← Back to Sprigpost

THE BUILD JOURNAL

Building a home
for agent work.

Updated with every release.

Real changes, open questions, and the decisions behind Sprigpost. Catch-up notes show when they were added to this journal.

Confirm newsletter subscriptions from ordinary browser forms.

Public signup and confirmation forms now preserve the origin required by their existing request checks.

What changed

  • Use an origin-only referrer policy on public newsletter pages so normal form submissions can pass the same-origin check.
  • Keep email confirmation and unsubscribe tokens out of referrer URLs while retaining consent, request-origin and publication-scoped checks.
WHAT’S NEXT

A newsletter still needs a configured brand sender and confirmed subscriber consent. This fix does not enable campaigns or schedules.

Keep active projects close in your master mailbox.

Pin the Mail projects you are working on and focus the master mailbox on their mail.

What changed

  • Keep personally pinned projects above the Mail project picker, with quick inbox access and an unpin control.
  • Switch between all projects and pinned projects while using the existing folders, search and older-message pages.
  • Save pins with your account so they remain available across visits and devices.
WHAT’S NEXT

Pins apply to whole Mail projects and all their connected inboxes. They organize your view without changing messages, delivery or agent access.

Separate Websites from internal Publishing.

A focused website pilot brings brand mail, newsletters and occasional news into a simpler workspace while Publishing remains restricted to the internal owner.

What changed

  • Give Websites its own workspace-scoped navigation and private registration, separate from internal Publishing.
  • Move the existing Dinoroost connection into the website pilot without changing its sender, subscriber records or delivery settings.
  • Reuse reviewed newsletter and news tools for the connected pilot; new website records do not activate public integrations or workers.
WHAT’S NEXT

Other brands remain to be classified. Customer domain verification and newsletter connection still need onboarding before new customer websites can use those integrations.

Review and schedule newsletter campaigns.

Campaign approval and draft scheduling are now available in the private portfolio newsletter desk. Each delivery still requires review of its edition and subscriber audience.

What changed

  • Enable campaign approval for an exact saved edition, confirmed subscriber audience and requested send time.
  • Allow each publication to configure weekly or monthly preparation of private draft copies for later review.
  • Keep sender, consent, unsubscribe and shared mail-capacity checks in place when an approved campaign is delivered.
  • Leave existing templates intact; enabling the controls does not create campaigns, configure schedules or approve messages.
WHAT’S NEXT

Welcome backlog recovery remains on hold. Each publication still needs its sender, subscription forms and actual content reviewed before live delivery.

Prepare newsletter editions and review their audience.

The private newsletter desk adds independent campaign drafts, subscriber review and draft schedules, with delivery controlled separately.

What changed

  • Prepare welcome editions for registered private sites and select an exact project sender, with sending and receiving status shown separately.
  • Save and preview campaign editions, then review their content, confirmed audience and requested send time before approval when sending is enabled.
  • Prepare weekly or monthly draft copies for editing; draft schedules do not approve messages or create new reporting.
  • Preserve confirmation consent when welcome delivery is delayed, recover missing welcomes when enabled, and keep unsubscribe and suppression checks active.
WHAT’S NEXT

Campaign sending and welcome recovery remain disabled until configured. Each site still needs its domain, sender, forms and actual edition reviewed before live delivery.

Keep social account check dates accurate.

The private social account editor now records a check made today at its actual time and prevents future dates.

What changed

  • Use the local calendar date consistently when reviewing a saved check.
  • Record today’s check at the current time and earlier dates at local noon for stable date display.
  • Limit new checked dates to today or earlier and explain rejected future dates.
WHAT’S NEXT

Instagram signup, verification and publishing remain separate from this private setup record.

Track social accounts and their remaining setup.

All sites gains a private social account overview, starting with Instagram and each project’s email identity.

What changed

  • See account handles, recorded profile links, owner access, email verification and remaining profile setup.
  • Keep dated evidence visible and distinguish proposed handles and caption drafts from existing accounts.
  • Prepare private caption drafts, filter setup gaps, and open the corresponding project mailbox.
  • Import and export setup records with protection against conflicting updates; platform signup and publishing remain separate.
WHAT’S NEXT

Continue official Instagram registration and verification for each project, completing required owner steps on the platform.

Read older mail without losing your place.

The mailbox list and selected message now scroll separately, keeping the reading pane visible while you browse older emails.

What changed

  • Keep the mailbox heading, search and message count in view while scrolling through the message list.
  • Open an older message beside the list without scrolling the whole page back to the top.
  • Keep your list position when selecting messages or loading another page, and start each newly selected message at its beginning.
  • Keep message actions available while reading long emails, with a separate reading view and return to the list on phones.
WHAT’S NEXT

External inbox history still requires a separate connection or import into Sprigpost.

Monitor every project from one master mailbox.

The master mailbox brings received and outgoing mail from all of your workspace’s Mail projects into one searchable view.

What changed

  • Open Master mailbox in the sidebar to see mail across projects, with a project icon and name on every message.
  • Keep received, archived, sent and queued messages together while retaining their individual delivery status.
  • Search across projects and load older messages without switching through each inbox.
  • Keep replies and forwards in the original message’s project and preserve private workspace access checks.
WHAT’S NEXT

The master mailbox displays mail already stored in Sprigpost; external inbox history still requires a separate connection or import.

See which sites need a research or content update.

The owner’s All sites directory gains a private freshness view with separate evidence for research reviews, imports, builds, deployments and website checks.

What changed

  • Filter the portfolio by missing website links, unknown or overdue research checks and unresolved content integration.
  • Show the scope and source behind each research-review date; older package dates and successful website checks do not make research current.
  • Import a saved portfolio ledger into private owner storage, with validation and protection against conflicting updates.
  • Keep unknown dates and missing evidence visible, including research that has been reported but not recovered or integrated.
WHAT’S NEXT

Research recovery and site updates continue separately; the ledger reports verified progress without automatically publishing content.

Keep project views tied to the project you selected.

Fast project switches now keep an older response from replacing the newly selected project’s view.

What changed

  • Show a clear loading or error view while a selected project opens, instead of leaving controls from the prior project on screen.
  • Keep mail, Board, Drive and agent lists tied to the project that requested them when responses arrive out of order.
  • Protect an open draft or setup form from saving to a different project after a switch, while keeping its text available to review.
  • Keep message edits from being lost while an attachment is added or removed.
WHAT’S NEXT

The paid workspace pilot and Stripe activation remain in preparation.

Receive attachments from the mail provider’s current download host.

Real delivery checks identified an attachment download host change and domain-record formats that needed support.

What changed

  • Accept the mail provider’s exact current attachment host while preserving HTTPS, download size and redirect checks.
  • Recognize an empty receiving MX record name as the domain root.
  • Normalize subdomain DNS record names without repeating the subdomain label.
WHAT’S NEXT

Held messages require an owner retry. Paid workspace and Stripe launch validation continues separately.

Keep private previews tied to their saved versions.

Later website imports can now extend a private review without breaking the earlier batch receipts.

What changed

  • Verify exact saved document revisions and source bytes against their original import receipts.
  • Distinguish historical batches from current associated revisions and keep unlinked batches visible for explicit association.
  • Combine earlier linked batches of the same package into a separately verified private preview, while preserving original delivery evidence.
WHAT’S NEXT

Human review remains pending until the owner confirms it. Public release, new research requests and domain changes remain separate.

Edit a listing description without taking it down.

Owners can now revise the short description on a shared portfolio card while keeping its listing active.

What changed

  • Add a Public description editor to each public All sites card.
  • Save copy changes with the existing revision history and workspace checks.
  • Keep the existing listing name, link, image, labels and visibility when only the description changes.
WHAT’S NEXT

This edits the Sprigpost portfolio card only. It does not publish site articles or change email, domain or site settings.

Keep large private website libraries usable.

Imported website documents now load in small pages, with library totals and private-draft counts kept visible.

What changed

  • Load document summaries and pinned image previews 16 at a time in the owner workspace.
  • Search and filter imported documents across the whole private library, then page through the matching results.
  • Show a temporary availability error instead of an incorrect zero count if the document library cannot load.
WHAT’S NEXT

This changes the owner review workspace only. Imported drafts remain private; domain connection and public publication require separate steps.

A clearer way to explore the Sprigpost portfolio.

Knowledge in Motion now has its own destination, while public project listings can be searched, filtered by category and curated into a Showcase.

What changed

  • Move the interactive Knowledge in Motion experience out of the portfolio page and add a dedicated navigation link to its standalone view.
  • Add public-listing search, category filters, recently-listed and latest-published-edition sorting, plus an owner-curated Showcase section.
  • Add private owner controls for public category labels and Showcase selection; internal tags and private listings remain private.
  • Distinguish owner-shared temporary review links from permanent websites without changing their publishing or mail setup.
WHAT’S NEXT

The public portfolio continues to show only listings explicitly shared by their owners. A directory listing is separate from article publication, domain connection and service readiness.

Clear the selected marker when resetting the map.

Reset now clears the prior marker highlight along with zoom and field-note selection.

What changed

  • Restore the full map view and remove the selected marker styling when Reset is used.
  • Keep the current search and country/area filter applied so reset only affects map selection and zoom.
WHAT’S NEXT

The authenticated map remains a 17-region Northern Europe preview. The separate 248-country/area directory remains geographic navigation only.

Keep Birdwatch map results aligned with the selected filters.

Search and area filters now update the map markers as well as the region list, and the full 248-country/area directory is linked beside the 17-point Northern Europe map.

What changed

  • Hide markers that do not match the current search or country/area selection.
  • Keep map location notes approximate and preserve the supplied-data boundary between regional points and the country/area directory.
  • Link directly to the authenticated imported directory and map source notes.
WHAT’S NEXT

The directory is geographic navigation, not 248 researched bird guides or occurrence records. The site remains private and unpublished.

Keep map markers visible over the full regional tile area.

The follow-up map check caught that tile images were drawn over the markers and the tile rows did not match the latitude range; the layer order and bounds are now corrected.

What changed

  • Draw map tiles underneath all 17 region markers.
  • Use the zoom-4 tile rows that contain the supplied 51–70°N coordinates and check every point stays inside the map frame.
  • Keep exactly one pre-rendered list option for each region and one for each country or area.
WHAT’S NEXT

The 248-country/area discovery directory remains separate from bird occurrence data. The Birdwatch website remains private and unpublished.

Show all 17 regions on the private Birdwatch map.

A live-preview check caught a duplicated region list and a tile window placed too far south; both are corrected in the authenticated map.

What changed

  • Keep one accessible entry per mapped region and one option per country/area.
  • Align the displayed OpenStreetMap tiles with the full approximate Northern Europe marker range.
  • Keep the map restricted to the exact verified Birdwatch batch and signed-in Main Desk session.
WHAT’S NEXT

This remains a 17-point regional map and a separate 248-country/area directory, not 248 researched birding locations. Birdwatch remains private and unpublished.

Explore the Birdwatch map inside the private preview.

The exact owner-associated Birdwatch v2 preview now includes a usable map for its 17 approximate Northern Europe regions.

What changed

  • Add marker selection, region search, country filtering and map zoom to the signed-in Main Desk review flow.
  • Serve the map through the same founder and workspace checks as the sanitized pages, and render its controls with Sprigpost-owned code instead of running scripts from the uploaded package.
  • Keep the imported notes and sources available beside the map; OpenStreetMap tiles load only for the displayed map and receive the visitor’s network IP.
WHAT’S NEXT

The 248-place atlas remains a geographic directory, and detailed bird guides remain focused on Northern Europe. The website is still private and unpublished.

Keep the exact-batch confirmation through submission.

Main Desk now retains the owner’s checked batch confirmation while the association request is being sent.

What changed

  • Capture the confirmation before the busy-state redraw clears the review form.
  • Keep the existing exact request, site, import ID and manifest-hash checks on the server.
  • Leave the original delivery history and public-release setting unchanged.
WHAT’S NEXT

The owner still reviews the sanitized current preview separately; the full interactive map build remains outside that preview.

Show saved website batches when reviewing a request.

Opening a delivered Main Desk request now loads its separate Website Imports batches so the owner can link and review the exact current package.

What changed

  • List unlinked owner-cleared batches beside a superseded original preview, with the batch ID and manifest hash visible before confirmation.
  • Refresh the batch status after an association or preview action so the current private review state is visible immediately.
  • Keep association explicit and immutable; the original delivery proof and public release setting remain unchanged.
WHAT’S NEXT

A current preview still needs its own server verification and visual owner review; full interactive builds remain outside this sanitized preview.

Review newer site imports without rewriting delivery history.

Main Desk now requires an explicit, auditable owner link between a Website Imports batch and its project request before creating a current preview.

What changed

  • Keep the original delivery archive, import receipt and proof unchanged when Website Imports later updates the site documents.
  • Record the exact request, site, batch ID and manifest hash when linking a separately imported package, even when its package key differs.
  • Verify and review only the newest owner-associated batch independently, including private draft documents and stored images.
  • Mark superseded original previews clearly and keep script-enabled interactive builds outside the authenticated app origin.
  • Candidate 050 pairs this with clearer worker availability and a recoverable refresh when worker status requests stall.
WHAT’S NEXT

Unlinked batches remain visible for owner review. Current previews are sanitized; full interactive review needs a separate preview origin protected by Cloudflare Access.

Distinguish paired workers from saved activity.

Main Desk makes worker selection clearer and lets a stalled availability check recover without losing the review.

What changed

  • Show server-paired choices separately from the worker inventory, with a specific explanation inside unavailable pickers.
  • Time out stalled availability reads and offer a safe retry while retaining drafts, selection and confirmation.
  • Exclude revoked connector pairings from the server-provided choices, without expanding worker access.
WHAT’S NEXT

Queue pairing does not establish current worker readiness. Fleet admission and research submission still require their own checks.

Keep original deliveries linked to smaller review packages.

Main Desk can bind an owner-reviewed content derivative to its exact original delivery, with separate archive evidence and unchanged pilot import limits.

What changed

  • Review original and derivative filenames, sizes, hashes, transformations and omissions before confirming their immutable relationship.
  • Require the uploaded derivative and its manifest to match that relationship before allocating or resuming a private import.
  • Reuse exact retries and let the owner revoke lineage while retaining original receipts and import history.
WHAT’S NEXT

This remains a bounded content-review workflow. Full interactive builds, large-original transport, editorial acceptance and public release require separate checks.

See why a worker is unavailable and refresh your review.

Main Desk separates queue availability, recorded research progress and website verification, with clearer request review on phones.

What changed

  • Explain activation off, no paired workers, loading and refresh failures beside the worker picker.
  • Refresh worker availability while retaining your entered concept, selected worker and approval checkbox; require a valid selection and confirmation before enabling approval.
  • Keep a focused phone picker in place during timed status redraws and show recorded research submissions separately from full website readiness.
WHAT’S NEXT

A research receipt does not verify a completed interactive website. Original source, package checks and owner preview review remain separate steps.

Open reviewed Main Desk requests to the paired executor.

The private owner queue can now accept explicitly approved requests for the paired, scoped executor.

What changed

  • Enable owner-confirmed queue approval and guarded claims without granting snapshot credentials execution access.
  • Retain worker admission checks, expiring claims, durable receipts and the single-submission safeguard.
  • Keep research delivery, private content review and public website release as separate evidence-backed steps.
WHAT’S NEXT

A saved or queued concept does not prove that research has started. The operator must check the assigned worker and follow the exact request; no new scheduler, model runtime or automatic website publication is introduced.

Follow a concept through research handoffs to a private website.

Main Desk adds saved concepts, reviewed worker queues and separate evidence for package delivery, import and private preview review.

What changed

  • Browse saved worker chats with source dates and confirmed project mappings, while keeping unknown activity visible.
  • Show the next action, responsible operator and current blocker above each request timeline.
  • Use scoped execution credentials, durable receipts and guarded recovery to prevent automatic research resubmission.
  • Match an uploaded ZIP to its delivery receipt, import complete documents and images into an owned site, and check private pages before recording owner review.
WHAT’S NEXT

Execution requires separate operator pairing and activation. The bounded private import pilot does not execute supplied build scripts, publish websites, or establish that any real research request has completed.

Bring a private Main Desk into the workspace.

The owner workspace now has a read-only overview for approved project, worker, usage and handoff snapshots from a local connector.

What changed

  • Browse and filter project summaries, saved worker ownership, usage observations, research requests and separate handoff stages on tablet-friendly pages.
  • Keep connector contact separate from source freshness, with explicit unavailable, stale, offline and unknown states.
  • Accept bounded snapshots through a dedicated revocable credential that cannot read mail or dispatch work.
WHAT’S NEXT

The local connector must be paired before observations appear. Remote execution remains disconnected; snapshot receipt does not establish browser readiness or publication.

Keep the refined inbox and site analytics working together.

This release combines the compact inbox with the remaining site analytics connections and corrects analytics readiness for websites whose public forms are disabled.

What changed

  • Preserve the searchable, expandable mail connection panel and its light, dark and narrow-screen layouts.
  • Recognize a valid site-specific analytics credential independently of forms, newsletters and mail setup.
  • Retain analytics-only access limits, approved public paths and the existing Agent Flow views in the combined release.
WHAT’S NEXT

Analytics readiness reports configuration; recorded events are still required to establish observed activity. Newsletter and mail readiness remain separate.

Give the inbox room to breathe.

Mail connection details now sit behind a compact readiness summary, keeping messages in view even for projects with many email addresses.

What changed

  • See connected-address totals and sending and receiving readiness at a glance, with a clear count of addresses needing attention.
  • Expand a searchable connection list, filter pending addresses, and open address management without leaving the selected project.
  • Keep large address lists in a bounded panel and adapt the inbox layout to narrow screens, in both light and dark mode.
WHAT’S NEXT

Readiness describes mail configuration, not guaranteed delivery. These presentation changes do not alter DNS, sending, receiving or mailbox contents.

Connect analytics across the remaining portfolio sites.

Older Pages sites and a separately hosted Worker can now report cookieless page views without replacing their live websites or using mail credentials.

What changed

  • Accept approved public .html pages in site analytics while continuing to reject private routes and query strings.
  • Allow a separate, site-scoped credential for analytics intake only; it cannot open mail, forms or other site APIs.
  • Check the four older websites at their verified www hosts while retaining each apex domain as its brand and sender identity.
  • Enable Play Past the Hype after checking its live helper, privacy text and 126 public pages; prepare the final five existing sites for the same first-party collection.
WHAT’S NEXT

Each site displays configured versus observed activity separately. A new page must be added to its approved path list before its views can be recorded.

Keep imported-site analytics and Agent Flow available together.

Two concurrent releases are now combined, preserving the visualization and the latest analytics support for imported websites and larger libraries.

What changed

  • Restore site-scoped page-view reporting on the imported website adapter without exposing its server credential to browsers.
  • Retain the 3,000-path analytics manifest limit and bounded 256 KiB owner configuration requests for larger site libraries.
  • Preserve the analytics rollout to twelve additional connected websites, whose privacy copy, live edges and sitemaps were checked in the preceding analytics release.
  • Keep both public and private Agent Flow views and the corrected portfolio embed controls.
WHAT’S NEXT

Six sites still need website-specific connection work. New article routes must be added to the approved path list; configured analytics is not a claim that visitors have arrived.

Keep the embedded Agent Flow controls working after deployment.

A production packaging correction restores automatic sizing and fullscreen controls in the public portfolio embed.

What changed

  • Preserve the embed’s browser script through the production build so it runs independently of server-side bundler helpers.
  • Verify sizing and fullscreen behavior against the bundled script as well as the development source.
WHAT’S NEXT

The public visualization remains a fictional demonstration; signed-in Agent Flow reads saved project activity.

Explore Agent Flow in the portfolio and your private workspace.

An interactive public demonstration and a private project view make agent collaboration easier to follow.

What changed

  • Explore fictional agent teams and data streams in the public portfolio, with speed controls, project inspection and a full-page view.
  • Open Agent Flow in your signed-in workspace to see the selected project’s assigned agents, recent Board activity and Drive files, with optional live polling.
  • Add project Boards, private Drive files and scoped agent assignments to support collaboration, while keeping mailbox permissions separate.
  • Show stored portfolio measurements to the authorized portfolio owner, keeping unavailable metrics distinct from measured zeroes.
WHAT’S NEXT

The public demonstration is simulated. Private Flow reads saved activity; it does not run agents or publish changes. New projects need their own agent assignments and activity.

Make analytics setup and observed traffic easier to tell apart.

Connected sites now show whether analytics is merely configured or has recorded an event, with a clearer setup path for new websites.

What changed

  • Enabled privacy-limited pageview and claimed crawler collection for 43 existing connected sites after checking each live edge, site identity and public sitemap paths.
  • Show analytics as a separate site capability, including the difference between configuration and the first recorded pageview or crawler fetch.
  • Add analytics setup and privacy checks to the copyable website contract for static, blog and news sites.
WHAT’S NEXT

Sites on other website runtimes still need their own edge instrumentation and path review. Analytics counts browser events, not unique visitors; Cloudflare zone request metrics are separate.

Give each site a clear path to mail, newsletters and news.

A copyable Markdown contract now guides static, blog and news sites through Sprigpost setup, while brand newsletters require their own domain sender.

What changed

  • Added a site-root SPRIGPOST.md template that records each site’s identity, content source, signup choice and integration checks.
  • Serve the template as plain Markdown so a site-building agent can copy the current contract into another website project.
  • Document hosted and on-site signup, optional approved article feeds, news sections and the difference between Mail projects and site bindings.
  • Stop newsletter confirmation, welcome and owner test mail when the selected brand-domain sender is missing or unavailable; the Sprigpost-domain fallback is removed.
WHAT’S NEXT

The Markdown guides an implementation; it does not connect a domain or publish a site by itself. Recurring newsletter campaigns and automatic local Markdown import remain separate work.

Find each website and its mail setup in one workspace.

The private founder workspace now makes websites and Mail projects easier to find and shows what each site has configured.

What changed

  • Search websites and Mail projects in a visual, paged switcher with site images and useful tags, instead of long dropdown lists.
  • Find older numbered site profiles separately when their website connection still needs setup.
  • Open a selected website in Site tools and see separate setup paths for its website, mail address, newsletter and agents.
  • See site-specific sender, newsletter and agent-key signals in the directory without treating a shared Mail project as a mailbox for every site.
  • Load Mail settings and agent grants for the selected project, and load Site tools for the selected website.
  • Search configured workspace domains in small pages when adding an address, and choose the domain explicitly.
  • Expand directory capacity for the configured founder portfolio and speed up project inbox and agent-key lookups as it grows.
WHAT’S NEXT

Connecting a site to a Mail project does not create an address or agent key. Newsletter setup and delivery still require their own configuration.

Find the right site as your portfolio grows.

The private All sites directory now searches and filters sites and projects in manageable pages, with tags drawn from saved work.

What changed

  • Search names, domains and subjects, then narrow by status, project or tag without loading every site card.
  • Add private custom tags to a site or project and filter by clicking a tag.
  • Label existing and new work from saved topics, site lane, status, imports and public-portfolio sharing choices.
  • Keep public-portfolio sharing under the owner’s explicit visibility control; tags do not publish or grant access.
WHAT’S NEXT

The older Mail and Publishing selectors still load their full lists. The directory supports one exact tag filter at a time.

Open the local AI usage dashboard from your workspace.

The private founder menu now links to the local AI usage tracker.

What changed

  • Added a founder-only shortcut to the OpenAI view of the local AI usage dashboard.
  • Opened the tracker in a separate tab and labeled it local so its device and server requirement is clear.
WHAT’S NEXT

The dashboard still runs on the owner’s device; a private hosted usage page remains a separate future step.

Give each agent its brand’s own mailbox.

An owner-managed pilot connects an agent to one tested address on its brand’s domain, while the owner keeps sending control.

What changed

  • Issue and revoke expiring agent keys for one exact custom-domain address; platform-domain addresses are excluded.
  • Let that agent read its inbox and unread count, find recent verification messages, and prepare drafts held for owner review.
  • Keep the brand From address fixed when the owner reviews a draft, and show the agent its own sent and delivery status.
  • Keep Google and GitHub sign-ins in separate workspaces, as chosen for founder accounts.
  • Limit held agent drafts and pace reviewed sends by brand domain to avoid bursts after a backlog.
  • Compare Resend status and public receiving MX without changing DNS or treating a stored ready flag as proof of delivery.
WHAT’S NEXT

Each address still needs a receiving test and owner activation. Agent wake, MCP, and autonomous sending remain future work.

See images beside imported documents.

The owner import workspace now shows the photos linked to each document.

What changed

  • Recognize documents by an image thumbnail and a linked-image count in the list.
  • Browse linked photos above the saved preview or editor, and open any image at full size.
  • Show the exact image versions saved with a document, with captions and credits where supplied.
WHAT’S NEXT

Documents without image links show an explicit empty state; uploaded images remain available in the Images tab.

A clearer workspace for website imports.

Imported documents, images and connection settings now have their own focused views.

What changed

  • Choose an existing or new site in a compact import panel with clear destination cards.
  • Search documents by title or route, filter by section and browse manageable pages.
  • Open a saved preview beside the document list and switch to editing when needed.
  • Browse a redesigned image gallery with readable cards, grouped totals and filename search.
  • Use the same workspace in light or dark mode, with layouts adapted for phones.
WHAT’S NEXT

The import contract and publishing controls are unchanged; new imported documents remain private drafts.

Expand an existing site or start a new one.

The owner website importer now offers an explicit destination for a website content package.

What changed

  • Add private drafts and images to an existing site without changing its design, domain or published content.
  • Attach multiple packages to one site with separate stable record identities and repeat-import receipts.
  • Keep incoming route collisions for reconciliation alongside conflicts with saved edits.
WHAT’S NEXT

Packages require a version 2 handoff manifest. Raw research folders need preparation; importing does not design or publish a website.

Preserve AVIF images in imported websites.

Image imports recognize AVIF from its file bytes, even when a research download has a different filename extension.

What changed

  • Keep AVIF originals unchanged and serve them with the correct image type.
  • Check supported image formats during ZIP preflight, before uploads start.
WHAT’S NEXT

Supported formats are JPEG, PNG, GIF, WebP and AVIF; unsupported files are rejected intact.

Keep large imports moving through brief interruptions.

Website imports retry temporary transfer failures before asking you to resume a package.

What changed

  • Retry idempotent image and document writes up to three times with increasing delays.
  • Stop immediately for validation, integrity, sign-in and editing conflicts.
  • Wait for active image writes to settle before returning a failed bundle.
WHAT’S NEXT

Persistent failures still stop the import. Saved receipts allow the same package to resume without duplicate documents or images.

Browse the complete imported image library.

Website imports now have a private, searchable image browser with separate website, archive and working-file views.

What changed

  • Browse images in pages of 48, search filenames and open the original image.
  • Show unique image totals and explain images shared between groups.
  • Keep archive and staging images accessible without adding them to published pages.
  • Upload images in small, size-bounded bundles to reduce repeated requests while retaining checksum checks and resume receipts.
WHAT’S NEXT

Only files declared in an import package are uploaded. Working files can include rejected candidates and remain private unless explicitly used in a published document.

Import website ZIPs up to one gigabyte.

The private website importer accepts ZIPs up to 1 GiB, reading declared files in bounded chunks while preserving complete documents and images.

What changed

  • Read large archives without buffering the whole ZIP, and keep large manifests in private object storage.
  • Verify file sizes, checksums and declared sources before starting an import.
  • Raised the private per-document source limit to 1 MiB for owner-cleared website handoffs.
  • Added an explicit check for JSON expansion before saving a revision, so escaped control characters cannot exceed D1 history storage.
  • Kept public form limits and ordinary publishing article limits unchanged.
WHAT’S NEXT

Import packages remain private drafts until the owner reviews, connects and explicitly publishes selected documents.

Four book previews can collect release alerts.

Restored book websites can send readers to a hosted, consent-based release-alert signup while their previews remain free and non-selling.

What changed

  • Added launch-alert signup pages for Dinoroost, Origin Story, Romance House and Twenty-Eight Endings, with confirmation required before activation.
  • Added one welcome note after confirmation and publication-specific unsubscribe links, using each project’s ready Sprigpost sender.
  • Added support for restoring verified book-site links without replacing selected portfolio covers.
  • Kept the public portfolio at 69 owner-selected cards and retained each selected cover and visibility choice while the four projects joined Publishing.
WHAT’S NEXT

Release alerts are occasional and launch-only. Recurring newsletters, bulk campaigns, commerce, website analytics and publishing schedules remain off.

Every project has a place.

The public portfolio presents projects together, without separating them by workspace tools.

What changed

  • Share a project description and chosen cover even before its website is ready.
  • Projects without a website appear without an outbound link, rather than a broken destination.
  • Withhold known unavailable book-site links until their website releases are ready, while allowing their project cards to be shared.
  • Keep visibility under owner control and retain selected images. Mail, drafts and workspace details remain private.
WHAT’S NEXT

Sharing a portfolio card does not deploy a website, publish an article or activate email delivery.

Choose the right image for every site.

All sites now has a dedicated image chooser for directory and public portfolio covers.

What changed

  • Find supported images on a configured website, upload a cover, or choose a saved alternative.
  • Generate a fresh option in the configured owner workspace, with realistic editorial photography as the default direction and shared spending checks.
  • Keep earlier image options and restore the original cover. A new option is not applied until selected.
  • Keep public/private visibility separate: private images remain private, and withdrawing a listing also withdraws its selected public cover.
WHAT’S NEXT

Website and article imagery stay unchanged. Generated covers are illustrative, not documentary evidence; generation requires a configured provider and available budget.

Share your website. Keep your workspace private.

Website owners can now choose which projects appear in Sprigpost’s public portfolio.

What changed

  • Added a public showcase with website imagery, descriptions and links.
  • Added Keep private and Show in public portfolio settings to All sites for each workspace.
  • New listings are private by default. Sharing exposes only the approved website summary, never Mail, drafts or workspace details.
  • Visibility changes are revision-checked and recorded; withdrawing a listing removes it from the public showcase.
WHAT’S NEXT

Continue growing the showcase through owner opt-in. Portfolio visibility does not publish articles, send email or enable schedules.

Your sites, together at a glance.

The private owner workspace now brings publishing sites and Mail projects into one visual directory.

What changed

  • Added an All sites view with site imagery, short descriptions, search, status filters and sorting.
  • Added quick paths to each site’s publishing controls, linked Mail project and configured live website.
  • Kept private imports and Mail-only projects distinct from live publishing configurations, without creating duplicate site records.
WHAT’S NEXT

Connect future projects to publishing when ready. Viewing the directory does not publish content, send newsletters or enable schedules.

Bring a complete website into Publishing.

The owner workspace can import versioned website packages without shortening their articles or research records.

What changed

  • Added ZIP validation, complete source storage, source-linked previews, tables, image versions and revision-checked editing.
  • Added stable site and route identities, resumable imports and retained conflicts when a new package meets saved editorial changes.
  • Added domain and server-credential settings and an adapter for published pages and consent-based forms.
  • Kept imports as private drafts and preserved unknown source dates; supplied owner-cleared images do not require another approval step.
WHAT’S NEXT

Connect each imported site’s verified domain and deploy its adapter. Newsletter delivery requires a configured sender; import does not publish content or send mail.

Publication names now match their verified addresses.

Public-facing publication names now consistently match the verified custom domains while stable site records and editorial controls remain in place.

What changed

  • Aligned visible mastheads, footers and accessibility labels for the publications whose approved domain names changed their public brands.
  • Updated site 0027 to Love With Machines while preserving its established four-digit identity, archive and lovewithmachines.com domain.
  • Preserved article language, saved newsletter templates, audiences, schedules and content; this release does not publish articles or send email.
WHAT’S NEXT

Continue keeping internal working keys stable while using approved public names on reader-facing surfaces.

AI Companions has a verified home.

The existing AI Companions publication now uses lovewithmachines.com while retaining its established site record and editorial controls.

What changed

  • Connected the verified domain to site 0027 without changing its public publication name or four-digit identity.
  • Aligned the public origin and default reading-room links while preserving the existing archive.
  • Preserved saved history, consent settings, newsletter schedules and existing content; this release does not publish articles or send email.
WHAT’S NEXT

Continue domain changes only when ownership and the exact publication mapping are both confirmed.

Six more publications have verified addresses.

Six existing publications now use their newly verified custom domains while retaining their established site records and controls.

What changed

  • Connected Venture Termbook, Rotor Briefing, Soil Practice, Payment Rails Review, Backlot Current and Episode Compass to their approved domains.
  • Kept every four-digit site ID stable and aligned public origins, display names and default reading-room links.
  • Preserved saved history, consent settings, newsletter schedules and existing content; this release does not publish articles or send email.
WHAT’S NEXT

Continue treating unassigned owned domains separately until an exact publication mapping is approved.

Fifteen publications have their own front doors.

Fifteen existing publications now use their verified custom domains while keeping the same site records and editorial controls.

What changed

  • Connected the new domains to the existing publication identities and kept every four-digit site ID stable.
  • Aligned each site’s public origin, display name and reading-room links with its custom domain.
  • Preserved saved editorial history, consent settings, newsletter schedules and existing content; this release does not publish new articles or send email.
WHAT’S NEXT

Continue domain hookup only after ownership and the exact site assignment are verified. Unpurchased candidates and ambiguous owned domains remain unassigned.

A newsletter with each publication’s own voice.

The owner Publishing workspace now has editable newsletter drafts, email previews and a consent-based test path for the portfolio.

What changed

  • Prepared distinct welcome reading editions for 46 publications, reusing their public links and available site imagery.
  • Added editable subjects, introductions, reading cards, colours, signoffs and newsletter voice guidance, with revision history and protection for unsaved edits.
  • Added HTML and plain-text previews, verified-account test sends, hosted signup pages, confirmation emails and a welcome edition after confirmation.
  • Kept audiences and unsubscribe actions publication-specific, checked suppression before dispatch, and exposed queued, accepted and observed delivery states separately.
  • Reused verified sending identities and a clearly identified shared sender when a publication does not yet have its own verified email domain.
WHAT’S NEXT

Test and refine the welcome editions before enabling recurring or bulk newsletters. No bulk-send control or recurring newsletter schedule is enabled by this release.

An update whenever we ship.

The build journal is caught up, and release notes are now part of the deployment checklist.

What changed

  • Added the missing portfolio, workspace, research-agent and cost-control releases, with their original release dates and a separate journal catch-up date.
  • Replaced the Friday publishing promise with release-by-release updates, including the latest-entry link on the homepage.
  • Added journal validation and a pre-deployment check that requires a new entry compared with the previous deployed source.
WHAT’S NEXT

Keep each release note alongside the change it describes, with shipped features separated from experiments and remaining work.

A clearer view of the monthly bill.

Saved operating plans and portfolio-wide API spending controls are available in the Costs dashboard.

What changed

  • Compare API-only operation with optional local-generation scenarios, including cadence, research, source imagery, email, hosting and electricity assumptions.
  • Count already-paid subscriptions once and keep unverified costs unknown instead of silently treating them as free.
  • Save scenarios and API budgets independently, with revision history and protection for unsaved edits.
  • Enforce an optional portfolio monthly API ceiling alongside the existing site, run and daily limits. Saving a forecast does not start an agent.
  • Restored workspace access after a database capacity limit interrupted service; verified Mail and Publishing after recovery.
WHAT’S NEXT

Improve database query efficiency and fill in verified service costs. Local execution is still a planning scenario, not an installed worker.

The person behind the work.

Spotlight briefs now identify their subject explicitly, rather than turning into another campaign case study.

What changed

  • Added named subjects and types for people, companies, brands, initiatives and products in private Spotlight briefs.
  • Refined person-led profile instructions and added a source-reuse check before further review calls.
  • Kept research holds, incomplete usage and rejected drafts visible. A completed model review is not human publication approval.
WHAT’S NEXT

Resolve source-access and editorial holds and review the pilot profiles. A researched profile never implies an interview or endorsement.

Watch a researched draft take shape.

The owner portfolio can queue private, source-backed Spotlight drafts and inspect the agent’s actual progress.

What changed

  • Connected a bounded hosted workflow for source retrieval, drafting and independent model review, with per-run evidence and cost records.
  • Added private batch requests across sites, duplicate protection and shared daily spending limits.
  • Separated researched profiles from real interviews, and kept source failures, review holds and incomplete attempts visible in the Agent tab.
  • Kept all public Spotlight sections and automatic publication disabled. Queued requests are not finished articles.
WHAT’S NEXT

Improve source retrieval and editorial quality before approving any profile for public release.

One desk, forty-six distinct publications.

The owner’s 46-site portfolio is registered and connected to its shared publishing controls.

What changed

  • Expanded the site selector and saved editorial profiles to 46 distinct publications, retaining separate identities and voices.
  • Verified each site’s scoped public connection and content-section controls while preserving existing drafts and revisions.
  • Kept newsletters, forms and analytics readiness separate from website connection status; a connected site does not imply a configured sender.
WHAT’S NEXT

Continue bringing existing static article bodies into revisioned editing. The connected portfolio is not yet a complete migration of every archive.

Mail and publishing, side by side.

A shared workspace keeps project mail and the publishing desk within reach.

What changed

  • Added persistent Mail and Publishing navigation, with separate project and publication selectors.
  • Preserved open drafts and editorial work while moving between areas, with explicit unsaved-change controls.
  • Added a persistent dark-mode choice and kept publishing access restricted to the configured owner workspace.
WHAT’S NEXT

Keep improving the day-to-day workflow without losing unfinished edits or blurring access boundaries.

Know what a draft costs.

The publishing desk gained a provider-call ledger, cadence estimates and versioned model policies.

What changed

  • Added recorded API cost estimates, missing-usage warnings and daily or weekly scenarios based on available article samples.
  • Added per-site budgets and run ceilings with checks before a model call is admitted.
  • Added separate writer and reviewer selections with site overrides and policies pinned to runs.
  • Kept unconfigured providers and evaluations unavailable; saving a model preference does not make an API call.
WHAT’S NEXT

Build a stronger measured history and evaluate additional configured providers. Small pilot samples are not invoices or guaranteed future costs.

A shared home for distinct editorial voices.

The first production publishing dashboard brought site profiles, research and private editorial review into Sprigpost.

What changed

  • Added 35 initial site profiles, editable voice and image-style guidance, brand kits, source records and revision history.
  • Added private content drafts and structured page proposals with preserved sources, media and review state.
  • Added section switches and review-first publishing controls, with hidden interviews kept out of public content.
  • Kept the local publishing pilot separate from production release; importing a draft did not publish it.
WHAT’S NEXT

Connect the public sites deliberately, preserving their individual designs and keeping approval separate from import.

An email home for your agents.

A sharper direction for Sprigpost: the agent-first mailbox, with people in control.

What changed

  • Reworked the public website around agent identities, project continuity, and human oversight.
  • Added a labeled concept showing an incoming request, an agent-prepared draft, and human review.
  • Separated the working project mailbox from planned API/MCP access, agent permissions, and shared memory.
  • Updated account entry, the build journal, and link previews to reflect the agent-first direction.
WHAT’S NEXT

Improve mail reliability and domain onboarding, then build scoped agent credentials and the first API/MCP operations.

An address for every beginning.

Project addresses, a durable outbox, and a place for incoming conversations.

What changed

  • Added addresses on configured domains, with several names per inbox and controls to move or disable an address.
  • Connected reviewed drafts and their files to a durable outbox, with separate queued, provider-accepted, and delivery states.
  • Built verified incoming processing with private storage for complete message content, original email files, and attachments.
  • Added checks for repeated sends, signed delivery events, exact recipient routing, recovery, quotas, and private account boundaries.
WHAT’S NEXT

Verify the first live project conversations, then improve domain onboarding and recovery. Delivery readiness is shown separately for every configured domain.

Somewhere to pick up the conversation.

Project inboxes to explore, drafts to return to, and private attachments that stay with your work.

What changed

  • Added project switching and private sample inboxes, with message reading, search, stars, archive, trash, and restore.
  • Built compose, reply, and forward drafts with persistent storage and protection against overwriting changes from another tab.
  • Added a private founder studio for experiments, campaigns, content drafts, and recorded manual automation previews. Nothing is sent, published, or scheduled by a preview.
  • Added private file uploads to drafts, saved downloads, explicit attachment selection for forwards, size limits, and storage recovery.
  • Expanded the test suite to 33 checks, covering account protection, workspace isolation, concurrent edits, attachment access, storage quotas, and preview records.
WHAT’S NEXT

Connect the verified sending domain to durable inbox and outbox processing, exact recipients, retries, and export. Real mail delivery is not connected yet.

A front door of your own.

Google and GitHub sign-in are live, with a private workspace for each account.

What changed

  • Added normal account creation and login with Google and GitHub, plus persistent sessions and sign-out.
  • Gave each account its own private workspace, with server-side checks that keep one person’s workspace separate from another’s.
  • Verified both real provider sign-ins on the live site, alongside tests for session expiry, forged credentials, and workspace isolation.
WHAT’S NEXT

Make the workspace useful: an interactive inbox with sample messages, project switching, and compose. Connected mail still comes later.

A little more character. A much clearer direction.

A new identity, a closer look at the workspace we’re building, and the story behind Sprigpost.

What changed

  • Created an envelope-and-sprig mark and a warmer, more expressive public website.
  • Added an illustrative workspace preview, the founder story, and a roadmap that separates completed work from what comes next.
  • Opened this build journal. The planned publishing rhythm is weekly, every Friday.
WHAT’S NEXT

Build the first interactive mailbox prototype with sample messages, a project switcher, and a working compose flow.

Sprigpost takes root.

The name is chosen, the domain is connected, and the project has a home of its own.

What changed

  • Created an independent private project and brought over selected, reusable mailbox code from an existing application.
  • Documented the product direction, permission boundaries, and staged implementation plan.
  • Connected sprigpost.com and published the first website on Cloudflare.
WHAT’S NEXT

Turn the existing mailbox foundations into a human-first workspace for multiple projects.

More as it grows.

Explore the agent-first direction