5 ms·

apps to optimize the space in their office layout for maximum efficiency That sound like some useless bullshit if I've ever heard it.

ygjb 7y ago
So does most software until you try doing stuff the hard way.
ch4s3 7y ago
Really? You think optimizing office layout to maximize rental income is ”useless bullshit”?
- stickfigure 7y ago
  In fairness, it might actually be. They have a lot of properties, yes, but is it so many that that can't just send a particularly skilled human around to design them? Binpacking is np-hard but humans still do a very good job of playing Tetris. It is hard to imagine getting a positive return on automating something they (in the long run) will do a few times a year.
  - ch4s3 7y ago
    > It is hard to imagine getting a positive return on automating something they (in the long run) will do a few times a year. That's totally fair. I think they actually do some other types of more mundane optimization that are geared towards managing the properties. I think a number of the engineers also work on the learning platform for their education services.
  - Someone 7y ago
    If I understand their business correctly, way more than a few times a year. I think they have problems of the kind ”there’s a team of 5 Monday to Wednesday, one of 7 Tuesday that also have a meeting room, one of 4 Tuesday to Friday, one of 3 for the next three weeks, one of 2 that have to sit side-by-side, …. Figure out who to place where, minimizing the times we have to tell customers to move desk, maximizing the ‘cohesion’ of desks assigned to each group. Oh, and do this knowing that the problem may change at any time” And all of that, hopefully, in a building with a high occupancy rate, all the time.
kasey_junk 7y ago
This sort of software is hugely valuable in things like manufacturing. I don’t know enough about commercial real estate to know if it can drive the same efficiency there but it’s certainly not on its face clearly useless. In fact it sounds way more useful than most software projects I see.
CydeWeys 7y ago
It's literally the bin packing problem, which is NP-complete. Plus you have another axis entirely on top of the already NP-hard problem: You want to minimize the number of times that you move any given group of people over time as you re-configure the office space to fit the maximum number of groups into it. This is an algorithmically hard problem, and I'm not surprised that there might be significant benefits to investing in solving it very well.
- oh_sigh 7y ago
  Can you give me an order of magnitude estimate of how many milliseconds you believe solving the 2d bin packing problem with added constraints would take on a commodity PC when n=the max number of clients in any wework building? The problem is interesting, but it isn't one you need to set up an entire engineering team to solve.
  - CydeWeys 7y ago
    Actually, yes, I think you would need an entire engineering team to write, maintain, scale, and run a system that solves this problem. For example, my employer has a whole team working on internal tools for booking/managing/re-allocating conference rooms. It's almost exactly the same problem, it's non-trivial, it's complex, and it takes an entire team.
mlevental 7y ago
lol here's a hot tip for you: anything with "optimize" in the description is not useless bullshit. like the person said: it's literally an np-complete problem.
dwoozle 7y ago
That’s a very weird take. That’s probably one of the highest value problems to solve in the entire economy.

ygjb 7y ago

So does most software until you try doing stuff the hard way.

ch4s3 7y ago

Really? You think optimizing office layout to maximize rental income is ”useless bullshit”?

stickfigure 7y ago
In fairness, it might actually be. They have a lot of properties, yes, but is it so many that that can't just send a particularly skilled human around to design them? Binpacking is np-hard but humans still do a very good job of playing Tetris. It is hard to imagine getting a positive return on automating something they (in the long run) will do a few times a year.
- ch4s3 7y ago
  > It is hard to imagine getting a positive return on automating something they (in the long run) will do a few times a year. That's totally fair. I think they actually do some other types of more mundane optimization that are geared towards managing the properties. I think a number of the engineers also work on the learning platform for their education services.
- Someone 7y ago
  If I understand their business correctly, way more than a few times a year. I think they have problems of the kind ”there’s a team of 5 Monday to Wednesday, one of 7 Tuesday that also have a meeting room, one of 4 Tuesday to Friday, one of 3 for the next three weeks, one of 2 that have to sit side-by-side, …. Figure out who to place where, minimizing the times we have to tell customers to move desk, maximizing the ‘cohesion’ of desks assigned to each group. Oh, and do this knowing that the problem may change at any time” And all of that, hopefully, in a building with a high occupancy rate, all the time.

kasey_junk 7y ago

This sort of software is hugely valuable in things like manufacturing. I don’t know enough about commercial real estate to know if it can drive the same efficiency there but it’s certainly not on its face clearly useless. In fact it sounds way more useful than most software projects I see.

CydeWeys 7y ago

It's literally the bin packing problem, which is NP-complete. Plus you have another axis entirely on top of the already NP-hard problem: You want to minimize the number of times that you move any given group of people over time as you re-configure the office space to fit the maximum number of groups into it. This is an algorithmically hard problem, and I'm not surprised that there might be significant benefits to investing in solving it very well.

oh_sigh 7y ago
Can you give me an order of magnitude estimate of how many milliseconds you believe solving the 2d bin packing problem with added constraints would take on a commodity PC when n=the max number of clients in any wework building? The problem is interesting, but it isn't one you need to set up an entire engineering team to solve.
- CydeWeys 7y ago
  Actually, yes, I think you would need an entire engineering team to write, maintain, scale, and run a system that solves this problem. For example, my employer has a whole team working on internal tools for booking/managing/re-allocating conference rooms. It's almost exactly the same problem, it's non-trivial, it's complex, and it takes an entire team.

mlevental 7y ago

lol here's a hot tip for you: anything with "optimize" in the description is not useless bullshit. like the person said: it's literally an np-complete problem.

dwoozle 7y ago

That’s a very weird take. That’s probably one of the highest value problems to solve in the entire economy.