eve's tour, or returning from a workflow dive — the canvas now shows the same glowing selection ring, and the inspector, the neighborhood dimming, and the ring always agree.
Keyboard users gained real selection for the first time: Tab to a star and press Enter to select it (inspector included), press Escape to deselect. Escape is polite — if a dialog, tour, or tooltip is open, it closes that first; selection clears on the next press.
Dragging a star selects it cleanly (never a second ring), and stars can no longer be accidentally deleted from the keyboard.
Known limitation: On the classic system page, clicking an already-selected star now deselects it (uniform with every other surface — previously it did nothing). Multi-select never existed as a real feature and is now explicitly off.
Activity feed groups gap bursts; empty proof panels gained a real next step
When Eve finds six gaps in one workflow in one sitting, the activity feed now says so in one line — "6 gaps found — 2 high · 4 medium" — instead of six identical cards. Runs, gap outcomes, and research findings keep their own lines because each one carries unique information.
The Runs and Evidence panels no longer stop at explaining what would fill them: an empty panel now carries a one-click "Start with the weakest workflow" action that selects that workflow on the constellation, so the first sandbox proof is two clicks closer.
Known limitation: The one-click action appears on the constellation level only — inside an open workflow the panels stay explanatory, since the target star is not on screen there.
“Watch it change” now rewinds reliably
The landing’s interactive workflow used to lose its reasoning-agent circle after a visitor chose Promote or Return and then scrolled back to reread the v7 baseline. A clean top-to-bottom visit looked correct, which made the defect depend on interaction history rather than screen size.
Decision animations now hand the scene back to the scroll timeline instead of destroying its instructions. Rewinding, leaving and re-entering the section, rapid decision changes, and switching reduced motion all restore one coherent workflow.
Known limitation: This corrects the public illustration’s animation lifecycle; it does not change workflow execution, evidence, authority, or production-readiness claims.
The inspector speaks like a colleague
The step inspector stopped saying the step’s name twice and stopped filling rows with placeholders that carried no information (“Open gap: Open proof gap” is gone) — a row now appears only when it actually says something, and real gap titles and next actions always do.
Connector approvals now come with Eve’s recommendation: one clear “Approve for this step” choice up front, with the narrower and broader scopes a click away under “More options”. Running a sandbox proof remains the single primary action.
Two small honesty fixes: the confusing “Redacted” tag next to a fully visible account email is gone (the panel already explains that Eve never receives secrets), and raw platform IDs no longer headline connected accounts.
Known limitation: Resolving a platform ID to a real person’s name needs provider profile data — until that lands, the account line names the provider and keeps the ID on hover.
A new workspace greets you like a colleague, not a form
The first thing Eve says to a brand-new organization is now one warm question — “What does your company do?” — instead of a system inventory of everything she doesn’t know yet. Once real context exists, her informative recap returns, because by then it has something to say.
The message box asks plainly too: “Tell Eve what your company does and how the work gets done...” — the bottleneck-and-boundary jargon is gone from first touch.
The status pill in the header now reads “Just getting started · you approve everything” — the same honesty as before, in human words. And empty panels (Runs, Evidence, Activity) now name the exact control that fills them instead of only describing the absence.
Known limitation: Empty panels point at the “Run sandbox proof” control by name; a one-click button in the panel itself is a follow-on.
Honest pricing: a real 300-credit allowance replaces "unlimited"
Pro now means something checkable: 300 credits a month, automatically added on every paid invoice. Before this release the page said "Unlimited Eve conversations" while the billing system never delivered a single monthly credit — a paying subscriber would have hit a paywall error on day one. Both halves are fixed: the promise is honest and the delivery exists.
Every credit price now covers what the action actually costs us to run at current model rates — deep score 15, Eve turn 5, sandbox dry-run 10, gap resolution 15, step re-score 15, export 2. The table on the pricing page renders from the same constants the platform charges, so the page and the meter cannot drift apart.
The Team tier is paused until multi-seat genuinely exists, the "Most popular" badge is gone (we have no customers to base it on — the Founding 50 offer stands in its place: first 50 subscribers lock $49/mo for life), and cost ceilings are enforced on every metered lane so no single request can turn into unbounded spend.
Known limitation: Credit packs and the annual plan are code-ready but dark until their Stripe prices are minted; in-app pack purchase UI is a follow-on — depleted Pro users top up via support until it ships.
Eve gets her surface
Eve’s chat history now sits on a proper card in both themes. In dark mode her messages used to float as naked text over the canvas labels behind them; the card gives the conversation its own quiet surface without borrowing drop shadows to do it.
When Eve thinks in several passes before answering, the transcript used to stack a duplicate “REASONED” row for each pass. A turn now carries one calm “Reasoned” line — expand it and every pass of her thinking is still there, in order, with honest timing shown only when it was actually measured.
The reasoning labels also stopped shouting: “Reasoned for 12s” now reads as quiet metadata instead of an uppercase section header.
Known limitation: The live “Eve is thinking” narration during a turn is unchanged by design — merging applies only once the turn settles.
The constellation becomes the hero
Workflow scores now paint a continuous color ramp instead of broad buckets — a 46 and a 69 used to render the same amber; now the color moves with the number, and teal is earned only as scores genuinely strengthen. Star size follows the score too, so your organization’s strongest and weakest work is visible at a glance.
Eve’s mark at the center of the constellation grew substantially — she is the heart of the system and now reads like it. In light mode her mark flows deep teal into mint instead of the muddy blend it had.
The constellation now fills the view when it opens instead of floating in a small band of empty space.
Known limitation: Sizing and color respond to Eve-estimated scores today; as sandbox and production evidence lands, the same visual language carries measured results.
The canvas gets its volume right
Open-gap counters on workflow steps were the loudest thing on the canvas — a solid red pill shouting over everything, even though gaps are normal work in Everform. They are now calm bordered counters that keep their severity color without dominating the view.
Two different facts used to wear the same amber warning chip: “this step’s output gets human review” and “changes here need sign-off”. They now speak distinctly — Review/Supervised for how the step runs, Approval required for what changes need — so a glance can’t confuse them.
Buttons on the draft bar gained clear ranks: one filled primary action, quiet secondary and text-only tertiary ones, so what to do next reads from form alone.
Known limitation: The companion constellation redesign shipped the same day after founder visual review — see “The constellation becomes the hero”.
A calmer canvas: one signal language, and unscored work reads as potential
Selecting a workflow step no longer stacks the same facts twice — the floating summary strip is gone, and the step’s signal chips now lay out in one readable row where labels like “Approval required” render in full (hover any chip or step name for the untruncated text).
Workflows that haven’t earned a score yet no longer look dead: their constellation stars wear a dashed “awaiting first proof” ring — the same estimated-versus-proven visual grammar the rest of the product uses — instead of a gray dash that read as disabled.
Constellation cluster regions now show their family name at rest, and stepping back up from a workflow keeps that workflow selected so the canvas and the inspector agree on where you are.
Earlier the same day, a bug wave fixed stale views after switching organizations, chat quick-actions overlapping Eve’s replies, a see-through menu in light mode, and the unbranded 404 page.
Known limitation: Connective edges between workflows only render where real data contracts exist — sparse constellations stay visually sparse rather than inventing links.
A calmer brand field and cleaner landing choreography
Proof and Trust now share Pricing’s quiet Everform-mark pattern and cursor-lit spotlight, without bringing the landing page’s workflow constellation into those calmer reading surfaces.
The landing’s workflow annotations, evidence cards, and Watch it change scene now keep separate visual lanes across full-screen and responsive layouts; the italic subtitles also return to the shared teal hierarchy.
Known limitation: The interaction is intentionally static for coarse pointers and reduced-motion visitors.
The complete launch site — proof, trust, changes, and honest checkout
Four new pages joined the landing: the Evidence Room (/proof) where one real dogfood workflow will be replayed as recomputable receipts, the trust ledger (/trust) where every claim carries how we know it and the gaps get equal prominence, this changelog with its RSS feed, and rewritten privacy and terms naming who is legally responsible.
Paid checkout went live only after verifying the actual Stripe prices — the audit found the Team plan’s billing config was missing entirely, so its checkout had never worked. Fixed, verified, then enabled.
The site is now machine-readable too: /llms.txt describes Everform truthfully for AI agents, including the honesty constraints they should relay.
Known limitation: The /proof record is pending its first production export; /about ships when the founder pages are complete.
The landing page is launch-ready
The scroll journey is ~24% tighter without losing a story beat. The hero now opens with the problem — "Your business runs on knowledge nobody wrote down" — before the category statement.
A new proof section discloses our dogfood setup honestly: the two companies running on Everform are founder-owned pilots, not customers, and the page says so.
Launch guards landed: dead GitHub links removed while the repo is private, plain-language privacy and terms drafts published (counsel review pending), and dishonest cancel-copy fixed on pricing.
Known limitation: Legal pages are drafts pending counsel review.
"Watch it change" — the organizational-CI loop as a film
The landing now shows the whole improvement loop: a working workflow, a tempting cheap change that breaks guardrails in front of you, Eve's hybrid revision growing around the failure, a 10% canary lane behind a sandbox boundary, and the owner's promote-or-return decision re-routing live particle flow.
Every number derives from one illustrative case record and is labeled as illustration — the act never pretends to be production evidence.
Production-truth spine: signed run witnessing, deployed dark
Runs can now be cryptographically witnessed at the platform boundary (Ed25519-signed attestations) and written into readiness evidence as directly observed — the machinery that will let "it works in production" be a provable claim instead of a sentence.
Deployed fully dark: no production evidence is minted until the enablement flag flips after verification.
Known limitation: Flag off in production; no anchors minted yet.
Porcelain — light mode for the platform
The signed-in app has an opt-in light theme designed as its own material world (matte porcelain, gel-teal accents) rather than an inverted dark theme. Dark stays the default and is pixel-locked.
Programmatic WCAG contrast checks now gate every themed token pair.
Eve’s proposals now apply and re-score in one loop
Approving one of Eve's workflow edits now applies it atomically and re-scores the workflow live on the canvas — the full see-it, approve-it, watch-the-score-move loop.
The apply path is guarded against version drift and double-application; failures surface honestly instead of pretending success.
Security: the billing table could be self-awarded. Now it can’t.
A platform audit found that a signed-in user could in principle write their own billing-credit rows. The hole is closed at the database-policy layer, and the audit that found it also hardened adjacent write paths.
We publish notes like this because a trust page without incident honesty is decoration.
Eve now runs on GPT-5.6 Sol
The platform default model moved to OpenAI's GPT-5.6 Sol for Eve's chat and drafting. Structural scoring is unaffected — it makes zero model calls by design.
Calm Precision — a design elevation across the whole canvas
Constellation focus choreography, an owned Eve message layer, a three-zone inspector, evidence-stage marks, and a strict motion budget landed together — hierarchy before decoration, everywhere.
Scoring self-heals across deploys
Workflow scoring jobs interrupted by a deployment are now fenced by generation and swept back to completion by a five-minute cron — a score request can no longer be silently lost to bad timing.