Constructable

At Constructable we yearned to be able to query our data in our frontend as easy as we could in our Rails API using ActiveRecord. So we built it! It's called the "Object Pool", and it's a great example of the power of sync engines: Each model has its own file that defines its attributes and relations to other models. For example, an "Account" from our accounting module. On page load, our sync engine Replicache pulls all the current data from IndexedDB and loads it into the Object Pool. Then, in any component, you can synchronously query for any model and access all its relations. In the background, Replicache keeps the data up to date with the server, and any changes are immediately reflected, making every component "real time". This also gives us a central place to define common logic related to each model, resulting in an easy to use domain model that enables us to build tightly integrated experiences across our product without rebuilding APIs and duplicating code. I'm beyond happy with how our bet on Replicache/sync is playing out. It's taken us a bit to find the patterns that work best, but as we continue to iterate on these systems, I find that I've never enjoyed working in a product as much as I do Constructable :)

?? ?Ghost Mode - a way more fun way to overlay sheets! Getting new revisions doesn't have to be scary anymore. ?? Another simple yet important feature for viewing drawing sheets and revisions. And yes, you can even overlay sheets that aren't revisions - Landscape and Civil anyone? #contech #overlay #blendmodetm #ghostmode #constructable

To wrap this year’s out-of-this-world UX Fest experience, we wanted to give a shoutout to our designathon judges Angie Delgado and Blake Feldman! Also we appreciate our amazing speakers Christine Lee Chiang, Eugene Chang, and Mark Edwards for setting us up for success with valuable UX skills! Everyone who participated came up with incredible designs and we can’t wait to see what’s in store for next year! ????

?? Why Search Alone Can’t Save You in Construction (And What to Do About It) We all know that a typical construction project comes with an insane amount of information, most of it being text. Let's say you’re working on a 30-unit apartment complex. There's a lot of reading involved—equivalent to cranking through many large novels. And that's just the text! Now add in images, details, and callouts, and you’ve got a full-blown scavenger hunt on your hands. We all know you obviously read every single word (of course!) But on that one rare occasion when you can’t find a callout on a sheet, you’re going to rely on search tools. So you type it in, hit search, and—drumroll—no results. Weird. You try a few different keywords. Nada. Eventually, you shrug it off, assuming it must’ve been from another job. Fast forward a few months, and surprise! You missed a crucial detail. The result? Thousands of dollars in delays, and let’s not even mention the sleepless nights that follow. Moral of the story: WE CANNOT RELY ON SEARCH ALONE. Turns out, there are two types of PDFs, and from the outside, you can't tell them apart. A vector PDF is like having the full blueprint, while a raster PDF is conniving and tricky–it looks like it has everything but, in reality, the computer only sees dots. Not words, not images—just a bunch of specks. When we first built Constructable, we thought we could ignore raster - most PDFs seemed to be vector anyway, right? Then one of our customers missed a crucial callout on a bid. Oops. So we dug in, did our homework, and realized something pretty alarming. A single PDF sheet can have both vector and raster elements. So even if you think everything on a page is searchable, a portion of it might be raster—leaving part of the sheet totally invisible to search. Our analysis showed that, on average, 25% of drawing pages have raster components, meaning search only works three-quarters of the time. So when you’re searching for "Trash Enclosure," you’re only getting about 75% of the results. This does not build confidence. Yikes. That’s when we said, enough is enough. Enter OCR—Optical Character Recognition. This nifty tech is like a pair of eyes that scans those pesky dots and turns them into actual, searchable text. So, the real question is: Is your software OCRing every drawing sheet, every PDF, and every submittal attachment? If not, you might be missing out on critical details. At Constructable, we’re here to make sure you never miss a thing. That sneaky Raster?PDF doesn’t stand a chance. And to close out this post, I will quote one of our customers “Your search is pretty Bitchin!” Note: (not an AI generated image) Also, this image would become searchable if it appeared in a drawing in our product! Check us out Constructable ??♀? ??♀? ??♂? #contech #construction

Building a local-first sync engine into the heart of Constructable was a big departure from the "Rails way" I came to know and love, but it has been a huge win for both us and our customers. Check out my latest post where I wrote about our journey into the local-first world: https://lnkd.in/eCw3ZajX

A flavorful discussion of how AI is used to search by meaning

A high level explanation of how AI fits into document search, with a particular focus on construction project management. Enjoy!

S23 + W23 Batch Retreat! Got a chance to see some absolute killer founders from our S23 and W23. It's awesome to see everyone's growth, wins, and challenges over the past year! Shoutout to some of the most talented folks building across the globe: Will Lawrence from Greenlite Atul Raghunathan and Sriharsha Guduguntla from Hyperbound Alan Wang from Serra John Yoder, Emma Nazim, and Molly Abbott from Constructable Amy Yan from Nowadays (YC S23) Adrian Johnston from Elyos Energy Sajag Agarwal from Movley Chris Parsonson, Angus Parsonson, and Sanj Ahilan from Solve Intelligence Gonna be awesome meeting up with all batch founders at Alumni Day today!