I'm a PE who has been tinkering with LLMs for a few years now. Opus 5.5 does seem to be a game changer for drafting. I don't generate with python tho, I have it drive AutoCAD headlessly and use our firm's drafting standards/templates. Its a junior-level EIT at first, but I give it redlines via Bluebeam MCP and it updates the relevant CAD SOP for all future sessions.
I am not an expert, but I’ve been tinkering architectural blueprints since Opus 5.5 released. From what I’ve learned, the vanilla results are a bit dated w.r.t CAD standards. Seems the only public version of the US NCS is v3 while the latest is v7. I ended up having to buy a licensed copy of v7 and develop my own MCP server to explore the standard. The result was a fairly significant rewrite of the application I wrote to comply with the latest requirements (see http://app.factoryplanner.net/try)
I am too. Doesn't that concern you for the future? I've been meaning to connect it to Revit but haven't had the time.
Edit: I'm not sure how fast any of this will be adopted with how slowly things move in construction. I'm surprised people are still using AutoCAD and making a profit.
Why would it concern a PE. 99.9% of their job is already stuff someone else could do and would rather not pay them to do but the government forces them to anyway. AI doesn't change that except perhaps remove a bit of the plausible deniability.
Now, I know, I know, I'm about to be hit with some bad faith retort about skyscrapers and bridges and other things engineered on a fine margin falling down. Your average multi tenant commercial building, your average parking lot, your average retaining wall ain't that. And that's the stuff that's the bulk of what PE firms do.
Because there are thousand of PEs across the country and eventually if a job requires 100x less time to produce, PEs will just be rubber stamping stuff after a thorough final review. My firm has like a dozen electrical PEs and we all spend the vast majority of our time in coordination meetings, reviewing small details, and working on drawings. If all I have to do is spend a few hours checking something at the end, they don't need all of us for that anymore.
Given how good the leading models got at 3D model creation over the past few months - Blender, Three.js, etc - it's not surprising to see that their 2D CAD-style abilities have improved to this extent as well.
(Amusingly the pelican-on-a-bicycle test doesn't capture this very effectively, despite testing 2D drawing with SVG - because pelicans can't write bicycles, so the results are inherently unrealistic. Having the model create an architectural diagram from a concept sketch is a much more useful 2D test.)
I worked almost 15 years in commercial concrete. My day to day tools were Revit and Bluebeam. As a little fun project I built a replica of them and they work surprisingly well. I was able to detail some of my previous jobs in an hour which took me weeks before. It is truly crazy the time we are living in.
Might be nice to scope out where exactly it is good and where it is lacking ; residential vs commercial vs industrial , A vs S vs MEP, style of architecture? We need better benchmarking here to substantiate.
Agreed. Also the consistency between the different drawings would be nice to have an understanding of. This is where a higher level 3d domain model (such as IFC), do validation, interference checking etc., and generating drawings from those makes more sense (in my mind; but I'm biased). We did a quick demo once [0] but it ran out-of-date quickly due to the hardcoded model ids.
“Permit level” is doing a lot of lifting. Standard vary greatly by locality and project. Plus, permitting is a much lower threshold than actual inspection and sign off (which is where the issues will be caught).
A build permit simply says that they reasonably believe you can build the thing such that it will pass inspections. It doesn’t mean that every single detail will be finalized or correct.
Have you considered the possibility that they meant exactly what they stated? That the output is good enough to reasonably believe it can be built and pass inspection,but it may still include flaws in details?
AI can draw that plan for you if you give it the prevailing wind and the summer and winter sun data. But it can't take a soil sample from the land for you, and it can't know what kind of soil it is or what that soil will or won't tolerate. From a distance it can offer a nice picture, but without data it can't reach reality.
Washington State is trying out a 'Permit-Ready Plans' program in a few cities, were pre-approved plans are offered at no cost. According to their initial study, building plans account for 5-20% of total construction costs. Software that might be able to shave 10% off new construction costs is a big deal.
I looked into one program and they are offered free of cost. Currently, they mostly consisted of ADUs and dense housing, which seems to align with they city's housing goals.
A big reason LLMs can write code productively is cause software dev tends to be pretty resilient to failure. And at the pessimistic end, lots of companies can make money off buggy products, so it’s easy to hand wave the failures away. Does that hold true for architecture?
That being said, I’m certain they’ll be extremely useful in other areas such as review, auditing, etc - if those even matter for architects.
I've reviewed dozens of sets of city and state AHJ review comments for housing. They sometimes take a full year and give less attention to the plans than even the worst contractor. Some aren't that bad, but many just have boilerplate review comments that are completely irrelevant to my drawings.
Civil engineering rules are mostly written in blood and costly and dangerous accidents. At least in my country. If somewhere else inspectors don't provide actual value, they should either get regulated or good riddance and AI might actually be better. I feel old, sigh...
That's not what we're talking about. By all means, follow the structural engineer's specification. But if you've ever submitted plans to a city to build a home, there is endless nonsense full of red tape: Drainage plan, soil analysis, tree survey, topographical survey, environmental site assessment, wind analysis, noise assessment, and more.
Sometimes these things make sense, but a lot of the time the work was already recently completed 3 years ago and yet they'll make you do it again. Last month I paid $3,900.00 to get 3 tiny trees and one medium sized oak surveyed here in the Bay Area. It's really about politics and control- the city wants to make a bunch of fake jobs and collect very high fees along the way.
Nearly everything you listed is designed to prevent your decisions, or changes to your property that haven't been identified yet, from harming you or people around you.
Removing a requirement might just leave someone else paying a much higher price, potentially with their property, their clean water, or even their life.
To be clear, the utility of these processes can be good - but it also gets perverted into weaponization. Local municipalities frequently deploy these regulations arbitrarily as a feature, not a bug, creating systemic friction to artificially create "jobs", drive up development costs, and suppress development.
The fundamental difficulty with regulations is that they've been written to apply the same to uncorrelated well-meaning individuals as to correlated constructively-malevolent corpos, based on some definition of "fairness" promulgated by the corpos. The things you've listed all seem pointless to you, because you're just trying to execute some small time project to make your own individual life better, not drive tens of millions of dollars of externality-producing economic activity through a loophole.
So the real answer is to reframe laws as applying to commercial operations, projects over a certain size, non-homeowners, etc. But of course the corpos cry foul over this, and would much rather you put your effort into working to repeal all regulations so they can operate without constraints.
Pragmatically, this is why I live in an area where if I wake up and decide I don't like a particular tree, it can be on the ground within a couple of hours. And most of that time is eating breakfast.
In the US, local inspectors are kind of like taxi medallion holders. They technically get you where you're going, but the standards are splintered from International Building Code and the whole process makes buildings far less safe than prefab components would be.
And at the same time I feel bad for inspectors being 5-10 years away from retirement, or worse, 5 years into the business. All inspectors in general, beyond some further construction engineering professionals and those making software for them.
1) Permitting is slow and, at some point, if you fuck it up big enough, the inspectors will probably take further steps, but otherwise not care much. You simply won't get a permit.
2) Who will care a lot are the builders and creditors. Building is expensive and if there's a couple of months of set back because of your fantasy drawings, that's not going to be in your favor.
Which is not so say AI planning is not going to happen. It will. But at that point it's probably mostly going to work.
This is more the other way around - there are already a couple of companies using AI to submit drawings that link back to previously permitted work, in order to reduce the arbitrary decisions by inspectors by saying "you personally approved this decision in another building on x y z". It's working well.
Lots of examples, sketches, and revisions, especially for the favorite model mentioned:
“Opus 5.5’s initial draft was the best, so I gave it the most feedback”
This shows that an experienced person is still needed:
“With the right guidance, Opus 5.5 is arguably better than the low end of the human market.”
Furthermore, the same problems arise as in software development:
“And, even after my redlines, the details contain some subtle errors that wouldn’t be made by a skilled architect with experience in this type of construction.”
And later:
“more than 15,000 lines of Python code”
as well as
“it took Opus 5.5 twelve hours to draft the house set, and the compute cost a few hundred dollars.”
The exact costs would have been interesting to know, since it’s safe to assume that prices will continue to rise (and must), which makes the comparison with a human interesting again—one who, in the end, will be needed once more.
I’ve been using various models to create 3d representations of various famous global landmarks when I’ve got some spare allowance … if you’re interested to see how they compare visit : https://civitas.reneo.app/compare.html#h=11&scope=models&tag...
If you want any, feel free to grab em by the source (MIT I think ?) and if you have better ones - please add !!
I'm a PE who has been tinkering with LLMs for a few years now. Opus 5.5 does seem to be a game changer for drafting. I don't generate with python tho, I have it drive AutoCAD headlessly and use our firm's drafting standards/templates. Its a junior-level EIT at first, but I give it redlines via Bluebeam MCP and it updates the relevant CAD SOP for all future sessions.
I am not an expert, but I’ve been tinkering architectural blueprints since Opus 5.5 released. From what I’ve learned, the vanilla results are a bit dated w.r.t CAD standards. Seems the only public version of the US NCS is v3 while the latest is v7. I ended up having to buy a licensed copy of v7 and develop my own MCP server to explore the standard. The result was a fairly significant rewrite of the application I wrote to comply with the latest requirements (see http://app.factoryplanner.net/try)
I am too. Doesn't that concern you for the future? I've been meaning to connect it to Revit but haven't had the time.
Edit: I'm not sure how fast any of this will be adopted with how slowly things move in construction. I'm surprised people are still using AutoCAD and making a profit.
Why would it concern a PE. 99.9% of their job is already stuff someone else could do and would rather not pay them to do but the government forces them to anyway. AI doesn't change that except perhaps remove a bit of the plausible deniability.
Now, I know, I know, I'm about to be hit with some bad faith retort about skyscrapers and bridges and other things engineered on a fine margin falling down. Your average multi tenant commercial building, your average parking lot, your average retaining wall ain't that. And that's the stuff that's the bulk of what PE firms do.
Because there are thousand of PEs across the country and eventually if a job requires 100x less time to produce, PEs will just be rubber stamping stuff after a thorough final review. My firm has like a dozen electrical PEs and we all spend the vast majority of our time in coordination meetings, reviewing small details, and working on drawings. If all I have to do is spend a few hours checking something at the end, they don't need all of us for that anymore.
[flagged]
Given how good the leading models got at 3D model creation over the past few months - Blender, Three.js, etc - it's not surprising to see that their 2D CAD-style abilities have improved to this extent as well.
(Amusingly the pelican-on-a-bicycle test doesn't capture this very effectively, despite testing 2D drawing with SVG - because pelicans can't write bicycles, so the results are inherently unrealistic. Having the model create an architectural diagram from a concept sketch is a much more useful 2D test.)
I worked almost 15 years in commercial concrete. My day to day tools were Revit and Bluebeam. As a little fun project I built a replica of them and they work surprisingly well. I was able to detail some of my previous jobs in an hour which took me weeks before. It is truly crazy the time we are living in.
Pretty big claim, attention grabbing for sure.
Might be nice to scope out where exactly it is good and where it is lacking ; residential vs commercial vs industrial , A vs S vs MEP, style of architecture? We need better benchmarking here to substantiate.
Agreed. Also the consistency between the different drawings would be nice to have an understanding of. This is where a higher level 3d domain model (such as IFC), do validation, interference checking etc., and generating drawings from those makes more sense (in my mind; but I'm biased). We did a quick demo once [0] but it ran out-of-date quickly due to the hardcoded model ids.
[0] https://ai-chat.ifcopenshell.org/
“Permit level” is doing a lot of lifting. Standard vary greatly by locality and project. Plus, permitting is a much lower threshold than actual inspection and sign off (which is where the issues will be caught).
A build permit simply says that they reasonably believe you can build the thing such that it will pass inspections. It doesn’t mean that every single detail will be finalized or correct.
> "Permit level” is doing a lot of lifting.
Have you considered the possibility that they meant exactly what they stated? That the output is good enough to reasonably believe it can be built and pass inspection,but it may still include flaws in details?
That was actually the intent of my point. Though, I clearly made it poorly.
AI can draw that plan for you if you give it the prevailing wind and the summer and winter sun data. But it can't take a soil sample from the land for you, and it can't know what kind of soil it is or what that soil will or won't tolerate. From a distance it can offer a nice picture, but without data it can't reach reality.
Washington State is trying out a 'Permit-Ready Plans' program in a few cities, were pre-approved plans are offered at no cost. According to their initial study, building plans account for 5-20% of total construction costs. Software that might be able to shave 10% off new construction costs is a big deal.
Any idea if these "permit ready" plans cost money for people to get vs being free?
If they cost money then it sounds like there's a potential incentive there for people to create a marketplace for them.
I looked into one program and they are offered free of cost. Currently, they mostly consisted of ADUs and dense housing, which seems to align with they city's housing goals.
I don't know how much time is spent 'drafting' at architecture practices vs modelling.
The drawings can be resolved from a data-enriched 3D model, which is normally where the hard hours go.
It would be interesting to add Deepseek 4.1 flash 1017 into this lineup. Deepseek's vision and 3D modeling capabilities are very impressive.
A big reason LLMs can write code productively is cause software dev tends to be pretty resilient to failure. And at the pessimistic end, lots of companies can make money off buggy products, so it’s easy to hand wave the failures away. Does that hold true for architecture?
That being said, I’m certain they’ll be extremely useful in other areas such as review, auditing, etc - if those even matter for architects.
I feel bad for all the inspectors who are going to have to do inspections on vibe-architected buildings in the next 12-18 months.
I don't. A lot of what they do is "make work" that slows down building homes and spaces with dilbert-esque nit picks.
I've reviewed dozens of sets of city and state AHJ review comments for housing. They sometimes take a full year and give less attention to the plans than even the worst contractor. Some aren't that bad, but many just have boilerplate review comments that are completely irrelevant to my drawings.
Civil engineering rules are mostly written in blood and costly and dangerous accidents. At least in my country. If somewhere else inspectors don't provide actual value, they should either get regulated or good riddance and AI might actually be better. I feel old, sigh...
That's not what we're talking about. By all means, follow the structural engineer's specification. But if you've ever submitted plans to a city to build a home, there is endless nonsense full of red tape: Drainage plan, soil analysis, tree survey, topographical survey, environmental site assessment, wind analysis, noise assessment, and more.
Sometimes these things make sense, but a lot of the time the work was already recently completed 3 years ago and yet they'll make you do it again. Last month I paid $3,900.00 to get 3 tiny trees and one medium sized oak surveyed here in the Bay Area. It's really about politics and control- the city wants to make a bunch of fake jobs and collect very high fees along the way.
Most of that stuff is important. That is, unless you like sewage pouring into the aquifers.
Nearly everything you listed is designed to prevent your decisions, or changes to your property that haven't been identified yet, from harming you or people around you.
Removing a requirement might just leave someone else paying a much higher price, potentially with their property, their clean water, or even their life.
To be clear, the utility of these processes can be good - but it also gets perverted into weaponization. Local municipalities frequently deploy these regulations arbitrarily as a feature, not a bug, creating systemic friction to artificially create "jobs", drive up development costs, and suppress development.
The fundamental difficulty with regulations is that they've been written to apply the same to uncorrelated well-meaning individuals as to correlated constructively-malevolent corpos, based on some definition of "fairness" promulgated by the corpos. The things you've listed all seem pointless to you, because you're just trying to execute some small time project to make your own individual life better, not drive tens of millions of dollars of externality-producing economic activity through a loophole.
So the real answer is to reframe laws as applying to commercial operations, projects over a certain size, non-homeowners, etc. But of course the corpos cry foul over this, and would much rather you put your effort into working to repeal all regulations so they can operate without constraints.
Pragmatically, this is why I live in an area where if I wake up and decide I don't like a particular tree, it can be on the ground within a couple of hours. And most of that time is eating breakfast.
In the US, local inspectors are kind of like taxi medallion holders. They technically get you where you're going, but the standards are splintered from International Building Code and the whole process makes buildings far less safe than prefab components would be.
Me too.
And at the same time I feel bad for inspectors being 5-10 years away from retirement, or worse, 5 years into the business. All inspectors in general, beyond some further construction engineering professionals and those making software for them.
I doubt that's going to happen.
1) Permitting is slow and, at some point, if you fuck it up big enough, the inspectors will probably take further steps, but otherwise not care much. You simply won't get a permit.
2) Who will care a lot are the builders and creditors. Building is expensive and if there's a couple of months of set back because of your fantasy drawings, that's not going to be in your favor.
Which is not so say AI planning is not going to happen. It will. But at that point it's probably mostly going to work.
This is more the other way around - there are already a couple of companies using AI to submit drawings that link back to previously permitted work, in order to reduce the arbitrary decisions by inspectors by saying "you personally approved this decision in another building on x y z". It's working well.
Ah yes, that I can totally see happening.
A single article by this person on Substack.
Lots of examples, sketches, and revisions, especially for the favorite model mentioned:
“Opus 5.5’s initial draft was the best, so I gave it the most feedback”
This shows that an experienced person is still needed:
“With the right guidance, Opus 5.5 is arguably better than the low end of the human market.”
Furthermore, the same problems arise as in software development:
“And, even after my redlines, the details contain some subtle errors that wouldn’t be made by a skilled architect with experience in this type of construction.”
And later:
“more than 15,000 lines of Python code”
as well as
“it took Opus 5.5 twelve hours to draft the house set, and the compute cost a few hundred dollars.”
The exact costs would have been interesting to know, since it’s safe to assume that prices will continue to rise (and must), which makes the comparison with a human interesting again—one who, in the end, will be needed once more.
It's more likely that token costs will keep falling over time.
> since it’s safe to assume that prices will continue to rise (and must)
Huh??