Lawyers is probably most of this. Police departments want to outsource their blame when something goes wrong (as do major corporations, mind you), so they're paying someone to handwave mistakes away.
99% of corporate/enterprise product delivery isn't about the product, its about contractual blame assignment.
Q&A. Compliance, e.g. correctly storing the data, while allowing audited access. Ongoing, around the clock support. Etc
Not that it doesn't make sense to look into government overspending in IT, it happens all the time. But this angle of approaching it is ineffective at best.
OK, this story gets posted on HN every now and then, and this question always appears there. Like, really always. People always recommend reading "The Mythical Man-Month", also this link is always posted somewhere in the comments: https://danluu.com/sounds-easy/
He didn't actually replicate an $86M project. The Victoria police signed a contract with Motorola for $17.3M in October of 2018 with Motorola part of which includes their ANPR system. It is a 5 year contract that will place cameras in 220 cars. The cameras will also record high quality video and audio. Among other things the contract includes 5 years of managed services regarding the operating of the ANPR and video/audio management system. A proper revised title for the article would be "I replicated an unknown fraction of a $17.3M project in 57 lines of code"
Combining off the shelf technologies for a novel / desired result with little effort isn't garbage. For example, podcasts are practically just media and image files listed in some XML. Also, replicating something existing in a vacuum with no restrictions is easier than prototyping in a legacy beurcratic environment.
He also used a compiler, but real programmers never do that. Maybe he used an IDE like Visual Studio that does a lot of real-time checking to make sure the code is correct before the compiler runs. Don't even get me start on the fact that he used off an off-the-shelf von neumann machine. Basically, the ghost of Von Neumann did allllllllllll the actual work and this guy is taking all the credit.
If you aren't writing your own Turing machine code in a language you developed, proved that language is Turing complete on the system you built, by hand(no help needed right?), that simulates the Turing structure on a state machine, you are not a programmer. Then, of course, wrote the OS(what's the name of your OS? I trying to become a real programmer and needs some pointers on scheduling). Don't forget to come up with your own machine learning model that doesn't use any of Shannon's llllllllame theories, doesn't rely on SVMs, neutral networks, or those weird upside trees(yuck) that those total losers with no programming skill what-so-ever, use.
Oh wait, you're still not a real programmer then; I almost forgot that the CT thesis is TOTAL unproven BS. No real programmer would stand for that. They are so logical and perfect in their understanding. So, if you can spare a minute from all the real programming you are doing, you should probably come up with your own model of computation, just so no one calls you out for not being a real programmer. Which, of course, you are. Wolfram did it and _everybody_ loved his "original" work. Of course, you'll then prove that this model can solve the halting problem in O(1). Wait, wait, wait, while you are at that, the PH hierarchy is also total unproven BS(look at me being so forgetful, I'm just not a real programmer, yet). Can you even believe that all those idiot nonprogrammers ever proved was P < EXPTIME, duuuuuuuuuuuh. You'll need to come up with your own classification system that isn't so broken. You don't even need to do much(what is NP even? And who decide to call it Merlin-Arthur? Is that a joke? Real programmer never joke about code.) because all your real code will run in polynomial time and always be deterministic...
What a Charlatan, amirite?
Programmers/Data Scientists stand on the shoulders of giants so large it is a serious undertaking to fully realize just how high up one truly is.
Actually an interesting project and in the authors follow up article he explains his motivation vs standing for the “click bait” knee jerk:
“Most readers saw it for what it was: a proof of concept to spark discussion about the use of open source technology, government spending, and one man’s desire to build cool stuff from his couch.”
> Since then, the reactions have been overwhelming. My article received over 100,000 hits in the first day, and at last glance sits somewhere around 450,000. I’ve been invited to speak on local radio talk shows and at a conference in California. I think someone may have misread Victoria, AU as Victoria, BC.
> I have met for coffee with various local developers and big name firms alike
But anyone who saw clickbait titles for what they are (putting the rest of the controversial claims aside) were just pedants:
> Teslas vehicles are already brimming with cameras and sensors with the ability to receive OTA updates — imagine turning these into a fleet of virtual good samaritans.
A few things that bug me (apart from filming open spaces and auto process those information is:
1) accuracy confidence of 90% is high enough for the author.
2) the recognizee plates are all send to a webservice. If grepped by IP you could "track" peoples cars/pinpoint the time to a location which is too much of an invasion of privacy for me. An offline database of stolen vehicles license plates would be better
3) they really put 86M$ in a system for this m-( makes me a bit sad
Really cool PoC. Though there is a lot of work that has to be done to get this production-ready, I really can't imagine it being more than $1-5m. I'd be really interested to see how the $86m breaks down.
It's government so that budget is paying for n levels of subcontractors subcontracting subcontractors, unions, oversight comittees to make sure the proper unions were involved, layers of beurocrats who do some things but nobody is really sure,regulators to make sure the beurocrats were following governmental standards, and regulators to to regulate the beurocrat regulators.
86m on a government project breaks down to about 1-5m in development budget.
“I replicated an $86M project by combining several open source libraries built on the backs of other developers whose names will go unnoticed using 57 lines of glue code.”
Using open source libraries is one thing, writing gratuitous articles about how you did something in X lines of code is another.
IMO, any article about how you did something in X lines of code is worthless unless you wrote every last line of code yourself. It only serves to impress people who know nothing about software. Lines of code you wrote is by no means a useful metric for evaluating complexity given the amount of library usage there is out there.
I could write articles about how I built $100M+ applications in a couple lines of code, but I have more integrity than that.
I don't think it's there to impress anyone about the author's skills, it's there to show how much open source can do. It's amazing that OP could mostly reproduce a huge software project without much effort.
I'm fairly convinced that the project cost a lot more than it should have. But the claim in the title is also a bit disingenuous. 57 lines of his own code is not "57 lines of code", and the results may not be at the same level.
Later edit: He does make the same point... in a another post where he admits to the clickbaity title, the overwhelming reaction (as many such titles trigger), and claims people pointing out the lack of support for the controversial core claims are pedants. [0]
Experts might be quick to point out that there is more to the system than merely being able to identify license plates; for example, it has to be able to operate on a national scale in a distributed way.
In spite of this, I still I think that the author's point holds - $86 million is a ridiculous sum for such a system.
Government contracts are corruption through ignorance (or is it pretend-ignorance?). Software does not cost that much to produce these days even at scale.
If you've ever run a large IT program before, you'd know that software development is probably less than half the cost and probably less than half of that will be on the core problem. There's a ton of governance involved in rolling out a system like this plus all the training, maintenance, operations, etc that all has to be done with full indemnity against failures. I could write a system to save text and images in a weekend, but that wouldn't make it Facebook.
These ridiculous prices are common in enterprise, not just the public sector. You just rarely hear about it when a bank spends 100 million on something that should have been build for 10 in the perfect world. One of the reasons you hear about this in the public sector is actually because we have systems to combat corruption. We don’t just hand out contracts to who we want, to get a public contract you go through a process called procurement, which typically requires a lot of upfront work both for the project but also legally and you have no idea if your bid is gonna to win.
It’s a lot more complicated of course, and one of the reasons is the massive bureaucracy. It’s both good and bad, on one hand it protects us from a public sector that buys software from sketchy people who go “I can do that 100 million project for 10”, exactly like I just did, and then completely fail to deliver. That’s not something you can do in enterprise, that whole move fast and break things simply doesn’t work when your results need to hold up in elections every four years.
Of course this exclusion of small agile companies who can’t afford to play into the public procurement system, means the companies who win the contracts are capable of milking us. Because the truth is, that we don’t get they alternative you think about. No one actually submits a 10 million bid for that 100 million system. Instead we’ll get a 93 million and a 86 million bid, and then we pick the 86 million one.
Typically less than 20% or the money goes to actual development though. The vast majority goes into whatever project management, governance, process modelling, lean, implementation strategies, training and all that other bullshit software companies sell you. And then 5 years later they fail to deliver and it ends up costing 10 million extra.
We really need a new model for this, and that’s where you and I probably agree, but it’s just not going to happen when not a single elected official in a country knows how to code.
All this bureaucracy does not seem to have stopped corporations like Oracle and IBM from legally stealing taxpayer dollars.
If anything, the bureaucracy is creating corruption. The money is going straight into the pockets of billionaires and thereby exacerbating the centralization of wealth, creating monopolies, reducing consumer options, and replacing good useful jobs with useless office jobs.
AFAIK, anyone who works in these kinds of projects uses open source libraries as they are the prevalent ones. So, I am a little perplexed with the emphasis on 'using open-source libraries'.
37 comments
[ 8.6 ms ] story [ 711 ms ] thread99% of corporate/enterprise product delivery isn't about the product, its about contractual blame assignment.
Not that it doesn't make sense to look into government overspending in IT, it happens all the time. But this angle of approaching it is ineffective at best.
But still, $86M is obscenely yuuuge.
Identify the problem, masquerade a randomly generated solution as being part of your perceptive abilities
KA-CHING! Thats a sound effect for getting rich from someone else paying you with other people’s money
my time this time!
If you aren't writing your own Turing machine code in a language you developed, proved that language is Turing complete on the system you built, by hand(no help needed right?), that simulates the Turing structure on a state machine, you are not a programmer. Then, of course, wrote the OS(what's the name of your OS? I trying to become a real programmer and needs some pointers on scheduling). Don't forget to come up with your own machine learning model that doesn't use any of Shannon's llllllllame theories, doesn't rely on SVMs, neutral networks, or those weird upside trees(yuck) that those total losers with no programming skill what-so-ever, use.
Oh wait, you're still not a real programmer then; I almost forgot that the CT thesis is TOTAL unproven BS. No real programmer would stand for that. They are so logical and perfect in their understanding. So, if you can spare a minute from all the real programming you are doing, you should probably come up with your own model of computation, just so no one calls you out for not being a real programmer. Which, of course, you are. Wolfram did it and _everybody_ loved his "original" work. Of course, you'll then prove that this model can solve the halting problem in O(1). Wait, wait, wait, while you are at that, the PH hierarchy is also total unproven BS(look at me being so forgetful, I'm just not a real programmer, yet). Can you even believe that all those idiot nonprogrammers ever proved was P < EXPTIME, duuuuuuuuuuuh. You'll need to come up with your own classification system that isn't so broken. You don't even need to do much(what is NP even? And who decide to call it Merlin-Arthur? Is that a joke? Real programmer never joke about code.) because all your real code will run in polynomial time and always be deterministic...
What a Charlatan, amirite?
Programmers/Data Scientists stand on the shoulders of giants so large it is a serious undertaking to fully realize just how high up one truly is.
“Most readers saw it for what it was: a proof of concept to spark discussion about the use of open source technology, government spending, and one man’s desire to build cool stuff from his couch.”
https://www.freecodecamp.org/news/remember-that-86-million-l...
> I have met for coffee with various local developers and big name firms alike
But anyone who saw clickbait titles for what they are (putting the rest of the controversial claims aside) were just pedants:
> Pedants have pointed out [...]
%s/good samaritans/rat finks/g
https://jonathanjanz.files.wordpress.com/2014/07/sutherland....
1) accuracy confidence of 90% is high enough for the author.
2) the recognizee plates are all send to a webservice. If grepped by IP you could "track" peoples cars/pinpoint the time to a location which is too much of an invasion of privacy for me. An offline database of stolen vehicles license plates would be better
3) they really put 86M$ in a system for this m-( makes me a bit sad
86m on a government project breaks down to about 1-5m in development budget.
IMO, any article about how you did something in X lines of code is worthless unless you wrote every last line of code yourself. It only serves to impress people who know nothing about software. Lines of code you wrote is by no means a useful metric for evaluating complexity given the amount of library usage there is out there.
I could write articles about how I built $100M+ applications in a couple lines of code, but I have more integrity than that.
https://i.imgflip.com/36py2c.jpg
I'm fairly convinced that the project cost a lot more than it should have. But the claim in the title is also a bit disingenuous. 57 lines of his own code is not "57 lines of code", and the results may not be at the same level.
Later edit: He does make the same point... in a another post where he admits to the clickbaity title, the overwhelming reaction (as many such titles trigger), and claims people pointing out the lack of support for the controversial core claims are pedants. [0]
[0] https://www.freecodecamp.org/news/remember-that-86-million-l...
In spite of this, I still I think that the author's point holds - $86 million is a ridiculous sum for such a system.
Government contracts are corruption through ignorance (or is it pretend-ignorance?). Software does not cost that much to produce these days even at scale.
It’s a lot more complicated of course, and one of the reasons is the massive bureaucracy. It’s both good and bad, on one hand it protects us from a public sector that buys software from sketchy people who go “I can do that 100 million project for 10”, exactly like I just did, and then completely fail to deliver. That’s not something you can do in enterprise, that whole move fast and break things simply doesn’t work when your results need to hold up in elections every four years.
Of course this exclusion of small agile companies who can’t afford to play into the public procurement system, means the companies who win the contracts are capable of milking us. Because the truth is, that we don’t get they alternative you think about. No one actually submits a 10 million bid for that 100 million system. Instead we’ll get a 93 million and a 86 million bid, and then we pick the 86 million one.
Typically less than 20% or the money goes to actual development though. The vast majority goes into whatever project management, governance, process modelling, lean, implementation strategies, training and all that other bullshit software companies sell you. And then 5 years later they fail to deliver and it ends up costing 10 million extra.
We really need a new model for this, and that’s where you and I probably agree, but it’s just not going to happen when not a single elected official in a country knows how to code.
If anything, the bureaucracy is creating corruption. The money is going straight into the pockets of billionaires and thereby exacerbating the centralization of wealth, creating monopolies, reducing consumer options, and replacing good useful jobs with useless office jobs.