Freesteel Blog

On the shape of Autodesk’s disinterest of innovation

Monday, August 3rd, 2015 at 11:40 am Written by: Julian

We will begin today’s rant (following last years post-quitting blog) by turning to Carl Bass’s favourite management consultancy firm, McKinsey&Company, and searching for “Autodesk” through the link:

http://www.mckinsey.com/search.aspx?q=autodesk

to get 17 hits — 10 of which are for exactly the same page.

Well done McKinsey. I hope your internal document management system is better than the one on your webpage, given that your only purpose is to write commissioned documents.

The McKinsey article begins:

How big companies can innovate
Who says innovation is only for start-ups? In these interviews, the heads of three large, established companies — Intuit, Idealab, and Autodesk — argue there’s no reason big players can’t develop the next big thing.

First up is Carl Bass. What’s he got to say?

The threat of somebody doing something is one of the biggest tools you have to motivate, encourage, scare people into taking risks they wouldn’t otherwise do. And most corporations are set up and, in some ways, structured and designed to maximize profit and minimize risk.

Yet what you need to do in order to become the disruptor, as opposed to the disrupted, is sometimes exactly the opposite. So, for example, this year we decided for the first time to build our own 3-D printer, which we are making with an open-source design.

Yep, that and the other 471 open source 3D printers listed in the RepRap Family Tree. I bet no one in the C-suite made themselves unpopular by bringing that up when he proposed this big idea.

Now let’s now hear from the slightly more earnest CEO of Intuit, Scott Cook:

If I had to point to one thing that’s made the biggest difference at Intuit — and there’s a package of things — it was to change how we make decisions, whenever possible, from decision by bureaucracy… decision by PowerPoint… persuasion… position… power… to decision by experiment.

Normally, companies put up a phalanx of barriers and hurdles and mountains to climb that may not seem hard for the boss or the CEO but are intensely hard, impossibly hard, for our young innovator to conquer. So our job as leaders is how do we get all those barriers out of the way… And a culture of experimentation can only work when it’s put in place by leaders. The innovators can’t do it.

I have a story about this.

There was a programmer called Julian Todd who had quite a long interest in building CAD/CAM systems in the browser. He even tried it back in 2006, as you can see from this brochure:

This experiment failed because the technology was not ready.

But by 2011, as any fool not using the Internet Explorer browser could see from the demos available, that fact had changed. [Click on the picture below!]

At the time Julian was working at ScraperWiki on a collaborative coding system in the browser with particular emphasis on webscraping.

This experiment didn’t work out commercially in the long term either.

However, technically competent people would find it easy observe that the tech was virtually identical to this new CAD-in-the-browser Onshape system, complete with simultaneous collaboration, version control and a business plan that required you to pay if you wanted to keep more than 3 scrapers secret.

The only difference between the two products was that ScraperWiki was based on Python at the server and Codemirror in the browser, while Onshape is based on Parasolid at the server and Webgl in the browser.

This is actually a similarity.

In October 2012, Julian was absorbed into the Autodesk borg and had a lot of positive feelings about it at the time, what with the massive opportunities and the resources that could be opened up within such a large organization with such a diverse set of industry-leading products. The CEO even appeared to be pretty well on the ball too.

Julian made his first connections within the company’s developer community to the ASM group in Cambridge, which is their proprietary solid modeling kernel forked from ACIS, that is essentially an equivalent component to Parasolid.

Just like Parasolid and all the other industrial solid modelling kernels, this component is a multi-million line mass of misdesigned quarter-century-old C and C++ that just about serves its purpose if you don’t notice that its architecture is constrained to an era when computers had but one processor and mere kilobytes of memory.

Through this team he contacted an employee in America who had experimented on his own unofficial “spare” time project to link the ASM under a SWIG interface into Python.

This was all he needed to build his idea quickly and cheaply.

Salvaging his old open source code from the Python-based Scraperwiki system — now that SWIG accessed the ASM into the Python environment — Julian was able to create a working CAD experiment in the browser by end of January 2013 in his spare time, after less than 4 months with the company.

Here are two screen shots of the demo, the second illustrating simultaneous editing. The only CAD feature available was boolean subtraction of a sphere by node selection, because this was just a quick experiment to prove the technology.

There is a recollection that Carl Bass breezed through the Cambridge office on his way to Legoland at around that time, and was reportedly not interested.

And that was that.

Strong intuition is a great tool. Sometimes. But in the event that the intuitive judgment is wrong there must be a forum where it can be factually challenged by persons with other points of view without being immediately steamrolled or, if they continue to stand up, condemned as impertinent.

I don’t think it is the CEO’s job to have the time and capacity to get to grips with any and all the technical details and development in the world as it happens.

But…

…if this man cannot somehow arrange for any person at any level of influence in the organization to have any curiosity about any challenge to any of the underlying assumptions and potential d-i-s-r-u-p-t-o-r-s to the business, then he hasn’t been listening to a anything he’s been saying about the future of the company.

Trust me: if anyone in the whole of the company was actually interested in this tech, I would have found them.

I pushed this demo to a few places, but excuses flowed easily. Interaction in the browser wouldn’t be responsive. Or precise enough to our special CAD tolerances. It probably couldn’t handle the massive levels of detail that you get in a realistic CAD model.

All these are, of course, testable hypotheses. No one can know the answers because, with so much development in the internet capacity and consumer hardware, the facts are always changing.

Hey, I said, you know your clunky Inventor CAD product doesn’t exactly have the most nippy user interface on the planet. I’m sure we could beat it in the responsiveness stakes with a browser-based program if its parameters of sluggishness were known and you didn’t move the goalposts.

Unfortunately, talk like that doesn’t go very far in Autodesk, because they think their software is so fecking awesome that any criticism of it doesn’t make sense.

Then Tinkercad was acquired by Autodesk in May 2013.

Hey, I thought no one was interested in CAD in the browser around here.

Shut up, Julian. You’re being very annoying.

As you’d expect, I’d been following the Tinkercad product for a while, pleased with the fact that they had coding in the browser, along with its own implementation of the CAD kernel, as explained in this google tech video:

minute 16: Kai Backman: [speaking about the solid modeling kernel] BRep [the model used by Parasolid, ACIS and Autodesk’s ASM] is industry standard. Everybody does it. The reason everybody does it is because it’s compact. Back in the day in the 1970s memory was a big concern. So you needed something which you could squeeze into as small space as possible.

“Volumetric” This is what we did. We take the whole space, subdivide it into cubes, and for every cube we store one bit of information: inside or outside. This is order n-cubed. But you can compress it so the actual growth is n-squared… Booleans are super straight-forward and parallizable…

[some discussion about moving the geometry between the client and server]

We don’t advertise this, but several people can edit one part at the same time. The reason we haven’t exposed it is we don’t know how to do the UI for it correctly yet.

There’s a lot more explanation of how they got the speed up to useability. And it’ll be the same work that was done at Onshape.

I shopped this tech video around a few people following the Tinkercad acquisition, saying: Hey, do you think that maybe there’s some ideas in this Tinkercad codebase that could perhaps inspire the folks over in the ASM team who are struggling with their single-core limited memory conventional solid modeling kernel?

No answer.

Then I went through the ridiculous rigmarole of obtaining access to the relevant source code repository of Tinkercad which took about a four weeks, zipped up the C-code files and emailed it to my colleagues in the Cambridge ASM group, with the message: Hey, do you want to get together and go through this code to find out if there’s any ideas we need to know?

And, lo, I got into a bit of trouble. My email was forcibly deleted from everyone’s account, and I realized that there was something fundamentally wrong with this software company. No matter how drunk you are, this is a no-brainer place to look for ideas.

Here is the joke:

A company like Autodesk expends all this fecking effort keeping its source code locked up and secret, massively inconveniencing its software developers and suppressing numerous lost opportunities for collaboration and consolidation — you cannot believe the time wasted by totally incompatible careless duplication of basic functions everywhere — and all for fear that someone in another company might see this code and acquire some amazing game-changing ideas from reading it.

And the same exact managers who appear to believe in this myth are involved year after year in corporate acquisitions that result in batches of previously confidential source code — sometimes from direct competitors — dropping straight into their Perforce repositories, and not one, not one damn developer or manager can be arsed to take the briefest cursory glance at it.

This is illogical!

They don’t even ask if the newly acquired developers would like to give technical presentations to the Autodesk developer community of the previously confidential tech in case someone wants to borrow, use, comment, and collaborate on the new tech. I dropped this idea into the “Suggestions” box after every acquisition.

That’s a good idea, Julian, they said. We’ll get round to it. And then they did nothing.

If I’d have concluded that the entire departments of corporate acquisition, HR and office of CTO were this bone-headed a few months earlier than I did, I’ve have been able to organize it myself by looking through the org chart and making a recorded Skype interview with a developer myself. We’d have talked programming languages, optimization, the shock of acquisition, software components, bug tracking, and the changes in what tech works or stopped working over the last decade. If the managers don’t understand what programmers need to know to get on, then we are wasting our time if we sit back and wait for them to deliver something useful.

And so, two years later, in 2015, and having spent a budget approximately half of what Carl Bass blew on acquiring his Socialcam porn site, and about equal to what he got paid personally for his valuable time during that interval, Belmont technology in Cambridge, Massachusetts launched Onshape.

Just like when Solidworks was launched, Autodesk refuses to see this as a competitor.

Over in my hackspace I have been enjoying using Onshape almost daily to print boxes for my electronic hang-glider circuits.

I have also helped a lot of visitors with their designs that they then 3D print later that day. They get their work done faster than it takes merely to download one Autodesk Fusion and install it. Fusion is Autodesk’s idea of an up-to-date product, even though is uses only 1990s technology.

Even if Fusion was the greatest thing ever, it still wouldn’t be useful here because most folks in the hackspace are on Linux, and this operating system is not supported.

The design fundamentals of Fusion are that you throw everything into it, including he kitchen sink, and people somehow work it out because they’ve just spent $5000 dollars on it and don’t have the option to back out.

I’m sorry, that’s the user model for Inventor and Autocad. Things are different now when you can’t rely on buyer’s remorse to get customers through the painful learning stage.

The design fundamentals of Onshape were explained by its founder Jon Hirschtick at the Develop3Dlive conference in the first 6 minutes of his talk:

Carl Bass was not impressed, and didn’t care what things they solved with their tech. I’m smarter than everyone in my company, and I’m smarter than them, he thought.

He channeled his inner Donald Trump and blogged:

In light of their upcoming launch, several people have asked me about Onshape. Given how hyperbolic and misleading their statements have been, it is time to set the record straight…

The first claim they make is that Onshape is new and revolutionary – the first to the cloud. It’s as if to say nothing happened in CAD while they weren’t building CAD tools…

[Two paragraphs of Autodesk vanity user-base numbers being much greater than a product that’s been on the market for less than a week omitted]

Is this the land that time forgot or were they just not paying attention?

Their second argument confuses the web browser with the cloud.

Let’s back up to the basics and explore what the cloud is good for when it comes to design and engineering. First and foremost, the cloud provides access to nearly infinite and scalable computing power at reasonable costs…

The second thing that the cloud is ideally suited to do is be the central coordination point for storage, data management and collaboration… For tools like Fusion 360 that have built-in collaboration and data management capabilities, A360 provides the underlying infrastructure to store, manage and version the data…

But for CAD applications today the browser means reduced functionality, less capacity for big models, slower performance (for anything that has to roundtrip to the servers) and most importantly, no offline access to your data.

This was spoken by a man whose company is totally organized around the crappy applications of Sharepoint and Microsoft Outlook email exchange.

I used those systems, and I couldn’t believe how bad they were. No wonder he feels that browser apps and the internet suck.

But at least they softened me up for when they launched A360, based on their $26million acquisition of Qontext, which was even worse. Everyone I spoke to agreed with me. Boy did we all try to use it and couldn’t. Those who had been in Autodesk longer than I had told me that shite like this happens all the time. Get used to it.

That was $26million spent on very bad software, and they would have been better starting from a free fork of WordPress.

Unfortunately, there’s no way they could choose to do this because everybody above a fourth level manager — and anyone connected to the legal department — believes the GPL is the second incarnation of the Ebola virus.

This seriously constrains the options and locks developer productivity to about the level it was back in 1993 when you couldn’t just go scavenging free code off the net as you do now in order to add stuff to your working product or check out how the most obvious ideas don’t work on implementation before you waste precious time falling for them.

I could not plumb the depths of ignorance within the Autodesk legal department. They wouldn’t return my calls — not even when I attempted to clarify some complete bollocks written into the employment contract which Mr Denmark begged me to sign one fateful night so that he could complete the deal. What was their problem? Do they think that if they responded to me in writing it would provide ammunition that I could later use in a legal dispute? Whose side are they on?

Meanwhile, software developers around the world are free to read the news on the Onshape blog about how their CFO has been testifying against the “corrosive non-compete clauses in software engineers contracts”.

Okay folks. Here’s a question. If you were a talented software engineer looking for a job, which company would you rather work for on the basis of the evidence?

Ultimately, in a big over-moneyed organization like Autodesk, the software developers count for nowt. The top executives look at their profit centres and see how the techies just cost money, smell bad, and are occasionally a bit rude to their superiors about their latest stupid ideas. (The only job of some of these managers is to come up with ideas and pontificate, so to say they’re no good is a serious insult.) Meanwhile, their tight team of lawyers and accountants over here will shuffle some bank accounts and papers through a system of shell companies in Luxembourg and create billions of dollars of pure profit by stealing from national tax bases. Who are you going to love? The shareholders — aka “The Street” as they like to call it — are ever hungry. You respect the team that delivers the dosh on time to throw down that money hole. The “investors” bleed you dry. All future earnings are already priced into the share price. The hill just gets steeper and nothing is ever good enough.

In the fullness of time, Onshape will eventually become vampire fodder like Autodesk, when its Angel Investors North Bridge Venture Partners, New Enterprise and Commonwealth Capital decide the time is ripe for an “Exit” and they drop this baby into hell.

Until then, everything is going great as this company enjoys a wonderful childhood where it can grow and develop to its fullest potential.

I remember at the discussion panel at the end of Develop3Dlive 2013 where Jon Hirschtick gave the game away about how he built Solidworks. There was nothing revolutionary about it, he said. All he did was take the old Parasolid CAD kernel in 1993 and compile it onto the new WindowsNT operating system (launched in 1992) which everyone was just getting into. He encouraged the developers to use modern (for the day) software design principles such as the new exciting Microsoft Graphical User Interface tools. They launched the product launched in 1995, and the rest is history.

Autodesk has been playing catch-up ever since, when they had to bodge together something from the ACIS kernel and launch Inventor in 1999.

What we have with Onshape is the same story offset by 20 years. The Autodesk executives knew this was going to happen, so by my calculations they’ll attempt to get their rival product to market in 2019.

Why is it going to take them four years to copy a product that was designed from scratch in two?

Because they will delay, and they will get it wrong because nobody in the company with commissioning responsibilities has the vaguest clue about the tech on which Onshape depends, and they’ll try and do it with Microsoft Windows servers, because they don’t understand the point of Linux (which is why none of their products support it). To them it’s just another brand of operating system — in the same way that a trained sheepdog is just a subspecies of wolf.

The basic technical intuition among the managerial class is obsolete. And they’re not aware of it. In response to the demand for new cloud products within the company, I circulated this seminal rant about how Amazon succeeded at it, and nobody understood a word. This ignorance cannot be remedied by a few coupons at Lynda.com. It’s gone too far. There’s a critical mass of stupidity which protects itself like a slag heap spilling over any barriers put up to contain it. And those who do know better are either climbing the greasy turd-pole or are instructed to shut the f*ck up so that this turd-pole climbing can be conducted in peace and tranquility.

Who knows what happens now? Autodesk’s top leaders, the ones who with the power to change the culture, have to a man been in the system for decades and are totally comfortable with it. Cannot see anything wrong. It’s going to take an outsider who has experienced the culture in another software firm to come in with enough power to make changes. You’ll know when that happens, when the bottom rung programmers are actually thanked and rewarded for constructively disrupting other teams and their out-of-date work practices, or for posting notes about how the website of your two-year old expensive acquisition is now full of porn.

I’ll just end this with this excerpt from Henry Ford’s memoirs about the revitalization of a railway:

The Detroit, Toledo & Ironton Railway was organized some twenty-odd years ago and has been reorganized every few years since then… All of the rolling stock was in extremely bad condition and a good part of it would not run at all. All of the buildings were dirty, unpainted, and generally run down. The roadbed was something more than a streak of rust and something less than a railway. The repair shops were over-manned and under-machined. Practically everything connected with operation was conducted with a maximum of waste. There was, however, an exceedingly ample executive and administration department, and of course a legal department. The legal department alone cost in one month nearly $18,000.

We took over the road in March, 1921. We began to apply industrial principles. There had been an executive office in Detroit. We closed that up and put the administration into the charge of one man and gave him half of the flat-topped desk out in the freight office. The legal department went with the executive offices. There is no reason for so much litigation in connection with railroading…

The railroads in general have broken down… Too many railroads are run, not from the offices of practical men, but from banking offices, and the principles of procedure, the whole outlook, are financial — not transportational, but financial. There has been a breakdown simply because more attention has been paid to railroads as factors in the stock market than as servants of the people. Outworn ideas have been retained, development has been practically stopped, and railroad men with vision have not been set free to grow.

Technical prediction: Autodesk’s clone of Onshape will be derived from Tinkercad, not started fresh by a team who doesn’t know the hardware. The fact is that toy/gaming software is far superior to engineering software because all your 3D graphics hardware has been made for it. The mind boggles at the amount of hours in management meetings it’s going to take to reach this conclusion.