Changelog

Everything new in Neural Garden — features, fixes, and improvements as they ship.

Latest

The iPhone home-screen app, fixed up

If you use Neural Garden from your iPhone's home screen, taps now land exactly where you touch and the screen no longer jumps around while the app opens.

  • Taps and buttons line up with your finger again on every launch
  • The app fills the screen down to the bottom edge, with no empty strip
  • The top of the screen clears the Dynamic Island, and the status bar matches your canvas in light and dark mode
  • The command bar sits back down by the bottom corners of the screen
  • Already added the app to your home screen? Remove it and add it again to get the new layout

Getting ready to open the doors

This release is public-launch polish: clearer sync status, easier help, and a batch of dark mode and mobile fixes. Garden invitations are also more reliable, and a session that expires now says so instead of failing quietly.

  • The sync pill is now a real button on every screen size, with a live status you can read and a “Sync now” you can tap
  • Reach Help or “Report a bug” from the profile menu or the command bar, with your device details filled in for you
  • Inviting the same person to a garden twice no longer leaves a stuck or duplicate invitation
  • A session that expires shows a clear “Session expired” message instead of quietly failing
  • Dialogs, cards, the footer, and pinch-to-zoom are fixed across dark mode and phones

A gallery of published gardens

A garden you publish can now be listed where anyone can find it, not just the people you hand a link to. Add yours from the Share panel and it appears with a picture of the garden itself.

  • Browse the gardens people have chosen to list under Gallery, and search or narrow by tag to find one
  • Put your own published garden there with “Show in the gallery” in the Share panel, under a category and a few tags
  • Each listing shows a small picture of the real canvas, so you can see what a garden is before opening it
  • Featured gardens lead the list, and you can take yours back down whenever you like

Neural Garden is in closed beta

New accounts now start on a waitlist while we open the door a few people at a time. Signing in still creates your account — you wait on the waitlist page instead of your garden until you are approved.

  • Tell us who you are and how you would like to use Neural Garden when you join the list
  • An email arrives the moment you are approved, and the sign-in page it links to already has your address filled in
  • Leave the waitlist page open and it carries you into your garden as soon as your invite lands
  • Already invited to someone else’s garden? You skip the list entirely

Search finds the right thought first

Text search used to answer with whatever you had touched most recently, which buried the thought you actually named. It now leads with the closest match to what you typed, and it keeps up in a large garden.

  • A thought whose title matches what you typed comes first, ahead of anything that only mentions it in passing
  • Results read the same whether you are online or offline
  • Thoughts you have moved to the Trash no longer turn up in search while you are offline
  • Typing in the search box stays responsive instead of lagging a keystroke behind

Work you do offline lands where you left it

Several ways an offline change could quietly go missing are fixed, and catching up after a stretch away no longer takes a quarter of an hour. Large gardens open faster too.

  • A thought you make offline stays put when you nest it or drag it afterwards
  • Delete a container offline and the thoughts inside it come back onto the canvas right away, instead of waiting for you to reconnect
  • A connection that flickers mid-sync can no longer drop a change or apply it twice
  • Coming back online after a long stretch away catches up in one pass rather than trickling in over many minutes
  • A big garden opens noticeably faster, and connection suggestions keep up as it grows

Find your way around a shared garden, and follow its links

A shared garden now opens with a list of the places you named, so a visitor can jump straight to one instead of panning around looking for it. Links look and behave like links — on the shared page and on your own canvas alike.

  • A shared garden lists its regions and bookmarks — pick one to jump to it, or open the list from the bookmark button on a phone
  • Links on a shared page are clickable now, and show the page’s own picture, title and site name instead of a line of grey text
  • A link whose page offers no picture shows the site’s icon instead, in your garden and on the shared page
  • Link titles read properly again — punctuation a site writes into its own title no longer shows up as code
  • A picture inside a thought stays put while you zoom a shared garden on a phone, instead of blinking on and off

Take a garden someone shared with you

A published garden is no longer only something to look at. Copy the whole thing into your own garden — signing up on the spot if you don’t have an account yet — or take the embed code and put it in a page of your own.

  • Copy a shared garden into your own, straight from the page you’re reading it on
  • No account yet? Signing up from the link brings the garden with you, and the copy is waiting on your canvas when you arrive
  • “Copy embed code” is on the shared page now too, so you can put someone’s garden in your own site without asking them for it
  • A shared garden moves like your own — scrolling keeps its momentum on a phone, and cards fade between zoom levels instead of snapping
  • A thought with a picture keeps showing it as you zoom out, and a long name on a small card is cut cleanly at two lines

Published gardens show their pictures, videos, and links

A shared garden's page now looks like the real canvas — pictures and videos show instead of just naming their file, and links to other thoughts actually take you there. Downloading a shared garden and importing it back in keeps those links working too.

  • A published picture shows the picture itself, not just its file address
  • A published video plays right on the page, with a link back to the original if it can't
  • A link to another thought on a published page takes you there instead of doing nothing
  • Downloading a shared garden and importing it back in keeps its links working, not just their look
  • Copying a garden no longer drops the words of a link pointing outside the copy

Share a garden as a link anyone can open

Publish any garden, or a single thought inside it, as a read-only link — no account needed to view it. It looks and feels like the real canvas, and only shows what you meant to share.

  • Publish a garden or one thought inside it as a link, then copy it, embed it on another site, update it, or take it down whenever you like
  • The shared page looks like the real canvas — zoom levels, colors, region highlights, and connections a reader can click to follow
  • A published thought shows its typed details, like a recipe's ingredients or cook time, while anything you've marked hidden stays hidden
  • Readers can download a copy of what they see and import it into their own garden
  • The page credits you by name, shows when it was last updated, and reads in light or dark to match the reader's own preference

An import lands inside one thought, not scattered across your canvas

Importing a file no longer drops its contents loose across your home canvas — everything lands inside one new thought named after the file. A garden you downloaded from a share link also keeps its typed properties when you bring it back in.

  • Importing a file creates one thought named after it, with everything from the file nested inside instead of scattered across your canvas
  • An imported thought with the same name as one you already have is renamed automatically so both stay distinct
  • A garden downloaded from a share link keeps its typed properties, like a recipe's ingredients, when you import it back in

Regions hold their shape, and share it right away

Resizing, fitting, and arranging a region now behaves the way you'd expect, and changes reach everyone in the garden immediately.

  • Resizing a region never leaves your thoughts hanging outside it, and its corner handle is easier to grab
  • Fit region to contents now matches exactly what's on screen, growing a region as well as shrinking it, in every layout
  • Switching a region's layout — Grid, Outline, and the rest — now reaches everyone in the garden right away, instead of waiting for a reload
  • Folding a thought's children re-arranges the region around the gap, and unfolding them no longer drops cards on top of each other
  • Deleting a region hands its thoughts back where they belong, and only a region you've sized by hand keeps that size for everyone else

Regions keep to what you put in them

Wrapping a selection in a color region now leaves room at the top for the region's own handles, without reaching over a thought you didn't pick. Regions also lay out a single thought, and leave AI suggestions alone entirely.

  • Create Region leaves a gap at the top for the region's drag handle and Arrange icon, so they no longer sit on the thoughts you just selected
  • That gap stops short of any thought you did not select, so making a region never takes a neighbour in with it
  • A region holding one thought now lays that one out — Re-arrange now used to do nothing there
  • Regions leave AI suggestions alone: never laid out, never carried along when you drag a region by its grip, and never counted when a region works out how big it has to be

Region names move out of the way

A region's name now hangs above its top edge under every layout, so a long name can no longer sit on top of the handles beside it. Fit region to contents sizes a region to what's actually on screen, and a region with no arrangement grows to hold one you drag into it.

  • Every region's name sits above its top edge, with the drag handle and the Arrange icon just inside it
  • The name is a label now rather than a handle — drag the grip at the top of a region to move it and everything inside
  • Fit region to contents measures the thoughts as they are drawn at your current zoom, so it can grow the region as well as shrink it
  • Drag a region into a region that has no arrangement and that one grows to hold it, instead of leaving it hanging over the edge

Color regions arrange what's inside them

A color region now has its own Arrange menu: pick a layout and the thoughts inside it lay themselves out within the box. It keeps itself tidy from then on, re-flowing when you drop a thought in, drag one out, or resize the region.

  • The icon at a region's top-right corner opens its Arrange menu — Grid, Outline, Tree, Radial, Force-directed, a Sort by list, Re-arrange now and Fit region to contents
  • Manual puts every thought back exactly where you had placed it by hand, even if a collaborator is the one who switches back
  • Drag a thought across the canvas and the region it will land in lights up; let go and that region re-arranges around it
  • Resize a region and its contents re-flow as you drag, and the region grows to hold everything rather than leaving a thought outside it
  • Draw a region inside another and the outer one treats it as a single block, and each gesture and the re-arrange it sets off undo together with one Cmd+Z

Regions show up as other people draw them

A color region or bookmark someone else adds, moves, recolours or removes now appears on your canvas straight away. Before this you had to reload the page to see it.

  • Regions and bookmarks from everyone in a garden arrive live, with no reload
  • The Bookmarks panel updates in place as bookmarks come and go
  • A region's arrangement belongs to the garden, so everyone looking at it sees the same layout
  • Renaming a view now updates the Saved Views panel straight away instead of waiting for you to reopen it

More ways to arrange a selection

The Arrange menu for a selection now offers the same layouts as the whole canvas, plus a Sort by list to set their order. The layouts themselves lay out more evenly.

  • Outline joins Grid, Tree, Radial and Force-directed for a selection, with Tidy Up still a single tap
  • Sort by sits beside Arrange, so a layout follows the order you chose
  • Radial wheels sit centred on their rings, and a crowded ring grows to fit instead of doubling back over itself
  • Force-directed spaces cards by their real size, so large cards no longer overlap, and re-running it gives the same picture
  • Every Arrange mode is now in the command bar, so you can reach it with Cmd+K

Copying a shared garden makes it yours

Copy to My Gardens used to make a second view of the original, so editing your copy edited the other person's thought and losing access left it stranded in your garden. It now makes real thoughts that belong to you.

  • A copy is now your own — editing it never touches the original, and it keeps working if the garden stops being shared with you
  • Nested thoughts, connections and typed properties come across, and a type you don't have yet is added to your Types garden
  • You stay on your own canvas, where the copy lands below your existing thoughts and pulses so you can see where it went
  • A copied thought named the same as one of yours arrives numbered, so you can tell them apart in search and the [[ picker
  • A note the AI suggested but nobody accepted stays behind, and a message afterwards names anything else that couldn't come across

Suggestions stay suggestions

Thoughts the AI proposes no longer turn up as gardens in the sidebar before you have decided anything about them.

  • An unaccepted suggestion no longer appears in the My Gardens list
  • Accepting one adds it to the list straight away, and dismissing it removes it for good

Your hand-placed layout stays put

Arranging a garden by hand and viewing it in an automatic layout somewhere else no longer costs you the arrangement. Each device keeps its own layout mode, and Manual always brings back exactly what you placed.

  • Outline, Grid, Tree and the other automatic layouts never overwrite your Manual arrangement, on any device
  • Switching layout mode is now per device, so changing it on one screen no longer changes it for everyone else looking at the garden
  • Thoughts created while an automatic layout is on land near the centre of your Manual arrangement
  • Arrange on a selection (Tidy Up, Grid, Tree and the rest) can now be undone with Cmd+Z
  • If an earlier version overwrote a hand-placed layout, the command bar's “Layout: Recover hand-placed positions saved on this device” puts back the copy that device kept

Types stays out of the way

The Types entry lives in the sidebar's System section and no longer shows up as a thought on your home canvas.

  • The Types thought no longer appears on the home canvas after opening the app in a new browser or while offline
  • The name Types is reserved for it, so a new thought can't be given that title in your home garden

Steadier on a phone

Three small things that got in the way on a touch screen, fixed. The bottom sheet and the dock now behave the same however you reach them.

  • Tapping the number beside a thought's title no longer brings up the keyboard on a touch screen; the list just opens
  • Opening a thought from a search result now raises the panel to the half point, the same as opening it from a menu
  • The dock's buttons sit centred in their pills, and its shortcut icons centre in the bar instead of bunching to the left

Thoughts in many places, one click to the right one

A thought mentioned in many notes no longer gets a copy added to each note, and following a link takes you straight to it instead of asking where. When a thought lives in several places, the number beside its title lists them all.

  • Linking to a thought that already exists just makes the link, so a person mentioned in many notes stays in one place
  • Click a link, a backlink, or a nested thought and you go straight to it on the canvas, or it opens in the side panel when it lives in several other places
  • Click the number beside a thought's title, on the card or in the side panel, for a scrollable list of every place it's placed, with a filter once the list gets long
  • Cmd+K shows each thought once with a count of its places instead of one row per copy
  • Clicking the title of a nested thought now takes you to it

Edits that stick, and a canvas that behaves on a phone

Whatever you were typing in the side panel is saved when you close it or move on to another thought. Search finds the thought you named first, and dragging a collapsed thought keeps its hidden children with it.

  • Closing the side panel, or jumping from it to another thought, now saves what you were typing instead of dropping it
  • Drag a collapsed thought and its hidden children come along, so expanding it later opens them right where you left them
  • Cmd+K search lists thoughts whose title matches first, ahead of thoughts that only mention the words
  • On a phone, Fit to Screen zooms out far enough to show everything, and opening the bottom sheet slides the view over instead of shrinking it
  • The command bar tucks away while you type in any other field on a phone, and the Nested and Backlinks lists read at one size on the card and in the panel

Search inside a thought's lists

The nested thoughts and the backlinks under a thought each get a search field, on the card and in the side panel alike. It can also find the thoughts that mention this one in plain text and link them for you with one click.

  • Click the magnifier beside Sort by to search a list — type several words, put a minus in front of one to leave it out, or type:book to keep one type
  • Pills under the field suggest the types in the list; click one and it joins the search inside the field, with its own × to remove it
  • In Backlinks, the unlinked mentions pill shows thoughts that name this one without a link, and the check mark turns each into a real backlink
  • On a phone the field opens clear of the keyboard, and the command bar rises with the keyboard only when you tap its own box
  • On a computer the command bar keeps one width whether it's waiting or open, so its results panel no longer jumps when you close it

A bar that fits your phone, and links that work everywhere

The command bar at the bottom now follows your phone's rounded corners, opens with a proper ease, and steps aside for the keyboard. Links in descriptions and backlinks behave the same in the side panel as they do on a card.

  • The bar and toast messages sit clear of the display's rounded corners on iPhone instead of cutting across them
  • Opening the bar eases it open over a dimmed garden, with its icons sliding in from the side it grew from
  • Start typing and the bar drops to just above the keyboard and widens to the screen edges, then eases back when the keyboard goes
  • Long-press a link on a card and its menu opens beside the link, not somewhere off screen
  • Links inside a nested thought's preview show their names again, and you can click one to go there
  • Backlink rows and renaming a thought work the same in the side panel as they do on a card, Merge offer included

Connect and create from the side of a card

Select a thought and a round handle pops out of each side. Click it or drag from it to make a connection or a whole new thought — at every zoom level, in place of the old four dots and the + button.

  • Drag from a handle onto another thought to connect the two, or onto empty canvas to make a new thought that arrives already connected
  • Pull from the left for a parent and from the right for a child, so the side you use decides which way the connection runs
  • Select several thoughts and the pair of handles serves the whole selection at once
  • Click a handle and you get a blank thought with its title box ready, instead of a placeholder name to rename afterwards
  • Suggestions have their own doors now — select a thought, right-click and choose Suggest Ideas, or press ⌘K — and they line up in a column beside it instead of landing on what you were typing
  • Cards reach their full detail earlier as you zoom in and stay legible further out, and the pill button that read Properties now reads Open in panel

Grok can look things up

With Grok as your AI provider, the assistant can now search the web and X when a question needs current or outside information. It also runs on xAI's newest model, so answers and suggestions get sharper across the app.

  • Ask the assistant something it can't answer from your garden — an address, the latest on a topic, what people are saying — and it searches, then folds what it found into the reply
  • Sources appear as small numbered links in the answer and as a Sources row underneath, so you can check where a fact came from
  • Two switches on the Grok tab in Settings turn web search and X search on or off
  • Each search is billed by xAI on your own key, and only the assistant ever searches — background suggestions never do
  • Type /ai in the side panel for a thought that isn't on the current canvas and the assistant now knows which thought you mean, and can write into its description

A cursor that lands where you point

The last of the iOS cursor trouble around links is gone, and the slash menu's text styles work again. A few small controls got easier to tap, and a lost picture now says so.

  • Glide the cursor with the keyboard trackpad over a [[link]], in a card or in the editor sheet, and it lands where you point instead of flying to the top of the screen
  • Type / in a description and pick a text style and it sticks, where before only Divider seemed to do anything
  • Small controls like the breadcrumb's home button and the description sparkle are easier to hit with a finger, without taking up any more room
  • A picture whose file has gone missing says so in its place, with a note to add the same picture again, instead of a broken icon or a blank

Ask AI from wherever you are

The assistant is now one step away from any thought, connection or blank patch of canvas, and it can write straight into a description. Pictures you add or generate also stop vanishing.

  • Right-click or long-press anything and the first menu entry is AI… — it opens the assistant already knowing which thought, selection or connection you meant
  • Type /ai while editing a description to ask for text or a drawing right at your cursor, and /image to drop in a picture of your own
  • Ask the assistant to add a summary or a few bullet points to a thought and it adds to the description without replacing what you wrote
  • Uploaded and AI-generated pictures now survive updates, and a picture a collaborator adds shows up for you without a reload
  • Permanently deleting a thought, or emptying the Trash, also clears out any picture nothing else still uses

Editing on a phone, without the fight

Placing the cursor, selecting text and moving around a description on iOS no longer trips over links, gestures or the editor sheet.

  • While you're editing, a [[link]] behaves like plain text — tapping it places the cursor instead of jumping away, and holding it starts a selection like anywhere else
  • Swipe to indent is a quick flick now, so dragging the cursor or a selection handle never nests the line by accident
  • Hold the spacebar to glide the cursor with the keyboard trackpad and it passes over links without jumping away
  • The editor sheet stays put while you select text inside it; the handle at the top still resizes it
  • One press of Backspace removes a whole link, and on a computer Cmd-click follows one while you're editing

One thought, one copy

Picking an existing thought from the title list sometimes left you with two of it — one you had to hunt down and delete. It doesn't any more.

  • Start a thought, type a name that already belongs to one, and picking it places that thought instead of quietly making a second one with the same name
  • This holds for a thought living in another garden, which is where the duplicates were coming from — and at every zoom level, including the compact cards on a phone
  • Duplicating several thoughts at once no longer leaves a copy that appears only after you reload
  • In the compact title list, the highlighted row stays where you put it, so Enter picks the one you meant

AI finishes your new thoughts

Create a thought and AI can fill in the rest — its type, a first line of description, and details it knows. A tap on the new sparkle does the same for any thought you're editing.

  • Create a thought on the canvas or from a [[link]] and AI picks its type, writes a short first sentence, and fills in details it knows — like a book's author
  • It only ever adds: nothing you wrote is touched, and when the AI isn't sure it leaves the thought exactly as you made it — turn it all off in Settings → AI autofill
  • A sparkle on its own line under the description completes any thought you're editing, on the canvas card and in the side panel
  • Descriptions are short and factual — no repeating the title they sit under, no describing the thoughts you can already see nearby
  • AI suggestions are easier to spot: brighter titles at every zoom, and they sit in front of nearby thoughts so Accept and Dismiss are always reachable

Sharing a guest can be trusted with

Sharing now hands a guest exactly the canvases you give them — nothing nested deeper, nothing by accident. Invites, roles and the Trash all follow the same rule.

  • Share a whole garden or a single thought; new invitations start as viewers, and you can change or remove a member's role from the Share panel later
  • Guests see only what you shared: lists, search, links and pickers all stop at the canvases they can reach
  • A guest's deletions go to the Trash where they can restore them; only the owner can permanently empty it
  • Your gardens and a collaborator's stay separate — pickers offer only thoughts belonging to the garden you are in, so content never drifts between accounts
  • The Share panel is roomier, shows each member's email, and Import/Export now says plainly what it does

Connect and disconnect from the menu

Editing a connection used to mean hitting a thin line with your cursor. Both jobs live in the menus now.

  • Connect To… links everything you have selected to one thought. Search for it, or create it on the spot, then say whether it is the parent or the child
  • Disconnect… removes a connection by name instead of by aiming at it. When a thought has several, you pick from a list
  • Long-press a connection on a phone or tablet to open its menu
  • Undo belongs to the canvas you were on, so it can no longer reverse something from a canvas you have left
  • Fixes: pressing Delete on a line that is not a real connection no longer fails silently, and the Disconnect menu no longer runs off the side of a phone screen

Dismiss a suggestion, and find deleted thoughts in the Trash

AI suggestions are sketches until you accept them, and they finally behave like it. Deleting a thought from inside another thought also stopped going missing.

  • A suggestion offers Accept, Dismiss and Delete. Dismiss turns it down and the AI offers you something else; Delete simply removes it
  • Select several suggestions and you get Delete on its own — aligning, locking or duplicating a sketch was never much use
  • Deleting a thought from inside another thought's list puts it in the Trash, where you can restore it, and the row disappears straight away instead of lingering until you reload
  • Connection arrows and labels sit on the line whichever line style you use, and hovering a line no longer covers the titles underneath it
  • Sharing is safer: an invitation to one garden reaches only that garden. Merging a type is now refused rather than quietly breaking every thought using it, and a merge survives going offline

Names that stay unique

Two thoughts answering to the same name made every [[link]] a coin toss. Aliases are now held to the same rule titles always were.

  • An alias has to be unique the way a title is — use one that is taken and you are offered a merge instead of a silent duplicate
  • The merge prompt reaches you from the card, the side panel and the zoomed-out title editor; renaming while zoomed out used to do nothing at all
  • Merging keeps more of what matters: it can no longer overwrite the surviving thought's text, aliases, types or properties
  • Fixes: a long press near the bottom of the screen no longer fires whichever menu item slid under your finger, and picking an alias from the title dropdown goes to the right thought

Steadier lists, smoother drags

Backlinks caught up with nested lists — their own rows, their own sort control, the same soft edges. Dragging on an iPad is smooth again.

  • Backlinks sit below a thought's nested list, each in its own row, so the two read as one card
  • Backlinks get a Sort by control of their own — Linked, Title, Type, Created, Edited, or any property — and each thought remembers its choice
  • A list inside a thought scrolls right after a two-finger zoom, instead of staying stuck until you tapped elsewhere
  • Dragging a thought is smooth again on iPad; the coloured type glow no longer redraws on every frame
  • Fixes: a full stop typed after a wikilink no longer opens the link picker, and expanding a collapsed branch no longer drops the thought on top of another

Nested lists you can sort and drag

The list of thoughts inside a thought became something you can work with. On a phone a row lifts under your finger the way it does in Reminders.

  • Right-click or long-press a row for Open in Panel, Move Out and Delete — or a backlink for Open as Garden and Open in Panel
  • On touch, hold a row and it lifts under your finger: drag to reorder, drop it onto another thought to nest it, or drop it on the canvas to lift it out
  • A Sort by control above each list — Manual, Title, Type, Created, Edited, or any property — with each parent remembering its own choice
  • Long lists cap their height and scroll instead of stretching the card down the screen, with soft edges showing there is more
  • Fixes: dragging a row downward lands where you dropped it rather than one slot too far, and lifting one finger out of a pinch no longer flings the canvas

Right-click anything, and move thoughts anywhere

Every thought and every selection now has a menu behind a right-click, or a long press on a phone. The headline command inside is Move Into…

  • Right-click a thought for Edit, Open as Garden, Add Child, Duplicate, Lock and Delete; select several and the group gets its own menu
  • Move Into… finds any thought and moves your selection inside it, or creates a new container on the spot — and what you pick stays exactly where your selection was
  • Create Region turns a selection into a labelled area, and Bookmark This View saves your exact camera position
  • The Nexus bar gained the same commands, acting on whatever you have selected
  • Menus fit the screen instead of scrolling — one near the bottom edge slides up until every item is visible

Pictures take root

Ask the AI for a picture and it lands on the thought you meant, or paste one straight into a description. However it arrives, you can spot it from any distance.

  • Ask for a picture in the Nexus bar and it arrives as an image thought, attached to the thought you meant
  • Paste or drop an image straight into a description — it renders where you put it and comes along in exports
  • A thought carrying a picture stays recognizable at every zoom, from a full-width cover down to the picture itself as the map marker
  • Images belong to the garden you add them in, so collaborators who share only that garden can see them
  • Grok connects properly now, and a Test connection button in AI settings checks any provider's key on the spot
  • Fixes: opening or closing a panel returns the canvas to the view you had

Sign in with just your email

Neural Garden no longer needs a Google or GitHub account. Type your email, click the link we send, and you are in — first visit included.

  • Enter your email and we send a one-time link; it works once, expires after 15 minutes, and there is never a password
  • The sign-in page leads with email now, with Google and GitHub below it
  • Continue with GitHub works again, after GitHub tightened how sign-in callbacks are validated
  • The frosted-glass top bar returns on the landing page, and matches across Help, Changelog and Graphs
  • Type glows on the canvas scale with each type's weight, so heavier clusters read stronger at a glance

Every command keeps its promise

Several commands in the > menu quietly did nothing. Now every entry does exactly what the toolbar would, and Undo covers nearly everything you do.

  • Every > command works: Undo, Redo, Select All, Zoom In, Zoom Out, Fit to Screen and Help used to do nothing at all
  • Undo now covers creating a thought, renaming, edits from the side panel, duplicating, nesting, connections, regions, group drags and every AI action
  • Dragging several thoughts together saves all of their positions; only the card under your cursor used to keep its place
  • Grok joins Claude, OpenAI and Ollama as an AI provider, and can generate images onto your canvas
  • Fixes: Help opens in a new tab rather than replacing your garden, and Fit to Screen frames your thoughts in the part of the canvas you can actually see

The Daily garden grows up

Getting in and out of the Daily garden is quick and clean now, and the journal column stands perfectly straight. Frosted glass gave way to solid surfaces that stay crisp at every zoom.

  • Entering a garden no longer flashes zoomed-out before settling, and leaving a big one for Home is quick instead of pausing for seconds
  • Coming back to the browser tab no longer freezes the canvas — the garden refreshes only when something actually changed while you were away
  • The Daily journal is one straight column, and gardens saved with the old skew straighten themselves out on their next visit
  • Zooming across detail levels pauses far less, and day cards no longer overlap while you scroll the timeline zoomed out
  • Panels, menus and toolbars sit on solid surfaces that stay sharp while you zoom; the blur was costing smoothness, most of all in Safari
  • Fixes: the browser Back button works during a drill-in, and switching gardens quickly can no longer paint the wrong one

Opening a thought is instant again

The first thought you opened after loading a garden could sit there for about fifteen seconds, even when it was empty. That is gone.

  • Opening a thought for the first time after loading a garden is instant, instead of pausing for around fifteen seconds
  • The cause: building your garden list was reading far more of the database than it needed, every thirty seconds, and everything else queued behind it
  • A thought now starts drawing from the copy saved on your device right away, rather than waiting on a bookkeeping request first
  • A failed request gives up quickly and falls back to your local copy instead of retrying for seconds
  • Zooming stays smooth: panels stop re-blurring the canvas behind them on every frame

Zooming works in Safari again

In desktop Safari, zooming shrank your thoughts along with the canvas instead of holding them readable. Fixed — and a tuning pass reset what a fresh canvas looks like.

  • Thoughts hold their readable size while you zoom in desktop Safari, exactly like every other browser
  • A timeline can no longer open stuck at a uselessly far-out zoom; it falls back to a clean reading-width fit
  • The Daily garden opens as a newest-first timeline on every browser and device
  • Smart detail starts off for a calmer default canvas, except in Town view where it is what gives the map its landmarks
  • Labels survive further into a zoom-out before collapsing to dots, and tree layouts space generations properly apart

Edits make room for themselves

Writing a longer description used to push a thought straight through the one below it until you nudged something. The layout re-flows the moment you save now.

  • Editing a description re-flows an arranged canvas as soon as you save, so a card that grew taller no longer overlaps its neighbour
  • The same applies when you edit from the side panel, not only by typing on the card
  • The coloured type glow stays on the thought it belongs to, instead of hopping to another and back every time you left the browser and returned

Real dates, and gardens that remember your view

Days, weeks, months and years now carry an actual date rather than leaning on their titles, and the canvas can sort by it — or by any property your thoughts carry. Every garden also remembers where you left it.

  • Day, Week, Month and Year thoughts have a real date property, filled in automatically from their titles on your next load
  • Sort by any property your thoughts carry — dates, numbers, text, yes/no — alongside title, type, created, updated and connections
  • The timeline's side rail follows whatever you sort by: dates label days and months, titles turn it into an A–Z index, types label each group
  • Timelines order by real dates now rather than title text, and the Daily garden opens at reading width on today
  • Every garden remembers its pan and zoom on each device, and coming back from a thought puts you where you were rather than centred on what you left

Big gardens open in a blink

A garden with hundreds of thoughts used to freeze for seconds after appearing, redoing a layout it had already saved. Suggestions also learned to step aside.

  • Large gardens open near-instantly — the canvas recognises when your saved arrangement is still correct and skips the work
  • Your device remembers how large each thought is, so only the handful on screen are drawn; a 300-day Daily draws the three days you can see
  • Auto-arranged gardens stop rewriting themselves on every visit, so manual arrangements no longer drift
  • AI suggestions stay out of your nested lists and out of bulk selections, and are swept away with the thought they belong to
  • One toolbar at a time: tapping a thought shows its own tools, drawing a selection box brings up the group tools
  • Connections meet a thought at its centre when zoomed out, so lines point at your thoughts rather than past them

The iPhone crash is gone

Zooming a garden with colour regions on could silently kill the Safari tab. Fixed at the root, and the whole zoom experience improved with it.

  • The iPhone Safari crash is fixed, and pinch zooming now does a fraction of the work it used to on every device
  • A long pinch no longer skips through part of the zoom range or snaps back when you reach the end
  • Region controls tuck away while you pinch and fade back when you lift, and are available at every zoom level
  • Thought cards hold their detail longer as you zoom out, and stay labelled dots to half the previous cutoff before becoming specks
  • Dark-mode glow blending is back: overlapping type zones brighten where they meet, rebuilt so it costs nothing
  • Fixes: a new thought can no longer vanish while it saves, and blank drafts always clean themselves up

The guide stays one step ahead

The guide now drafts two levels at once, so keeping a suggestion hands you the next three instantly. Suggestions became tidy chips that follow your selection at any zoom.

  • Each background pass prepares your next three suggestions and each of their next three, so keeping one opens the following doors with no wait
  • Suggestions are compact chips beside their thought, one readable size at every zoom; keeping one grows it into a full thought
  • Suggestions follow your selection — select a thought and they bubble out, click away and they sink, reselect and they are back
  • Quiet gardens: a list in Settings of gardens where suggestions never appear on their own, with Inbox, Daily and Types out of the box
  • Fixes: no more crash when selecting a thought with prepared suggestions inside a collapsed branch, and auto-arrange is snappy again

Your garden opens from your device

Every canvas paints from the copy on your device first, online or offline, with the server catching up quietly behind it.

  • Gardens and sub-gardens open near-instantly, in a few hundredths of a second rather than around a second
  • Work done offline survives a shaky reconnection — your device stays the source of truth until every change has uploaded
  • Fixed a crash on iPhone when slowly zooming into an area with a colour region
  • Dropping a thought onto something that already contains it now explains itself instead of just snapping back
  • Kinder to phones and batteries: less idle work while you read, less work per keystroke, and no doomed AI retries while offline

The guide learns you

The Socratic guide reads far more of your garden before it speaks — including a goal you write and every suggestion you have turned down.

  • Each suggestion is made knowing the thought's parent, its siblings, its children, its properties and the path back to the root
  • The guide is no longer questions-only: a planning branch gets next actions, a direct question gets an answer, an untested assumption gets a counterpoint
  • Give your garden a Goal and the AI steers every suggestion toward it, without ever rewriting it
  • Occasionally a Coach note appears with an observation about your thinking itself — a branch running deep with no synthesis, or drift from your goal
  • Your no's are remembered across sessions and devices, so an angle you turned down is never offered again
  • Pre-think prepares suggestions while the canvas is idle, so selecting a thought pops them out instantly

Offline covers your whole garden

Offline used to cover only each garden's top canvas — drilling into a thought showed an empty space. Every nested level comes with you now.

  • Drill into a thought while offline and its nested thoughts are there, even in sub-gardens you never opened online
  • Connection lines are part of the offline copy at every level, including links that reach into a sub-garden from outside
  • Fixed a background error that repeated in the browser console every few minutes while a garden was open

Layouts that keep themselves tidy

Auto-arranged layouts re-tidy themselves as you zoom, instead of overlapping until you nudged something.

  • Tree, grid and the other layouts re-tidy about half a second after a zoom changes how much detail thoughts show
  • Timeline measures off-screen thoughts before arranging, fixing the misplaced and overlapping cards
  • Picking an existing thought from the title suggestions at low zoom reliably becomes that thought, rather than creating your half-typed text
  • Backing out of a new thought without typing no longer leaves a blank one on the canvas
  • The merge dialog sits above the command bar on phones — its buttons were hidden underneath

Everything reads better

Long titles stay legible at every zoom, and the AI's answers arrive with real formatting instead of raw markdown.

  • Long titles get two lines before trimming on compact cards, and the full card always shows the whole title
  • Hover a thought whose title is cut off and the full title appears, so you need not zoom in to read a name
  • The AI's answers render properly — headings, lists, dividers and code, instead of raw ### and --- characters
  • The Garden Intelligence summary is a rich-text field now, so you can edit or rewrite it in place

Big gardens stop tripping over themselves

A garden with months of daily entries used to flood the server on opening and jam everything else. Saves are batched now.

  • Opening a large arranged garden saves the whole layout at once rather than a request per thought
  • Adding a date to the Daily garden is instant again, after those saves stopped exhausting the server's budget
  • The same applies to every multi-thought move: align, tidy up, arranging a selection, dragging a branch, and undo

At home on Windows and Linux

Ctrl+scroll zooms the way you would expect, and every shortcut hint shows your keyboard's keys rather than Mac symbols.

  • Hold Ctrl and scroll to zoom toward your cursor on Windows and Linux; it used to hijack the view with a Mac-only gesture
  • That eagle-eye gesture is Mac-only now, where the Control key is free, and is documented in Help
  • Shortcut hints match your keyboard: Ctrl+K and Del on Windows and Linux, rather than the Mac symbols
  • Redo works again everywhere, and Ctrl+Y also redoes on Windows and Linux
  • Keyboard shortcuts no longer break when Caps Lock is on

The AI becomes a Socratic guide

AI suggestions moved out of chat and onto the canvas. Press + on a thought and three suggested paths fan out as real, movable thoughts — sketched in dashes until you decide.

  • Press + and three suggestions fan out, drawn dashed so they read as sketches; keep one with ✓ and it becomes an ordinary thought, or dismiss it and a fresh path grows
  • Direction carries the meaning: + on the right explores where an idea leads, + on the left explores what it rests on
  • Suggestions are short questions, not essays, and the guide reads what is already around the branch so it never repeats you
  • Until you keep one, a suggestion is a read-only sketch — out of search, clustering and garden intelligence
  • A new page, /socratic, explains the thinking behind the guide: questions over answers, one step at a time
  • Fixes: collapsing a branch hides its connection lines, and large gardens no longer fail to load their nested thoughts

On the trail of the last mobile crashes

Recent releases made the app far steadier on phones, but rare crashes remain. Now when a tab dies, Neural Garden notices and reports what happened.

  • If your phone's browser force-closes the tab, the app notices on your next visit and sends a crash report
  • Reports carry technical details only — device, memory, how many thoughts were on screen — never the content of your thoughts
  • Zoomed all the way out, dragging in the blank space beside a dot pans the canvas instead of grabbing the thought

A cleaner canvas and a panel you can resize

Drag the thought panel to fit your work, reach panel actions on your phone, and enjoy a tidier desktop canvas.

  • Drag the thought panel's left edge to resize it, and the width you pick is remembered
  • Panel actions are reachable on touch through a ⋮ menu, rather than icons that only appeared on hover
  • The bottom-left zoom, fit and lock buttons are gone — zoom on scroll or pinch, fit with F, and lock in Canvas settings
  • Colour regions appear only in Manual arrange, where they stay anchored to the thoughts you placed
  • The canvas orientation rings fade toward the edges again, restoring a subtle sense of depth

Wikilinks save instantly again

  • Saving a thought whose description contains a [[wikilink]] is fast again; it could stall for seconds, or leave the editor stuck open
  • The Smart detail canvas setting now says (variable LOD), so it is clearer what it changes

Cooler and steadier on your phone

The canvas background moved to one efficient layer that no longer crashes phones, and your local data moved off the main thread.

  • Rings, type zones and colour regions draw on a single light layer, ending the out-of-memory crashes when zooming on a phone
  • Your local data runs in the background, so the app stays responsive and your phone runs noticeably cooler
  • Type zones read as soft light glowing on the dark canvas, like terrain rather than a flat tint
  • Colour regions respond to touch: drag the label to move one, pull the corner to resize, tap to rename or recolour
  • Rings, type zones and regions are on by default on phones now that they are light enough to always show

Your canvas becomes a place

A full background for the infinite canvas: distance rings, automatic type territories, and colour regions you draw yourself.

  • Subtle rings around the centre replace the dot grid — their curve always points home, so you can tell how far out you are
  • Type zones tint the neighbourhood around each type's most important thought, sized by how much of your garden it holds
  • Draw your own colour regions: name them, recolour them, resize from the corner, and move a region with everything inside it
  • Regions are saved with your garden and sync across devices, and appear in Bookmarks as coloured entries
  • Bookmarks belong to the canvas you are on, and new ones appear in the list instantly

Lighter on its feet

  • Clustering runs in the background, so organizing related thoughts no longer freezes the canvas
  • The canvas does less work while idle and during drags, fixing a crash when pinch-zooming on phones

A smoother ride from card to dot

Zooming got a polish pass: cards unfold in stages, and the deepest zoom-out stays tidy, readable and clickable.

  • The full card opens condensed at 80% zoom and unfolds its properties, backlinks and nested thoughts as you zoom in
  • On desktop the detail levels morph into one another as you zoom, rather than switching in steps
  • At the deepest zoom, dot labels fade away to keep the overview clean; hover or select a dot to bring its name back
  • Thoughts with children keep their name and chevron at dot zoom, and connection lines point at the dot's centre
  • Every garden can zoom all the way out to the dot map, small gardens included

Aliases show up when you link

  • The wikilink picker lists each thought's aliases beneath its title, so you can search for a thought and link it by another name

Faster ways to start a wikilink

Optional shortcuts get you into the wikilink picker without typing brackets.

  • A full stop at the start of a sentence, or a capital letter at the start of a word, opens the picker — each toggled in Settings, both off by default
  • Multi-word titles work with these shortcuts, so a space no longer splits your search in two
  • Cancelling the popup sticks; it reopens only on a fresh edit
  • A tappable Cancel row makes the popup dismissible without a keyboard

Canvas zoom reads more like a map

Zooming out settles thoughts into clear levels of detail rather than continuously resizing them, closer to how a map labels cities and countries.

  • Thoughts settle into one of five detail levels as you zoom, with importance deciding when one gets more detail
  • A new mid-level view shows a thought's icon, title and counts before it collapses into a plain dot
  • Recently updated and well-connected thoughts bloom into full detail first as you zoom in
  • Your most important thoughts keep their name visible even at the deepest zoom-out
  • Town view pulls connected thoughts toward each other rather than scattering them

Small fixes and polish

  • Hover tooltips appear on every canvas toolbar and sidebar button; several showed nothing at all

Town view, rebuilt for mobile

The whole-garden map is smooth and crash-free on phones now, and zooming out stops where your thoughts actually are.

  • Town view no longer freezes or crashes on a phone, even for a large garden, and stays responsive while you pan and drag
  • Zooming out is crash-free everywhere: thoughts collapse to simple dots, keeping their soft glow so they still read on the map
  • Zooming out stops once your thoughts fill the screen, rather than drifting into empty space
  • Town view opens where you left it and lays out the same way each time, so large maps stop re-shuffling
  • Dimming the rest when you select a thought is now a lighter fade, and off by default on phones

Backup and Restore, made clearer

  • Restore is a simple choice: Import brings everything in as copies, Reset puts your thoughts back exactly where they were
  • Restoring keeps your custom types intact — they are matched by name, so nothing is dropped or duplicated
  • Light and dark mode moved into Settings, alongside a simpler Appearance section
  • Imported and adopted top-level gardens can be deleted

Back up and restore your whole account

A new Backup and Restore section in Settings captures everything you own in one file you can save, schedule, restore, or mirror to Markdown.

  • One click downloads every garden you own — nested ones included — with thoughts, connections, wikilinks, bookmarks and types
  • Restore two ways: bring everything in as fresh copies, or put your original thoughts back exactly where they were
  • Automatic snapshots every 6 hours, daily or weekly, kept in your browser to download or restore anytime
  • Mirror your garden to a folder of Markdown files that works with git, Dropbox, Drive or iCloud
  • Load demo data with one click to explore a fully populated example garden

See your whole garden at a glance

New ways to read a big garden: a map of every nested thought, a focus mode that dims everything else, and zoom that keeps your important thoughts legible longest.

  • Town view lays your whole hierarchy out as one living map, where children cluster around their parents and nothing overlaps
  • Dragging a thought rearranges the map around it without disturbing your real layouts
  • Focus dimming fades everything except the thought you select and its neighbours; press Escape to bring it back
  • Your most connected thoughts keep showing their card as you zoom out while everything else shrinks to dots
  • Zoomed-out thoughts show the right type colour and icon, matching the full card

Readable notifications on mobile

  • Notifications sit above the bottom dock instead of hiding behind it
  • They use the dark surface now, rather than a bright tint that clashed with everything else
  • They match the dock's rounded shape and keep clear of the home indicator

See at a glance when thoughts are locked

  • A locked thought shows a blue open-lock button, so it is obvious the placement is held and one click releases it

Breadcrumbs follow the path you took

  • When the same thought lives in more than one place, each copy shows its own trail, so going back up lands where you expect

Never lose your thoughts off-screen

Panning gently stops at the edge of your thoughts, so a stray swipe cannot fling the canvas into empty space.

  • You can push the outermost thought almost to the edge of the screen, but never so far that everything leaves the view
  • Zooming out near the boundary glides your thoughts back into view rather than snapping
  • The boundary adapts to each layout and zoom level, and timelines keep their own scrolling

Steadier on mobile

  • Opening or resizing the bottom sheet while dragging no longer locks the app into a loop that could crash the tab
  • Zooming to fit on a small screen stays within safe bounds

Lock thoughts in place

Pin a layout you like — a locked thought holds its spot and will not be nudged by a stray drag or an auto-arrange.

  • Select one or more thoughts and hit the lock button to freeze their position; selecting, editing and connecting all keep working
  • A locked thought renders slightly dimmer so you can see what is pinned, and the button flips to Unlock
  • Align, Tidy Up and the other arrange tools skip locked thoughts rather than shuffling them
  • Locked thoughts hold their spot when you zoom out too

Capture a backlog idea without leaving the canvas

  • A new Add to Backlog command drops a task into the Roadmap and opens it ready to type, the same way Add to today works

More ways to follow a wikilink

  • Shift-click a wikilink to open it in the side panel instead of moving the canvas, handy for a look while you keep writing
  • Right-click a wikilink, or long-press it, for Go to Node and Open in Panel

A calmer Daily timeline

  • Daily opens as a newest-first timeline, so today sits right at the top when you arrive
  • Drilling into a day and coming back fits the view to the timeline rather than overshooting
  • Panning reaches the edges of your entries instead of stopping at their centres

Fold everything under a heading in one click

  • Option-click a heading's chevron to fold every sub-heading, paragraph and list beneath it, and again to open it back up
  • The same works on the chevron in front of a paragraph or quote that sits above a list

Captures and AI results appear instantly

  • A captured page or a freshly classified thought shows up live on whatever garden you are viewing, with no reload
  • Each Agent Task links straight to where the work happened, so you can jump in and review
  • The Agent Tasks panel is one smooth scroll area, with a scrollbar that appears only while you scroll

Agent Tasks — your captures, typed by AI

A new Agent Tasks area is where background AI work shows up. Capture a batch of links and the agent sorts them into the right types, using your own key, on your command.

  • A red dot flags items needing attention; hit play and watch them get typed, with a history of past runs below
  • The AI reads each page and gives it a type, a clean title, properties and a one-line description
  • It types what a page is, not what it is about — a startup's site is an organization even when the startup builds AI
  • Turn on Auto-run to have new captures processed without clicking play each time
  • Captured links refresh on the canvas when you return to the tab or a run finishes

Capture any page into your Inbox or today

The Chrome extension works on every web page now. Pages arrive already typed — recipes, articles, films, podcasts, products — and land wherever you choose.

  • One click captures any page, with the type detected and the matching properties filled in
  • A recipe arrives with its full ingredient list and steps, so the capture is useful before you organize it
  • A Beatport album arrives as a rich Album, with artist and label as links, catalogue number and cover art
  • Choose where it lands: your Inbox, today's entry, or both — and the choice sticks
  • Capturing the same page twice is safe; the button says so before you click

Sharper types

All 45 built-in types got a once-over: shorter descriptions, tighter aliases, unique icons and a cleaner hierarchy.

  • A new Ethereal type for things that are not material, in a misty periwinkle that sits apart from AI, Org and Person
  • Every built-in type has a one-line description that is easier to scan in pickers and search
  • Aliases pruned of ambiguous overlaps, so type-matching feels more decisive when you make a thought
  • Org shifted to a bluer indigo so it stops competing with AI's purple
  • A type definition shows its own icon and colour, so each card previews itself

The same search everywhere

Every place that pops a search list — the title field, wikilinks, slash commands, the type picker, the command bar — now looks and behaves the same.

  • The same visuals, keyboard and selection behaviour across all five
  • A list flips above its anchor when there is no room below, and follows the cursor when the phone keyboard pushes the input up
  • Hover changes the selection only when you actually move, so the list stops grabbing whatever the cursor was resting on
  • Tapping a row on a phone picks it; it used to go grey and do nothing

Beatport tracks, and tidier properties

Drop a Beatport link on the canvas and Neural Garden builds a fully typed Track — artwork, player, artist and label linked up.

  • A new Track type with DJ-ready fields, and the artist and label nested inside the track so the whole release is one click away
  • A new Chrome extension adds the track you are looking at with one click
  • Drag, paste and the canvas menu all recognize a Beatport link and do the same thing
  • A Track shows its artwork with a player over it, and a way back to Beatport always one tap away
  • Noisy technical fields can be marked Hidden and collapse behind a toggle at the bottom of the thought

Small fixes

  • The + on a nested list creates a child when you type a name and press Enter, instead of quietly nesting an unrelated match
  • Help, changelog and landing pages scroll with one finger again
  • Drilling into a thought lands it where you can see it rather than behind the chrome

A pass over the phone experience

Text sized for phones, a dock that shows your real thoughts, and a swipe to indent while writing.

  • Titles, body text, panels and lists all step up in size, and the body sits large enough that iOS stops zooming your inputs
  • The dock shows your actual thoughts, each in its type's icon and colour; tap one to go there
  • Search matches across title, body, alias and type, so two words like event birthday find the right thought
  • Swipe right on a line to indent it, swipe left to outdent
  • Backlinks sort newest first, in the thought and in the panel

A smoother canvas on touch

  • Pinch to zoom works anywhere, including over a card, and keeps the spot under your fingers anchored
  • One finger pans over Timeline cards, so cards no longer create dead zones
  • Flicking the canvas coasts and slows naturally instead of stopping the instant your finger lifts
  • Long timelines feel less choppy
  • Page zoom no longer stacks on top of canvas zoom in Safari, so cards stop growing past their limit

Timeline, redesigned for journaling

A new vertical timeline stacks dated thoughts like a journal, and the horizontal one gained a date rail above the cards.

  • The vertical timeline stacks dated thoughts with a date rail, month dividers, and a + between dates to fill in a missing day
  • The horizontal timeline has a date rail that follows the cards as you pan
  • Entering a timeline fits to a comfortable reading width and centres on today
  • Scrolling stops cleanly at the edges
  • The vertical timeline runs newest-first and the horizontal one oldest-first, unless you sort them yourself

Calmer collaboration

  • Each thought remembers its own arrangement, so Outline at home and Grid inside a thought stay independent
  • Cursors are scoped to the canvas you are on, rather than shared across every part of a garden
  • A wikilink to something a collaborator cannot reach still shows its title, so the reference stays readable
  • An unreachable wikilink is faded and says so, rather than leading to an error

Read nested thoughts without opening them

  • Each nested thought shows a two-line preview of its text under the title
  • Hover any nested thought for the full tooltip — title and formatted body, with no need to drill in
  • Tooltips work across gardens, loading the text when it is not already on your device

Drop or paste any link onto the canvas

  • Drag a link onto the canvas and it lands with the page's title, description and preview image already filled in
  • Paste puts it in the middle of your view, or right-click to drop it at the cursor
  • If a link with that title already exists, you get a second copy of that thought rather than a duplicate

Wikilinks stay live

  • A wikilink reflects a rename immediately, including one made by a collaborator
  • A new wikilink no longer shows as crossed out before its target has settled; it crosses out only when the target is really gone

Descriptions that fold like an outliner

Bulleted lists inside a description fold and unfold, and a wikilink shows its target's type at a glance.

  • Any list item with children gets a chevron to fold it, and the state is remembered across reloads
  • Option-click folds the whole branch, and Cmd+. folds the branch under your cursor while typing
  • Numbered lists fold too, with the chevron beside the number rather than replacing it
  • A wikilink wears its target's type icon in that type's colour, including links to another garden
  • A new /Body Text command resets a line to a plain paragraph from inside any heading, quote or list

Daily journaling, and a smarter command bar

A Daily garden with one entry per date, commands to link any date, and a command bar that picks the best match rather than always offering Ask AI.

  • /today, /yesterday, /tomorrow and /date drop links to dates, and Add to today jumps straight into today's entry ready to write
  • Ask AI shows up only when nothing else matches, so typing a title selects that thought
  • The pencil puts your cursor at the end of the body so you can keep adding lines; a brand-new thought still starts in the title
  • Pressing F re-runs the layout as it fits the view
  • Auto-arrange waits for you: a new thought stays under your cursor while you name it, then slides into place

Aliases, and a calmer canvas top

Give a thought other names it answers to, and get back the vertical space the canvas chrome was taking.

  • Add nicknames or alternative spellings; typing one into a [[ search finds the thought and links it under that name
  • Removing an alias rewrites every wikilink that used it back to the real title, straight away
  • A book's author, a film's director and a track's artist are proper thoughts now — type [[ to pick or create one, already typed
  • New thoughts start with no type, and adding a child no longer copies the parent's type onto it
  • The canvas top is one row instead of two, giving about 80px of height back

Live renames and richer properties

A rename flows through every wikilink and backlink as it happens, and a property value can hold a link.

  • Rename a thought and every link to it updates immediately, everywhere it appears
  • Type [[ in a property value to link a thought; if the property is tied to a type, the new thought arrives already typed
  • Renaming a property carries every existing value over to the new name, rather than leaving them orphaned
  • The side panel mirrors the card's sections, so what is open stays open as you move between them

Smoother nesting, wikilinks that flow

A new thought made inside another one now lands there directly, and brackets pair themselves as you type.

  • The + on a nested list creates the thought inside its parent straight away
  • Typing [ pairs the bracket, a second one opens the link picker, and backspace removes both at once
  • Press [ around selected text to wrap it, and again to turn it into a link with that text already searched
  • Choosing Create new from a link puts the new thought inside the one you are writing, rather than dropping it far away
  • A backlink shows the sentence around it, with the link highlighted

Sturdier delete and restore

  • Deleting a thought no longer takes its grandchildren with it — the level below moves up intact
  • A restored thought reappears on the canvas immediately and the view centres on it
  • The command bar, wikilinks and backlinks all land on the right thought, already centred
  • The command bar matches aliases, so searching a nickname finds the thought and says why
  • Breadcrumbs show where a thought sits, so the same thought always shows the same trail

Viewing and editing feel like one surface

Switching between reading and editing a thought no longer shifts anything under your cursor.

  • Title, text and type stay in the same place with the same font, on the card and in the side panel
  • Your cursor lands at the end of the title rather than selecting all of it
  • Editing survives a zoom, so you can change detail levels without losing what you were writing
  • Picking an existing thought from the + autocomplete now draws the connection, which it used to skip
  • Escape saves and returns to reading

Steadier canvas loads

  • A single failed background save can no longer leave the canvas stuck loading; it paints with what it has
  • Text arriving from a collaborator or an import is filtered down to the formatting Neural Garden supports

The canvas opens like a notebook

Returning to a garden you visited recently paints instantly from your device, while fresh data arrives behind it.

  • A recently visited garden paints in under 50 milliseconds, with no wait on the network
  • Backlink counts and badges arrive a beat after the canvas, rather than holding up the first paint
  • On slow wifi every load used to be a cold load; that is fixed, so repeat visits really are fast
  • The in-browser database downloads once ever, instead of on every page load

Tighter layouts

  • Grid stays tight however wide your thoughts get, with a simple gap control
  • New, tighter defaults across the arrange modes
  • Switching into Gravity no longer crashes the canvas

Links that land where you expect

  • Clicking a backlink from inside a thought takes you to the copy in that thought, not back out to the root
  • The here badge in the placement picker marks the garden you are actually looking at
  • Renaming a thought updates every link pointing at it, immediately, everywhere on your canvas

Outline mode, and richer previews

An outline that behaves like one, and thought text that shows its real formatting even at a glance.

  • Outline indents by exactly the gap you set, down to nothing, and its lines snap to clean right angles
  • Bullet lists, line breaks and wikilinks look the same in a preview as they do in the editor
  • Nested rows declutter: the drag handle moves right and the pop-out arrow is gone
  • Softer thought surfaces, with connection handles that appear only on hover or selection

Quieter, calmer canvas

  • Expanding shifts the layout first, then fades new thoughts into the space that opened, instead of popping
  • Smaller, subtler connection handles that stay out of the way until you need them
  • Outline draws clean vertical lines, so a branch reads as a single column
  • A new Zoom to 100% option in the canvas menu

Types you can edit

The built-in types can be shaped on the canvas now. Changes stay a draft until published, and your own edits survive an update.

  • Edit colours, icons and properties directly, with no code deploy needed
  • A banner counts your unpublished changes, with Publish and Revert
  • When types are published, your own edits are preserved and only untouched fields take the new values
  • The thought action pill no longer appears on hover, only when a single thought is selected, so brushing past stops flashing menus

Groups and connections stay in sync

  • Drag a collapsed thought onto another and everything hidden inside comes along
  • Connections between thoughts survive being nested together — drill in and they are still there
  • Drawing a connection in any auto-arrange mode re-runs the layout with the new shape
  • Wider, more readable thoughts with tighter padding

Reliable AI actions everywhere

  • A thought the AI creates inside another one stays there after a reload
  • AI edits land on the right thought, fixing the phantom not-found errors that caused long waits
  • AI actions get a realistic 30 seconds, so a slow network no longer times out mid-way
  • The sparkle opens suggestions on click only; hovering near an edge no longer triggers it

Every thought is its own garden

Drill into any thought to see what is nested inside it on its own canvas, and link thoughts together with wikilinks that stick.

  • Open any thought as a garden — what is nested inside gets its own canvas, with full editing, connections and AI
  • Wikilinks are tracked properly, so backlinks update live for collaborators and survive a reload
  • Title autocomplete appears everywhere you make a thought, so you link to what exists rather than making duplicates
  • Rename a thought to a title that already exists and you are offered a merge, combining both into one
  • Deleted thoughts go to the trash first, and can be restored to where they came from

Type badges show on every thought

  • Type chips and coloured borders appear on full-size thoughts, not only the small ones
  • Existing notes, ideas, tasks and questions are matched to their types automatically
  • Type assignments survive an update, so a refresh never quietly orphans your typed thoughts

The command bar becomes an assistant

Create, connect and nest thoughts by typing what you want, and the canvas rearranges itself around the result.

  • Ask the AI from the command bar to make thoughts, connect them, or put them inside one another
  • The canvas rearranges when thoughts are added, removed or resized, in every layout mode
  • Settings, shortcuts and every toolbar action are searchable and runnable from the command bar
  • Settings became a quick modal rather than a separate page

The view keeps its place

  • Opening or closing a panel adjusts the view to keep what you were looking at centred, with no jarring jump
  • Tapping a thought on a phone makes room for the sheet immediately rather than waiting for the animation
  • The hover pill on small thoughts gained Select children and Properties, matching the full-size menu

Gravity mode, and settings per garden

  • Focusing a thought in Gravity shows its whole family — parents, siblings, children — fading in by how close they are
  • A focused thought shows its title and text together
  • Arrange mode, sorting, grouping and filters are saved per garden and restored when you return

Nested thoughts, and a cleaner panel

  • A nested image shows as a full-width picture, and a nested link shows its favicon and domain
  • A + at the bottom of every nested list makes a child without leaving the thought
  • The sparkle suggests new nested children, and offers to set up an AI provider if you have none

Types survive an import

  • An export carries its full type definitions, so imported thoughts keep their types when shared
  • On import, types are matched against your own and link up automatically
  • Anything unmatched appears in the preview with Create and Skip

A properties sheet that feels native

  • The bottom sheet snaps to a quarter, half or full height, and a fast flick skips between them
  • Pull down when the content is already at the top to shrink the sheet rather than stretch the content
  • Flick down or tap outside to dismiss, with real spring physics
  • Only the chevron opens a section on a phone, so scrolling no longer opens things by accident

Canvas layout polish

  • Switching layout modes animates thoughts to their new places instead of snapping
  • Every layout spaces thoughts by their real size, so different sizes no longer overlap
  • Dragging a thought onto others pushes them apart smoothly
  • Connection lines are easier to click, and their arrows hug the thought at every zoom

Types system polish

  • A type's colour picker opens from the colour chip and works properly
  • Colour, icon and properties sit under one section rather than being shown twice
  • Pick a type's icon from a searchable list of over 200, grouped by theme, instead of typing its name

A consistent look across every menu

  • Menus, panels and dropdowns share one style, so they look and behave the same everywhere
  • Sidebar panels open anchored to their icon, making opening and closing more predictable
  • Menus animate with a subtle scale so the frosted backdrop stays visible throughout

AI intelligence per garden

  • Each garden has its own AI summary and tags, so drilling in reflects that context rather than the parent's
  • It writes itself the first time you visit a garden, with no setup
  • It stays fresh: half a minute after you change something it updates using only what changed
  • The AI orb pulses while it is thinking

Each garden remembers its own canvas

  • Every garden remembers its arrange mode, so switching no longer resets your layout
  • Sort, grouping and hand-placed positions are kept per garden

Gravity and Outline arrangements

  • In Gravity, thoughts orbit a focused one, with children above and grandchildren clustered around their parents
  • Tap any thought to refocus, and the hierarchy reorganizes around it
  • Children sort left to right by whatever you sort on, with grandchildren grouped under their parent
  • Drag a thought sideways to override its position, and it stays until you change the sort
  • A new Outline mode arranges the whole canvas as an indented list

Reliable types everywhere

  • A duplicated thought keeps its type
  • A new child stays in the canvas you made it in, rather than jumping to Home
  • Undo and redo on a child keep its type
  • Collaborators see the owner's types, and they survive a refresh

Import and export your garden

  • Export a garden as JSON, Markdown, OPML or CSV
  • Import from the same formats, so files from Obsidian, Miro and XMind come across
  • Choose how an imported hierarchy lands: nested inside each other, or flattened into connected thoughts
  • An import lands on the canvas you are looking at, rather than always at home

True offline mode

  • Create, edit, delete, nest and rearrange without a connection; everything syncs when you reconnect
  • Every arrangement runs on your device, with positions saved locally
  • An indicator shows what is still waiting to sync
  • Dark mode no longer flashes white while the page loads

Steadier collaboration

  • Type definitions no longer vanish while collaborating
  • A type you create, edit or delete reaches collaborators straight away
  • Someone you share a garden with stays inside it, rather than being able to navigate up into yours

Admin tools and user management

  • An admin panel in the sidebar for managing people and shared gardens
  • Promote or demote admins, and remove accounts for a clean reset
  • The first person to sign in becomes the administrator

Types keep themselves in order

  • Improvements to the built-in types reach your garden automatically on your next load
  • Four new built-in categories: Abstract, Emotion, Place and Thing
  • Fixed types multiplying on every page load and flooding your home canvas; existing duplicates are cleaned up

Adaptive zoom, and a sheet you can resize

  • Thoughts change shape as you zoom — full cards, then compact strips, then labelled icons, then dots
  • The properties panel no longer blocks zooming; scroll and pinch freely with it open
  • Drag the mobile sheet's handle to resize it, or pull it right down to close
  • Tapping a thought when zoomed out opens its properties, so you can look without zooming in

A type system that grows with you

  • Assign a type on the thought itself — search, pick or create one without opening a panel
  • Types nest, so Media opens into Book, Film and Song as a submenu
  • Properties flow down: define author on Media and every Book, Film and Song has it
  • Thoughts show coloured type chips, and the minimap dots match

Editing connections, and tidier layouts

  • Right-click a connection to flip its direction, edit its label or delete it, with full undo
  • Drag-select a region and the connections between the selected thoughts highlight too
  • Select Children now takes the whole branch, not just the level below
  • A new Tidy Up arranges thoughts as an indented outline
  • Keyboard shortcuts work even when focus is in the sidebar

Editing reliability

  • Rapid Enter and Escape while editing no longer risks losing your changes
  • The context menu shows keyboard shortcuts, so faster ways are discoverable
  • A failed save or delete now tells you, instead of failing silently
  • The + on a thought ignores rapid double-clicks, so you stop making duplicates

Apple Pencil on iPad

  • Draw with the pen to select thoughts, tap to deselect, double-tap to make one
  • Switch between finger and pencil freely — finger pans, pencil selects, with no ghost boxes in between
  • Double-tap a thought with the pencil to start editing

Nesting that behaves

  • The hover buttons no longer steal clicks from the thought itself
  • Dropping into a tall thought works across its full height
  • A dropped thought lands at the bottom of the list, or exactly where the insert line shows
  • Drag a thought over a nested list to place it precisely, with an indicator showing where

Steadier under the hood

  • Heavy layout work moved off the main thread, so the canvas stays responsive
  • One crash no longer takes down the whole app
  • Keyboard navigation and screen-reader labels across the toolbar, menus and thought actions
  • Fixed duplicate thoughts on rapid clicks, missing children after un-nesting, and orphans after undo

Touch and Apple Pencil

  • Full touch support: tap to select, long-press for a menu, drag to select several
  • Apple Pencil works naturally, tolerant of the small movements of a resting hand
  • Drag-select a group and move them together
  • Notched devices are respected, so nothing hides under a screen cutout

Canvas interaction polish

  • AI suggestions slide out smoothly and dismiss naturally
  • Bigger targets for drawing a connection, and a cleaner canvas while you drag one
  • Hover a connection to see where it leads, and click the arrow to travel there

Nesting stability, and one way to delete

  • A nested thought no longer reappears as a ghost on the root canvas
  • Deleting from the gardens panel and from the canvas now do exactly the same thing
  • Children of a deleted thought settle neatly below where their parent was
  • A confirmation appears before deleting a thought with more than five children
  • Drag several selected thoughts onto another to nest them all at once

Frosted glass and a dark theme

  • Frosted glass across the sidebar, panels, toolbar and menus
  • Dark mode is the default now, with light mode remembered if you choose it
  • Softer borders and muted text for a calmer look
  • Panels line up with the sidebar icon that opened them

Smoother canvas interactions

  • Zooming from the toolbar and menu animates smoothly, without the stutter
  • Long-press opens the context menu on a phone
  • Click a thought for a blue selection, and Shift-click to select several
  • A floating pill with align, distribute, tidy, duplicate and delete
  • A new Saved Views panel for storing and restoring layouts

Gardens panel redesign

  • Home is pinned at the top, with sorting and the create button below it
  • The sort label cycles through Edited, Created and Name, and the arrow flips the direction
  • Smooth zoom when drilling into a thought and back out
  • Your position and zoom live in the URL, so a view can be shared

Drilling into thoughts

  • Full drill-down navigation between canvases
  • Your home garden cannot be deleted, and looks distinct
  • A proper favicon and social preview image

Grouping and nesting

  • Drag thoughts onto each other to group them
  • Drill into a group to see what is inside
  • Drag back out to un-nest, with clearer drop indicators
  • Landing page polish: brand colours, spotlight cards and a hero glow

Landing page launch

  • The landing page, with the Hex Bloom logo and a new typography system
  • An interactive demo canvas in the hero
  • A warm neutral dark theme, and support for notched devices

AI suggestions and garden intelligence

  • Hover a connection handle to see what the AI thinks connects
  • Garden Intelligence gives each garden its own summary and context
  • An animated AI orb replaces the old button
  • Local AI: available Ollama models are detected for you
  • Undo now covers edits to a thought's text

Canvas overhaul and collaboration

  • A large pass over connections, settings and general canvas polish
  • Your arrange mode syncs to the people you are working with
  • Editing controls are hidden in view-only and locked modes
  • Your first garden is named after you

Real-time collaboration

  • Work together in the same garden, with live cursors
  • Share a garden by invitation, with roles deciding what each person can do
  • A dot flags a garden newly shared with you
  • A user menu in the sidebar for signing in and out

Initial release

The first version of Neural Garden goes live.

  • An AI-powered second brain on an infinite canvas
  • Thoughts you can connect, arrange and drill into
  • Local-first: your garden lives in your browser