#iterative — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #iterative, aggregated by home.social.
-
"I genuinely don't believe you can build complex systems any other way..."
-
I wonder how much battery, inverter, grid forming etc one could put in a standard 2TE container sized box, build one a week, and plug together somewhere convenient.
One might not add much inertia to a Grid, but as the year goes by you'd be steadier and steadier. -
Yeeaaa... it's a different approach.
(Don't read this as a defence of Musk, he's a turd, but SpaceX has competent technical people below their chimpanzee-on-a-string PR person)
NASA's traditional approach was to basically achieve perfection of design and manufacturing before trying to launch anything. Look at every possible failure mode of every component, down to the tiniest screw or wire or bit of plastic. Keep redesigning parts until you eliminate all failure modes that you don't have triply-redundant backups for. Test the living snot out of everything on the ground, in the lab. Have massive technical and safety reviews to ensure nothing was missed, anywhere.
It worked about as well as anything could, but it was extremely slow, bureaucratic, and above all incredibly expensive. Tons of rework when issues were found meant having to go back 3 steps to change something, and then redo the massive amount of work that had been done since then to make sure no new failure modes were possible, etc.
SpaceX is doing things differently - #iterative design. You design, build, #integrate, and #test-to-failure as often as possible to learn where the weak spots are -- you then rapidly iterate when you find the problems. "Rapid Unscheduled Disassembly" is an expected part of the process - it's how you learn the limits of what you've built, where the problems are.
Neither one is "the right way". They both work.
-
Yeeaaa... it's a different approach.
(Don't read this as a defence of Musk, he's a turd, but SpaceX has competent technical people below their chimpanzee-on-a-string PR person)
NASA's traditional approach was to basically achieve perfection of design and manufacturing before trying to launch anything. Look at every possible failure mode of every component, down to the tiniest screw or wire or bit of plastic. Keep redesigning parts until you eliminate all failure modes that you don't have triply-redundant backups for. Test the living snot out of everything on the ground, in the lab. Have massive technical and safety reviews to ensure nothing was missed, anywhere.
It worked about as well as anything could, but it was extremely slow, bureaucratic, and above all incredibly expensive. Tons of rework when issues were found meant having to go back 3 steps to change something, and then redo the massive amount of work that had been done since then to make sure no new failure modes were possible, etc.
SpaceX is doing things differently - #iterative design. You design, build, #integrate, and #test-to-failure as often as possible to learn where the weak spots are -- you then rapidly iterate when you find the problems. "Rapid Unscheduled Disassembly" is an expected part of the process - it's how you learn the limits of what you've built, where the problems are.
Neither one is "the right way". They both work.
-
Yeeaaa... it's a different approach.
(Don't read this as a defence of Musk, he's a turd, but SpaceX has competent technical people below their chimpanzee-on-a-string PR person)
NASA's traditional approach was to basically achieve perfection of design and manufacturing before trying to launch anything. Look at every possible failure mode of every component, down to the tiniest screw or wire or bit of plastic. Keep redesigning parts until you eliminate all failure modes that you don't have triply-redundant backups for. Test the living snot out of everything on the ground, in the lab. Have massive technical and safety reviews to ensure nothing was missed, anywhere.
It worked about as well as anything could, but it was extremely slow, bureaucratic, and above all incredibly expensive. Tons of rework when issues were found meant having to go back 3 steps to change something, and then redo the massive amount of work that had been done since then to make sure no new failure modes were possible, etc.
SpaceX is doing things differently - #iterative design. You design, build, #integrate, and #test-to-failure as often as possible to learn where the weak spots are -- you then rapidly iterate when you find the problems. "Rapid Unscheduled Disassembly" is an expected part of the process - it's how you learn the limits of what you've built, where the problems are.
Neither one is "the right way". They both work.
-
Yeeaaa... it's a different approach.
(Don't read this as a defence of Musk, he's a turd, but SpaceX has competent technical people below their chimpanzee-on-a-string PR person)
NASA's traditional approach was to basically achieve perfection of design and manufacturing before trying to launch anything. Look at every possible failure mode of every component, down to the tiniest screw or wire or bit of plastic. Keep redesigning parts until you eliminate all failure modes that you don't have triply-redundant backups for. Test the living snot out of everything on the ground, in the lab. Have massive technical and safety reviews to ensure nothing was missed, anywhere.
It worked about as well as anything could, but it was extremely slow, bureaucratic, and above all incredibly expensive. Tons of rework when issues were found meant having to go back 3 steps to change something, and then redo the massive amount of work that had been done since then to make sure no new failure modes were possible, etc.
SpaceX is doing things differently - #iterative design. You design, build, #integrate, and #test-to-failure as often as possible to learn where the weak spots are -- you then rapidly iterate when you find the problems. "Rapid Unscheduled Disassembly" is an expected part of the process - it's how you learn the limits of what you've built, where the problems are.
Neither one is "the right way". They both work.
-
Yeeaaa... it's a different approach.
(Don't read this as a defence of Musk, he's a turd, but SpaceX has competent technical people below their chimpanzee-on-a-string PR person)
NASA's traditional approach was to basically achieve perfection of design and manufacturing before trying to launch anything. Look at every possible failure mode of every component, down to the tiniest screw or wire or bit of plastic. Keep redesigning parts until you eliminate all failure modes that you don't have triply-redundant backups for. Test the living snot out of everything on the ground, in the lab. Have massive technical and safety reviews to ensure nothing was missed, anywhere.
It worked about as well as anything could, but it was extremely slow, bureaucratic, and above all incredibly expensive. Tons of rework when issues were found meant having to go back 3 steps to change something, and then redo the massive amount of work that had been done since then to make sure no new failure modes were possible, etc.
SpaceX is doing things differently - #iterative design. You design, build, #integrate, and #test-to-failure as often as possible to learn where the weak spots are -- you then rapidly iterate when you find the problems. "Rapid Unscheduled Disassembly" is an expected part of the process - it's how you learn the limits of what you've built, where the problems are.
Neither one is "the right way". They both work.
-
Musk isn’t an engineer and doesn’t understand iterative design, and now SpaceX and NASA are facing a sunk cost fallacy.
You never achieve iterative design with a full-scale prototype. It is incredibly wasteful and can lead you down several problematic and dead-end solutions. I used to engineer high-speed boats — another weight- and safety-sensitive engineering field. We would always conduct scale model tests of every aspect of design, iteratively changing it as we went so that when we did build the full-scale version, we were solving the problems of scale, not design and scale simultaneously.
https://www.planetearthandbeyond.co/p/starship-was-doomed-from-the-beginning
-
'Statistical Inference of Constrained Stochastic Optimization via Sketched Sequential Quadratic Programming', by Sen Na, Michael Mahoney.
http://jmlr.org/papers/v26/24-0530.html
#optimization #iterative #iteration -
All progress here is iterative, currently waiting patiently for some components to setup the DC power system in the house , primarily for an led lighting system 😁
-
Moments away from opening my Professional Scrum with User Experience course. (PSU)
It's one of my favourite courses to teach. The material is important for everyone who attempts to blend #UX and #Iterative&Incremental #Software #Development.
-
Some dancing starfish for today's #inkyDays drawing. With my recovery from surgery, I can't really dance yet, but I'm dancing on the inside.
Here's an in-progress shot.
And that's a wrap for this month's daily drawings. In-progress and complete drawings for this month are now all posted for my supporters on patreon and kofi in this month's #inkyDays posts:
https://www.patreon.com/posts/101521238
https://ko-fi.com/Post/InkyDays-April-2024-N4N0WEJT0#ink #drawing #art #starfish #seastars #MastoArt #iterative #GenerativeArt #wip
-
I see many software teams using user stories, making estimates, predicting how many of these stories fit into the next Sprint, and when a particular set of stories may be done. But what about getting feedback and reacting to it? Thinking that the first version of a feature will be great is ignorant. Get feedback, improve and iterate! Yes, now, estimates are useless. But your product will be useful to your users.
#PseudoAgile #Iterative #SoftwareDevelopment -
'Online Stochastic Gradient Descent with Arbitrary Initialization Solves Non-smooth, Non-convex Phase Retrieval', by Yan Shuo Tan, Roman Vershynin.
http://jmlr.org/papers/v24/20-902.html
#stochastic #iterative #optimization -
'Generalization Bounds for Noisy Iterative Algorithms Using Properties of Additive Noise Channels', by Hao Wang, Rui Gao, Flavio P. Calmon.
http://jmlr.org/papers/v24/21-1396.html
#generalization #stochastic #iterative -
Has anyone used the #PersonalSoftwareProcess (a.k.a. #PSP)?
How did you learn it? I am trying to learn by myself, but the learning process highly recommends taking classes with a coach/trainer. Is it really worth it?
I have also found that the exercises are aimed towards a #waterfall approach, so have you used it with a highly #iterative / #agile approach like #tdd?
As a reference, I'm following the #psp for the 21st century (https://softwareexcellencealliance.org/psp-for-the-21st-century/) but there are also materials for the whole thing:
* PSP for Engineers - https://softwareexcellencealliance.org/psp-for-engineers-version-4-course/
* PSP Fundamentals - https://softwareexcellencealliance.org/psp-fundamentals-and-advanced/(It's also difficult to understand which one to select. There are overlaps in both versions)
I read the book from W. S. Humphrey but I found it somewhat difficult to follow after the first few chapters.
Call to all #developers #coders #programmers #devs
#SoftwareEngineering #SoftwareDevelopment #software #Development #personal #process #CMMI
-
If you rely solely on "big bang" A/B #testing without #iterative components, you will never find out what actually brought regressions or improvements.
Even if you succeed, you never learn.
#ux #software #softwareengineering #marketing, #fundraising #manufacturing