What's new in August 2026

A background job you had just started could claim it had been waiting for weeks, because a re-run kept the clock from its first attempt. The queued and working readouts now restart with each attempt, and a genuinely long run reads as "2h 5m" instead of a wall of minutes.

A submittal of the exact parts the spec named came back mostly blank, with false not-mets on the specified modules. Now a product the section names satisfies its own article's rows, the datasheet values already extracted answer temperature, humidity, voltage and I/O in code, and ranges are compared as numbers, so "0 to 60" meets "0 to 40". The check reads the pages each requirement actually needs, a filed part is always judged, and equipment files itself from its own name. The matrix lists what needs you first, and a not-met you stand behind takes a declared deviation that prints with its reason in the package.

A PLC section against a sixty-part panel came back as sixty columns of not confirmed, because every requirement was asked of every part. Under the matrix each subject nobody answers to now asks which entry it is, File equipment walks the register once, and a requirement is checked only against the entry filed as its subject. A catalog number the section names is matched against the part itself, so a FortiSwitch submitted where the book says Stratix reads not met before a datasheet is opened. Feature requirements are rows again.

An item you add from the catalog is now called what the manufacturer calls it, and it stays that way as the catalog learns more names. The ordering code sits beside it as a chip rather than standing in for it, so a register, a submittal workspace or a search result tells you what the thing is before it tells you how to order it. Name a piece of equipment yourself and that name is yours; clear the field and it goes back to following the part.

Compiled manuals carry it too. The contents page prints the product name with the catalog number beside it, and the Equipment Schedule swaps its Equipment and Description columns for one Product column with room to print a full name, so nothing truncates. Issued Revisions are untouched.

Submittal rows are almost half as tall, because the description no longer needs a line of its own.

A catalog number like 5W5B3H meant nothing until you looked it up, and a spec that asked for an "Endress+Hauser ProMag W 500" never matched the part you actually had. Parts now carry the manufacturer's product name: it is read off the datasheet you attach, fills in for older parts with a one-time pass, shows up wherever the catalog number does, and new equipment lands in your register under its real name. Compliance checks see the name too, so a requirement that names a product can finally be met by it.

Starting an RFI used to land on a bare form: no breadcrumb, no title, and two identical-looking buttons for writing an RFI or uploading one you received. The page now opens like every other page, with the breadcrumb trail and a header that carries Preview and Mark sent. A clear "New RFI / Upload RFI" switch replaces the old buttons, the recipient comes from your project directory instead of a typed address, and the sign-off block only asks for the names that matter while drafting.

The Field Guides page was one flat list, so finding the guide for the phase you were in meant scanning every title. Guides now group the way a job runs: reading the spec, assembling the package, surviving review, then closeout. The landing page also answers the four questions that decide a trial before they become an email (time to first package, who sees your projects, what happens when you leave, portal workflows), and links shared from the site now unfurl with a proper title and picture instead of a bare address.

Picking from a dropdown inside a log's detail panel — a change order's "grew out of RFI", a punch item's classification — opened the option list somewhere else on the page entirely, and it would not always go away. The list now opens directly beneath the field you clicked, everywhere in the app, and closes the moment you pick, press Escape, or click away.

Reloading a page, or opening the package layout or cover page editor, used to white out the whole window for a moment while everything loaded again from scratch. The sidebar is now part of the page itself, so it is there the instant the window paints, and Submittal Kit remembers your workspace between loads instead of asking before it can draw anything. Opening an editor no longer reloads the app at all: it opens in place, the way moving between tabs does.

Checking a part against the specification used to read its whole datasheet, every page, for every requirement. The check now finds the handful of pages likely to answer the section's requirements and reads those, so a three-hundred-page manual costs no more than a cut sheet. Reading a part's datasheet for its description and attributes works the same way, and scanned datasheets are read through OCR instead of coming back empty. Verdicts still quote the datasheet word for word, with the page to turn to.

A section demands things of kinds of component: the CPU module's catalog number, the switch's operating temperature, and a few things of every part. Each requirement now sits under that subject, and the subject bar says which part categories it covers. The compliance check runs a requirement only against those parts, so a terminal block is never asked about the PLC's catalog number, and a subject with nothing on the register says so instead of costing a call.

Uploading a PDF used to wait while its text was read in the request, and a long manual could hold the page for a while. The upload now lands right away, the file is read in the background, and a look-alike offer shows on the document's row once the read finishes, with a way to keep the existing file or keep both.

A section that asked for twelve things offered twelve submittals, most of them procedures and notifications no datasheet answers. The pane now lists what describes the equipment, plus the O&M, and holds the rest under More suggested submittals, where they keep every answer they had. An O&M or a procedure no longer carries a compliance band or a matrix it has nothing to put in.

A curated section sets aside more requirements than it keeps, and every one of them sat in the list struck through, so the page was twice as long as the half worth reading. They now wait under Set aside, closed, with their reasons and their Restore intact. Read it again is the same button in the same place on the scope document, the requirements and the demand list.

Opening a section on the Specifications tab pulled the whole book out of storage every time, and a section with two hundred requirements sent six megabytes of page to draw a list. Neither happens now: a section opens in a few tens of milliseconds, and a requirement's checks and controls arrive the first time you open that row. Lines on the Scope tab load their edit form and section finder the same way.

An instrument section came back with 146 requirements for four instruments, a third of them from Part 1 articles specifying accessories nobody was buying. Reading a section now takes a blanket Part 1 rule but leaves the accessory specifications alone, keeps each row at the size a datasheet has a line for rather than one per screw, and cites a real article or none at all instead of writing "article number not shown". The same section now reads 89.

A scope document, a section's requirements and a section's demands can each be read again from what is already uploaded, instead of deleting the file and starting over. The list is cleared and rebuilt rather than added to, so a second read under better reading actually improves it. Anything you decided is kept: lines you typed, requirements you accepted, declined or judged, and demands you answered with a submittal.

A read, a compile, or an import used to report in on a timer, and every tick could close the step list you had just opened or drop you to the bottom of the Scope tab. Screens now hear about a change the moment it lands and refresh only what moved. Disclosures stay open, the page keeps its place, and answering one line of your scope repaints that line.

A section could be read again and come back with every requirement kept, even the ones the trimming pass had plainly decided to drop. The verdicts named rows by their article reference instead of the row itself, so none of them matched and each unmatched row quietly defaulted to being kept. Rows now carry a label an article number cannot be confused with, and a response that names rows we never sent is treated as a failed pass rather than as approval of everything.

Sending an RFI used to wait while the PDF was drawn, its attachments merged and the file stored. The record is now numbered and marked sent at once and the PDF is built behind it; the page shows Preparing the PDF until it is there. If the build fails, the record says why and offers Build PDF again.

A section like PRIMARY SENSORS AND FIELD INSTRUMENTS specifies every instrument a plant could carry, and you are buying four of them. Trimming its requirements now sees only the lines filed under that section instead of the whole project's scope, so requirements about equipment you are not furnishing drop out. And a section that specifies equipment but asks for no submittal of its own now says which section does ask, instead of showing an empty list.

Rows arrived that nobody could check: "Automatic zeroing" whose value was "automatic zeroing feature", "Transmitter construction" whose value was "solid state construction". A requirement now has to be something two datasheets could differ on, so those are gone and the ones that hid a real number came back with it: the low-flow cutoff reads "below 0.5 percent", the damping "1 to 25 seconds", and an output that was one row is now its signal, its action and its load.

Reading a section again for what it asks you to submit cleared the list and then sat silent for the whole minute the read takes, which looked exactly like a button that deleted your work and stopped. The section now says it is reading, how long it has been going, and how many sections are still to come.

"6.3 GHz to 26 GHz" was read as article 6.3 of Part 2, which cut the radar level transmitter's specification in half and left every requirement from it citing an article the book does not contain. An article of Part 2 is numbered 2-something, and that now settles it. A bare "7.5" in an equipment section was doing the same thing.

Reading a section used to hand you every restatement and every installation step as its own row, and you declined your way down to the ten that matter. The read now merges duplicates, sets aside anything a manufacturer's datasheet does not actually carry (lab test acceptance criteria, resin chemistry, jacket print legends), and lands the keepers on your list directly. Every removed row stays visible with a one-line reason and a Restore beside it, and anything you decided yourself is never touched.

Reading a scope document now shows its four steps with the finished ones ticked, instead of one sentence for the whole three minutes. The project's setup checklist keeps up on its own while those reads run, and says which one is running, so a page that used to need a refresh to notice a step was done no longer does.

You no longer have to open a drawer to find out whether a package answers the spec, or guess whether the matrix will reach the PDF. The submittal page now carries a compliance matrix panel with the counts on its own line and a switch that says, in words, whether the compiled submittal includes the table. The matrix is scoped to the sections that submittal is linked to, so a package stops claiming requirements nobody checked it against. Both bands on the page now start collapsed and remember what you opened.

You had to scroll past every section to reach the sharing panel, and it sat there greyed out on packages that were not compiled yet. Share now sits in the submittal header beside Settings and opens a window with the same recipient list, activity, and one-time-code links. On a package with nothing to send yet, the window says why instead of the page carrying a dead panel.

A supply line that never landed on a section had nothing you could press: the read had already run, and adding the line by hand did not start another one. The Scope tab now offers Match unfiled lines whenever any are waiting. Uploading another volume also re-reads the scope against a book whose earlier volume did not finish, instead of leaving those sections unmatched for good.

You were answering the same question twice: once on the Submittals tab and once beside the section that asked for it. Now you answer it once, on Specifications, where the section's own words are in front of you. The Submittals tab says how many are waiting and links across, and says nothing when none are. Two answers moved with it: you can point a demand at a submittal you already made, which links that section to it, and you can reopen any demand you have answered without touching the submittal.

A long list of checks before a submittal goes out used to push the sections halfway down the page. The band now collapses to one line, and opening it shows every check in a compact table that scrolls in place, with the kind of check called out beside each one. It remembers whether you left it open, and the count on an equipment row still jumps you straight to the list. Checks still never block a submittal.

Run check on a submittal reads each piece of equipment's datasheet against the accepted requirements for the sections it answers, and records met or not met with the datasheet's own words quoted as evidence; anything the datasheet does not prove lands as not confirmed. Verdicts land on the requirements panel with nothing to accept: each stays marked as suggested until you confirm it, and you can override or remove any of them. Anything not met shows up in the checks band before the package goes out, and if a requirement or a datasheet changes after a check, the affected answers are flagged as checked against older facts.

Matching your scope against the specification can only answer one question: is what you said you supply specified here. It cannot tell you what is in the book that you should have said, and those are the sections most often missed. All sections now sweeps the whole book for controls work, by section number where the numbering says so and by title where it does not, and lists anything you have not decided on yet. Work that only touches controls, raceway and grounding and surge, is listed apart from controls work itself, because who owns it is usually worth an RFI rather than an assumption.

Claiming a project used to mean answering "is this section yours?" over and over about a book that covers every trade on the job. A new Scope tab turns it around: tick what you provide from a list of control-system deliverables, paste your scope, or upload it, and each line comes back with the section that specifies it beside it. Sections nothing you supply reaches are never put in front of you, every line stays editable for the life of the project, and each one shows the submittals covering it, however many batches you sent it in. The old "Is this yours?" queue on the Specifications tab is gone; claiming lives in one place now.

Reading a dense section for what the equipment must be used to take one long call, and a section with a hundred requirements in it hit the answer size limit and came back empty three attempts running. Sections are now read a few articles at a time, so the requirements appear on the panel as each part is read instead of all at once at the end, and the panel says which part it is on. If one part does not come back the rest still do, and the panel tells you the list is incomplete rather than letting it look finished.

Deleting a specification removed the file and left every section it had been read into standing on the project, with nothing behind them and no way to take them off. All sections now has a tick beside each division and each section and a Remove button, so you can clear a division or a handful of sections; deleting an upload offers to take its sections with it, ticked by default; and if any sections are already stranded the tab says how many and clears them in one press. Removing a section drops its claim, its requirements and its register entries, but lines of your scope of supply are kept and go back to unmatched.

Every section of a specification was read one after another, so a project with a dozen claimed sections waited for the sum of all of them. They share nothing, so they now run several at a time and the wait is closer to the slowest section than to the total. A long scope document's pages overlap the same way. If every pass fails, the import names the failure instead of calling it too large. Background workers can pick up any kind of work now, so a burst of reading no longer queues behind itself while other workers sit idle.

Bringing a section into scope, by claiming it or by importing your scope, now reads it for what it demands you submit and what the equipment must be. Neither read has to be asked for, and the results appear as they arrive instead of after a manual refresh. The Specifications tab also keeps your place: reload and you are back on the section you were reading rather than the first one in the list. When your scope matches more than one section, each choice now says what that section is and why it was offered, instead of showing bare numbers.

Deciding which of two thousand sections are yours used to mean matching your scope against a table of contents, so "Fiber Conversion Panels" never reached the section that says "media converters". The import now reads the sections themselves and searches by meaning as well as by word, quoting the line behind every call, and it keeps what the contract actually said including work you supply but somebody else installs. It is reachable on a project that has never used it, reports every stage while it runs, and drops citations your book has no section for instead of claiming them.

Three cells printed the wrong thing: PROJECT # printed the project's name, PROJECT NAME printed its description, and the job number a reviewer files the package by appeared nowhere on the sheet. Each cell now prints what its label asks for, and PROJECT #, SUBMITTAL #, SPECIFICATION # and CONTRACT NO. all print the value when the project has one and otherwise leave a line to write on. The submittal date now comes from the same place the cover page and the comment response log read it from.

You write VFD, the engineer writes "variable frequency drive", and the search that should have found both found neither. Type either one now and you reach both, across spec sections, library PDFs and the palette; shorthand that means two things, like LS for a level switch and for a lift station, looks for both readings. The same vocabulary reads equipment tags back in plain English, so VFD-101 on a register somebody else built tells you what it is.

A package can be complete and still come back, because the panelboard's rating is below the available fault current or the transformer's temperature rise was never stated. Read a claimed section and it returns the characteristics it demands, each quoted with the article it sits at; accept the ones that apply, mark each Met, Not met or Not confirmed, and point at the page that proves it. Add a Compliance Matrix section to a submittal and all of it prints inside the package, so the reviewer audits your checking instead of repeating it.

Finding anything on the Specifications tab meant reading past everything else on it: an endless queue, panels counting how many sections had been read, and three drawers to open four levels deep. The tab now shows the sections you own down the left and one section, whole, on the right: its pages, what it asks you to submit in the spec's own words, and anything that has changed it. A section that names nine deliverables can be answered in one press rather than nine.

You watched a reader work, then watched the progress bar sit there long after it had finished and saved everything it found. Reads now turn the page over the moment they land, whether that is an uploaded RFI, comments pulled from a returned document, or a scope import. Previews open where you clicked them instead of reporting that the site refused to connect, and a specification you scanned rather than exported now reads from the text recovered on upload.

The Table of Contents no longer doubles its rules, and an empty section gets its tab number on the contents page instead of numbering that runs 01, 02, nothing, 04. The Compliance Matrix prints landscape with wrapping cells, so a required value lifted from the specification arrives whole rather than cut off mid-number. Divider pages stay portrait behind a turned section, and a long project name on the Transmittal stops at its own column instead of running under the ITEM # label.

Thirteen field guides are now up at /learn: what goes in a submittal package, how to read a spec section, what Approved as Noted obligates you to do, numbering schemes that survive long jobs, the closeout checklist, and what our own submittal logs say about real resubmittal rates. Written from the founder's own project experience, useful whether or not you use Submittal Kit, and each one is plain markdown at its address plus .md if you want to hand it to your assistant.

When an imported parts list spells a manufacturer differently than your library does, you used to work through every unrecognized name by hand. Now the review screen comes pre-filled. Names that look like one of your existing manufacturers arrive already pointed at it, marked Suggested, with the alias checkbox ready so the spelling is remembered next time. You still confirm every row, and a name with no good match is left exactly as before.

A change proposal or a G703 continuation sheet arriving as a PDF used to mean retyping every field by hand. The New Change Order page and the Budget tab now take the PDF itself. Drop it and the form or the line table fills in, with the numbers, dates and headings pulled straight from the document. You check every row before anything is logged, and if the file can't be read you get the ordinary form exactly as before.

After uploading an RFI you used to hunt for what the reader had filled in, behind a collapsed form and a pop-up demanding the sent date. The record now leads with one card, "What the PDF said": every field the reader filled, next to the line on the PDF it came from, and one "Looks right" folds it away. Two fixes went with it, a dropped PDF actually starts the read now, and the sent date you record sticks the first time.

Searching document contents used to be literal to a fault: "brackets" missed the page that says "bracket", and a scanned datasheet was a dead end with no text layer to index. Search now matches word forms and puts the pages carrying all your terms first, and pages that turn out to be images are read individually and indexed like any other upload. Catalog numbers and ratings still match exactly as typed.

What does hand-assembling packages actually cost you a year? The new worksheet at /roi does the arithmetic with your numbers: your package count, your hours, your rate, your resubmittal share. No industry averages baked in anywhere. The defaults are a worked example drawn from real project submittal logs, and every one of them is yours to change. Share it with whoever signs off on tooling.

Answering a page of review comments means writing the same kind of sentence over and over. Pause while typing and a grey draft of the next words appears, written from the comment itself and the way you have answered similar ones on this project; press Tab to keep it, or keep typing and it steps aside. Nothing is saved until you accept it, so your log only ever holds words you chose.

The transmittal always printed an empty NOTES block, so what a package contains and what changed since the last revision got typed into an email or nowhere at all. The transmittal section now has a cover note. Type it once and it prints on the next compile, and Draft note suggests a starting paragraph from the files that changed and the review responses the package carries. Issued revisions are records and stay exactly as they went out.

A recent update tucked Download PDF, Preview, Recompile and the source document export behind a "..." menu on the submittal page. They are visible buttons again, right where they used to be. The "Pending issued package" box is gone too, because the confirmation you get when you Mark as Submitted already names exactly what will be issued.

Filtering the Contact List and changing a role, expanding a firm, or keeping a punch candidate used to redraw the whole page. Your filter snapped back and open panels shut while your search text still sat in the box. Each of those now touches only the rows it changed, so your filter, your scroll position and your half-typed edits stay put. A rejected save hands your typing back too, so fixing a near-miss is an edit, not a do-over.

Getting a compiled package to the reviewing engineer used to mean attaching a 40 MB PDF and hoping their server took it. A submittal now has a Share panel: pick a contact or type an email, and each recipient gets their own link to a transmittal page with a download button. The link is bound to their address and opens only after a one-time code, and you see per-recipient views and downloads, with a revoke that works the moment you click it.

Hard to argue for a tool when the value it returns is invisible. Settings > Billing now shows your numbers: packages issued, hours recovered, labor recovered, and your subscription cost with the return multiple, with a compact version on the dashboard. The two assumptions behind the math, hours saved per package and your loaded rate, sit right next to the figures and are yours to edit, so the numbers stay honest to how your shop actually works.

A corrupted or oversized file used to fail somewhere mid-import, or grind the page trying. Every spreadsheet import now checks a file's real size before parsing it and tells you exactly what is wrong, and PDF and image uploads are checked the moment they arrive. The SOV import no longer reads legacy .xls files, so re-save as .xlsx or .csv before uploading.

Finding a phone number used to take two steps: pick a firm from a narrow list, then read its detail pane. Firms & Contacts is now one table with people, projects, phone and email in columns, and clicking a firm expands its team right in place. Delete only appears on records nothing references yet; everything else deactivates, so history keeps its names.

Kick off the wrong compile or point a scan at the wrong datasheet, and until now your only option was to watch it finish. Every progress surface now carries a plain Stop: datasheet scans, compiles, source exports, part descriptions, BOM imports, punch extraction, scope parsing, document finding, RFI reads, and MOF exports. A stopped job lands in a clear stopped state and stays there until you choose to run it again; nothing requeues behind your back. This also fixes the datasheet scan that could sit at "Looking for this catalog number on the page" for hours. A stuck handoff kept recycling the same work forever, and scans now run to a real finish.

A library document can back parts across many projects, and editing its category or Always Full Sheet setting used to apply everywhere silently. When the file is shared, the edit now tells you first exactly how far it reaches, with live counts of the parts and projects linked to it, the same warning the part page already gave. A document used in one place still edits without ceremony. Also new: What's new lives in the help menu when you're signed in, opens as a quick panel instead of leaving your page, and shows a dot when something shipped since your last look.

After a reviewer returned a package, two questions had no answer on screen: what did this revision actually change, and did anyone go through the returned document? Each revision's drawer now answers both. A changes-since line states what the package added, replaced, or removed compared to the revision before it, with the full list one click away. Under the returned document, a summary shows the outcome, the reviewer, the date, and any pages the reading could not cover, so you know at a glance whether the response still needs your eyes.

Running several jobs meant answering "what's late, everywhere?" by opening each project and checking five surfaces one at a time. A part twelve days past its order-by date, a slipping FAT, a change order waiting in a customer's inbox: none of it reached the dashboard. Now it all does. A schedule rail lists every dated item across your projects in time order, broken dates first, then the next thirty days. Below it, your open work splits into "In your court" and "Out of your hands", each row tagged with the register it came from, so the next action is always one click away.

Figuring out a corner of Submittal Kit used to mean poking around or emailing support. Now every feature has a written guide with screenshots at /help, public and shareable. Search the guides with Cmd+K on the help site, and matching guides now appear in the app's own search under a Help group. Your Help & feedback menu links straight there, and any guide is plain markdown at its address plus .md, easy to hand to your assistant.

Opening the site while signed in now lands on your dashboard directly instead of taking a redirect hop first. Logos, icons, and fonts are cached by the browser for the long term rather than being re-downloaded, and the Part Library page trimmed redundant counting work. The difference is most noticeable on company laptops whose security software slows every request.

Milestones were ten loose dates you pushed around one at a time. Now each one can carry a duration, follow another milestone with an optional lag, and the dates work themselves out: slip FAT and Ship, Materials on Site, and everything after move with it. A timeline above the table draws the plan as bars you can drag and resize, with week, month, and quarter views and a today line. An empty schedule can start from the standard sequence, Kickoff through Substantial Completion already chained with typical spacing, so the first draft takes one date and one click. Recording an actual date pins that milestone where it really landed, and the rest of the plan flows from it. Procurement's need-by dates and the Submittal Log's proposal hint read from the same schedule, so there is one plan, not three.

The schedule also works the way a schedule is used. Durations and lags count working days, so weekends stop hiding inside "10 days of fabrication". Every change that ripples shows an Undo, and a history keeps who moved what, when. The chain that decides the finish date gets a heavier outline, milestones that drive purchasing say so on their row, and dragging a bar shows the date it will land on. Dragging a dependent bar slides its lag behind its predecessor, so the whole chain stays honest. Project followers hear about it when one edit moves the finish date a week or more, and the whole timeline exports as a PDF Gantt or a spreadsheet.

Submittals put together by hand, outside the application, had nowhere to live: the log could note that a revision went out, but not keep the document that actually went. Drop that PDF into a submittal's Revisions section, fill in the revision number, dates, and what came back, and it becomes a real revision on the Submittal Log, marked as recorded so it reads apart from one the application compiled. The next revision is built here as usual and continues the numbering.

The Revisions list on a Submittal only offered a download, so checking what went out in an earlier revision meant saving the file and opening it yourself. Each revision now opens in the same preview window the rest of the application uses, with the download still there beside it.

A recorded revision is typed in after the fact, so the usual mistake is the wrong file or the wrong date rather than a bad build. You can now correct one: change its number or dates, drop in a replacement PDF without re-entering everything around it, or remove it entirely, even when it is not the newest. All of that stops once the reviewer's returned document is attached, at which point the revision is part of the permanent record like any other.

Importing a BOM could stop with an error when a part was already in your Part Library under the manufacturer's saved spelling and the spreadsheet shouted it in capitals. The import now files the row against the part you already have instead of trying to create a second copy of it.

The importer is also better at reading a workbook. It matches column headings by whole words rather than stray letters, so "Manufacture", "Part #" and "PartNumber" are all understood and a column like "Finished Water QMS" is no longer mistaken for a quantity. It reads the rows underneath a heading before deciding, so a leftovers tab loses to the live parts table and a column that simply counts 1, 2, 3 is never taken for a part number. When one sheet is the clear winner, the import skips the mapping step entirely.

Adding a section to a submittal updated the section list, but the button you added it with kept offering to add it again and the section itself did not appear further down the page until you reloaded. All three now update together, as do renaming, deleting, and reordering.

The Submittal page shows every comment on the project along with the review or site visit that raised it. It used to look that source up one comment at a time, so a project with a long comment list spent longer and longer loading the page. The page now collects all of that in a single pass, and its loading time no longer grows with the number of comments.

Progress used to look different depending on where you were: a yellow-striped band on imports, spinning circles under some drop zones, and a "Background work" box that popped up in the corner of the screen. All of that is now one quiet design. When you choose a file, the drop zone becomes a small card with the file's name and size, a thin blue progress bar, and one line saying what is happening right now, with the real percentage when we know it, the step count when the work moves in steps, or how long it has been running. When a job keeps going after the upload, the same line tells you that you can leave the page. Finished work turns the bar green with a plain-words receipt; failures turn it red and say what actually went wrong, with the file still attached so Try again means try again. The corner box is gone, progress stays where you started the work, and it is waiting there when you come back.

Pressing Compile used to sit for five to ten seconds before anything changed on screen. Behind the button, the server was re-downloading every source PDF in the submittal just to double-check nothing had changed, on every compile, every download, and every submit. Those checks now use the record the app already keeps, which answers in a blink, and the heavy file work happens only inside the background compile job where it belongs. The submittal page also stopped running a specification search on every visit; it now runs only when you open the settings drawer where those suggestions appear.

Uploading a file you already had could add a second copy of it, so one datasheet appeared twice and each copy could be linked to different parts. Re-uploading the exact same file now always lands on the entry you already have. A vendor datasheet downloaded a second time is harder, because the file differs even when the document does not, so those uploads still finish as before and the library and part pages tell you when a new file reads the same as one you already have. They show what it matched, when it was added, and a link to open both, since two sheets in a product line can read alike and differ only in a photo. One click links the part to the file already in your library and drops the copy, or you can keep both.

Compiling a submittal could fail with an error page when one of its source documents could not be read, with nothing on the page explaining which document was at fault. The compile button now tells you what needs fixing before it starts, and the same documents are flagged on the submittal itself instead of looking ready. Clearing the Submitted date on a log entry also failed instead of saving, and now reports the problem inline. Compiling is substantially faster, most of all on the packages that took longest: drawing-heavy manuals, and any submittal that covers part of a large project. A package now pulls only the documents it actually contains rather than everything in the project, and documents it has compiled before are reused instead of fetched again. The Purchase Orders tab also stays quick as a project's order history grows.

With the sidebar collapsed, the help and account menus opened underneath the page behind them, so log rows and other content showed straight through the menu. Both menus now sit above the page wherever they open.

Finding the spec paragraph or RFI answer you half-remember used to mean opening page after page. Search (Ctrl/Cmd K) now reads the text inside RFIs, punch responses, change orders, equipment notes, and recorded spec requirements, shows the matched line with a preview beside the results, and pressing Enter on a spec section opens its pages right there, at the matched page. Your search sticks around if you close and reopen, and the empty search remembers where you have been: press Enter to jump back to your previous spot.

Someone joining a company mid-project used to land on a dashboard listing every package the whole company owed, under a heading saying only they could move it. Now their first dashboard asks which projects are theirs, showing what is out for review and what has come back on each so the names mean something, and it turns into their real board the moment they pick. Admins can put a new teammate on their projects right from the invite form, and the invitation email names those projects. On the Team and Personnel rosters, the amber flag on a name now reads "Name is a guess" and explains that the name came from the person's email address and is what an Emergency Contacts page prints. Setting your name on your Account page finally corrects the roster too, which it never did before, so the flag clears and documents stop printing the guess.

A dropdown opened near the bottom of a scrolling panel could be cut off, leaving the last option impossible to pick. Every dropdown in the app now opens over whatever it sits inside, with the full list reachable.

Every project now keeps a Change Order log beside its RFI Log and Punch List: numbered records with status, dollar amount, and the RFI each change grew out of. Open a record right in the log and it edits in place, saving each field as you leave it, marking one approved, rejected or void without the page moving, and taking several quotes or markups at once dropped straight onto the record. Bring your existing log in by dropping the spreadsheet on the page: it reads layouts that aren't laid out like ours, including headers split over two lines and the Pending / Approved / Rejected columns a lot of PCO logs use instead of a status column, picks up the day each change order came back whether your sheet calls it a response date, a return date or a decision date, and leaves the subtotal at the bottom out of the log. Anything already logged is spotted by number, or by matching title and amount, and you decide row by row in a window over the log whether to skip it, bring it in again, or update the record from the sheet. When something lands wrong, correct a change order's dates or its number on the record, or delete it outright.

Procurement now lives beside the equipment it covers. Drop a vendor quote on a project and the line items are read from the document, matched against your parts, and priced. Build purchase orders from the quotes you accept, then see at a glance what has been ordered, acknowledged, and received.

Every part now shows its full price history across projects, the quotes behind it, and what is on order, right on its page. Duplicate parts (a quote's full ordering code next to your catalog number) are spotted automatically, and cleaning them up is two clicks: move the quote line to the right part, then delete the empty one. The Parts and Document Libraries share a faster table layout with last-quoted prices and linked-parts columns.

Exporting the MOF failed with a generic error when a Part Category name contained a character Excel does not allow in sheet tabs, such as a slash. Those characters are now replaced in the tab name, and the export reports its real progress instead of failing immediately.

The dates that drive your submittal schedule (order-by, ship, FAT) now live on the project itself instead of a spreadsheet on someone's desktop. The milestones strip shows what is coming due across the work you follow.

Merging two parts failed with an error when the duplicate had quote pricing or purchase order lines recorded against it. The merge now carries that pricing history over to the surviving part.