LandSight is the solution of how fast, accurate and effective to review, evaluate and manage the project include but not limited to architecture and landscape
Any where near to AI
Joined December 2023
- Tweets93
- Following35
- Followers63
- Likes76
Pinned Tweet
One drawing sheet can stand for 4 identical floors. The bill priced 1.
Mistakes like that hide in every construction drawing package, and someone has to find them by hand before a firm can quote. For a landscape package, that's about 2 weeks.
So I built Landsight.
You drop the developer's folder in as it came: CAD drawings, scans, Excel bills, backups. It gives every file a verdict, counts and measures what's drawn, and checks the bill against the drawings, the firm's price list and the building code. Then it drafts the change request and the reply to the developer.
Every number traces back to the drawing it came from, and nothing counts until a person confirms it.
We planted 30 errors in a sample package. It found 28, flagged 2 for a person and missed none. About 10 minutes instead of 2 weeks.
Under the hood:
- Muse Spark 1.3 reads what geometry can't settle and writes the prompts SAM 3.1 uses to measure colour plans.
- Muse Wonder learns each firm's habits.
- And because drawings are under NDA, there's a private mode on Muse Spark Open, Meta's open weights, with the same result.
Built for the Meta Global AI Developer Hackathon @AIatMeta @cerebral_valley
What's the worst drawing error you've seen?
#MetaAI #ConstructionTech #AI
2 weeks → 10 minutes. On open weights. ⏱️
That's checking a landscape drawing package by hand, then with Landsight.
30 planted errors in a sample package:
✅ 28 found
🙋 2 flagged for a person
❌ 0 missed
Same scoreboard on Muse Spark Open, @AIatMeta's open weights, on a server the firm chooses. Open weights aren't just for chatbots. 🔒
Checking the drawings now takes minutes. Building is still the slow part.
How far are we from building on Mars? 🚀
Try it in your browser, no install 👉 landsight.co/try
The full story 👉 teapaw.org/landsightdotai/status/…
One drawing sheet can stand for 4 identical floors. The bill priced 1.
Mistakes like that hide in every construction drawing package, and someone has to find them by hand before a firm can quote. For a landscape package, that's about 2 weeks.
So I built Landsight.
You drop the developer's folder in as it came: CAD drawings, scans, Excel bills, backups. It gives every file a verdict, counts and measures what's drawn, and checks the bill against the drawings, the firm's price list and the building code. Then it drafts the change request and the reply to the developer.
Every number traces back to the drawing it came from, and nothing counts until a person confirms it.
We planted 30 errors in a sample package. It found 28, flagged 2 for a person and missed none. About 10 minutes instead of 2 weeks.
Under the hood:
- Muse Spark 1.3 reads what geometry can't settle and writes the prompts SAM 3.1 uses to measure colour plans.
- Muse Wonder learns each firm's habits.
- And because drawings are under NDA, there's a private mode on Muse Spark Open, Meta's open weights, with the same result.
Built for the Meta Global AI Developer Hackathon @AIatMeta @cerebral_valley
What's the worst drawing error you've seen?
#MetaAI #ConstructionTech #AI
Doing a AI project within a weekend is easy
Go to market of even find A users is harder than most ppl thought
People always enjoy on “creating” and AI make the cost of money or time lower enough to see the outcome and enjoy the dopamine
But the next step is much harder because these reason:
This week's results make me optimistic. We have an unprecedented bipartisan opportunity to take on Big Tech’s abuse of market power and open up markets for the next generation of founders.
Inventing powerful AI is only half the job. People must be able to use it to make their lives better and their businesses more productive. That gets harder when a few trillion-dollar companies can block new products from reaching them.
Congress can open those markets for Little Tech. Let thousands of startups compete to lower costs and grow the economy for everyone. We should spread the gains from AI as widely as possible, not concentrate even more power in the companies that already control distribution.
Checking a landscape drawing package by hand takes about 2 weeks. Landsight does it in about 10 minutes.
30 planted errors in a sample package: 28 found, 2 flagged for a person, 0 missed.
Nothing counts until a person confirms.
What's the worst drawing error you've seen?
All built by #Meta #Muse #SAM31 @finkd
One drawing sheet can stand for 4 identical floors. The bill priced 1.
Mistakes like that hide in every construction drawing package, and someone has to find them by hand before a firm can quote. For a landscape package, that's about 2 weeks.
So I built Landsight.
You drop the developer's folder in as it came: CAD drawings, scans, Excel bills, backups. It gives every file a verdict, counts and measures what's drawn, and checks the bill against the drawings, the firm's price list and the building code. Then it drafts the change request and the reply to the developer.
Every number traces back to the drawing it came from, and nothing counts until a person confirms it.
We planted 30 errors in a sample package. It found 28, flagged 2 for a person and missed none. About 10 minutes instead of 2 weeks.
Under the hood:
- Muse Spark 1.3 reads what geometry can't settle and writes the prompts SAM 3.1 uses to measure colour plans.
- Muse Wonder learns each firm's habits.
- And because drawings are under NDA, there's a private mode on Muse Spark Open, Meta's open weights, with the same result.
Built for the Meta Global AI Developer Hackathon @AIatMeta @cerebral_valley
What's the worst drawing error you've seen?
#MetaAI #ConstructionTech #AI
Drawings are confidential, so there's a private mode on Muse Spark Open, Meta's open weights, in a deployment you control.
Same result on the sample.Firm memory runs on Muse Wonder: it remembers what your firm confirms and dismisses.
Anyone interest in join the waitlist, please filled:
landsight.co/#trial
What a cool stuff, and we could make it faster
(Maybe not within 10mins if we are talking about Taipei 101 of course😂)
I built an interactive #Taipei101 damper simulator.
Turn up the wind and watch a 660-tonne steel ball swing against the tower. Then lock the ball and watch the same building sway on its own.
Illustrative sim with assumed building parameters, not an engineering model.
The private mode runs Muse Spark Open (meta-models/muse-spark-open-1026) on Baseten
and it can move to a firm's own servers, so drawings never have to touch a public cloud.
Try it in your browser, no install: landsight.co/try
Website: landsight.co
The full 3-minute film: youtube.com/watch?v=ObB8pePC…