6 3d Shapes

[Hebe Zuelke]

Whencandidates ask what keeps me up at night, my answer is the fear that we lose the high-alpha execution of a startup and slow to the pace of a big company. The goal is that the Nth employee is able to be just as effective and impactful as the 10th. There are a number of reasons why that isn't true at most companies, and I wanted to explain one of them below.

Scopes and optimism significantly impact how quickly shit gets done, due to a combination of how humans work and bugs with common incentive structures. Scopes, in an engineering context, are time estimates for how long things will take to finish. They are produced for planning purposes, so that we can plan ahead for how long things take and figure out how much stuff we'll get done.

The scope influences how long things takes. When we say things will take a long time, they will take a long time. When we say things will take a short amount of time, they will take less time.

A version of this that we all have seen is the student completing assignments right before the deadline. Almost everyone I know completes homework right before the deadline, even when they really enjoy the homework. The existence of the deadline warps their distribution to finish it right before the deadline most of the time (sometimes right after).

You already see the problem. While the median in Fig 2 is probably slightly lower than the median in Fig 1, the mean is a lot higher in Fig 2. At a company, we mostly care about the mean since we're constantly pushing new things out. So simply introducing the scope has made us move slower.

One assumption here was also that engineers are good at estimating the distribution of how long something will take. Unfortunately, that's actually pretty hard, and due to the incentive structure (it's far worse to under-scope than over-scope in most product teams), most engineers end up being pretty pessimistic in their scopes. And, with a pessimistic distribution over outcomes, stuff only looks worse.

I sent this memo to the Scale AI team back in 2019. At the time, I viscerally remember the magical feeling that we were going at warp speed, and I didn\u2019t want that to go away. This memo was my manifesto to continuing warp speed. I hope it\u2019s helpful to anybody who wants to make a big dent on the world.

To begin justifying my claim, see below a histogram of about 9 million marathon finishes. Zoomed out, it looks like an approximate Gaussian distribution, which is roughly to be expected. But, when you look more closely, you'll see there are noticeable large spikes in the distribution at \u201Cround numbers\u201D, at 3:00, 3:30, 4:00, 4:30, etc. These spikes aren't by accident; they're produced because humans naturally try very hard to hit goals, and we become very good at adapting to these 0-1 loss functions.

Let's see how this affects the startup. Let's first consider how long something takes in a \u201Chigh urgency\u201D environment\u2014one example is during site downtime. Every moment the site is down is really bad, so the incentive is to finish it as fast as possible. Another high-urgency environment is an early-stage startup. In most scenarios, the company dies and fails, so every moment that new stuff isn't accomplished is really bad. There aren't really \u201Cdeadlines\u201D in both environments\u2014the deadline is now.

What happens when we set a \u201Crealistic\u201D scope? Let's assume we know this distribution exactly, and we set a scope that is the median of that distribution. The existence of the scope means that we're now planning for that amount of time to be taken, so the incentive structure actually changes. Versus the \u201Chigh urgency\u201D environment, now there is no visceral utility if we take less than the scope. The scope is the goal, and ostensibly we've planned-in that much time.

There is little utility to completing it far before the scope, but there is still disutility to completing it after the scope (we don't want to seem like bad planners!), so we warp the distribution to \u201Cplay limbo\u201D with the scope.

The mean to this distribution is still probably higher than Figure 1 (the \u201Chigh urgency\u201D environment), but not by too much more. In 10% of outcomes, we might go a bit slowly, but after that we still are \u201Cunder the gun\u201D and try to get it done. So, now with an optimistic scope, we're actually moving quite a bit faster. Put another way, optimism shapes reality.

To end, this directly ties into two of our credos at Scale AI: \u201CUp the tempo\u201D and \u201CAmbition shapes reality\u201D. Doing things as fast as possible, without regard for the scope, is the only antidote against the Limbo Effect. Optimism is another less good antidote, and if you are deeply optimistic, you can use it to consistently limbo under your goals, and over time, warp reality.

This effect might seem small in isolation, but it\u2019s death by 1000 papercuts. A dearth of optimism will slowly kill any product, team, or mission. Execution will slow to a halt, and even the most minor tasks takes weeks to do. Our optimism and resolve have immense influence in what we accomplish, both at a micro task-by-task level, and when summed up, what we can do over a lifetime.

I need to do this so I can make a pixel layer to add some texture without the various shapes showing lines/edges. I';ve succewssfully done this with the rest of the image as you can see in the attached photo. But those were all single shapes.

I made this alien's body in 4 different shapes but I want them completely merged. I thought I could group them and add new pixel layer... but no matter where I place the pixel layer... it's just not right.

To follow these steps, you need to open or create a report. To do that, on the Report tab, click Recent to open an existing report, or click More Reports to select a report from the list in the Reports dialog box or create a new report.

To format and align your text, click the Home tab, and then choose options from the Font, Paragraph, or Alignment groups, depending on the program you're using and what kind of formatting you want to apply. (Text formatting options are limited in Project.)

Quick Styles let you apply a style to your shape with one click. You'll find the styles in the Quick Style gallery. When you hover over a Quick Style thumbnail, you can see how the style affects your shape.

After adding a shape to a document, you can modify it by adding connector lines, changing the style, adding a mirror image, changing the color, and applying a variety of other effects, such as shadows, glow, and 3-D.

To format and align your text, click the Home tab, and then choose the available formatting options. For example, you can change font color, styles, sizes, change alignment or indentation, change the text orientation, to name a few.

You can add shapes, such as boxes, circles, and arrows, to your workbooks and presentations. (Word for the web doesn't support shapes.) To add a shape, select Insert on the ribbon, select Shapes, and then choose a shape.

When a shape is selected in your document, a Shape tab appears on the toolbar ribbon. It has buttons for things you can do with the shape such as adding a fill color or outline, or choosing a predefined shape style.

At the bottom of the gallery there are also options for the thickness (Weight) of the outline, and whether the outline is solid, dotted, or a dashed line. Point at Weight or Dashes to see a pop-out list of options.

Thanks guys! I figured it out, kind of. I was group the objects lines and faces before trying to cut out shapes. Do you know if there is a way to cut out of shapes from objects where the various components have already been grouped?

To cut out some geometry (you need to clean up after the intersect operation) you can enter the group or component, to be in its editing context. Then select all its geometry or the part that you suspect will intersect > 'Intersect faces with model > clean up > back out of the grouped environment again.

3a8082e126