3 ms·

This is an inevitable aspect of the MVP fake-it-until-you-make-it culture that infects even established companies. They are telling themselves that if enough pe

This is an inevitable aspect of the MVP fake-it-until-you-make-it culture that infects even established companies. They are telling themselves that if enough people come to depend on it doing what they claim it does then they'll make it do that, but not before. I've encountered people making surprised Pikachu faces when called on problems with their systems while they roll out a new version that fixes it and claim incidents never occurred. Appearances are everything in tech today.

dbish 3y ago
This is pushed from the smallest startup upwards as a way to make sure you’re building something people want. There are pros and cons of course, but I can understand why companies do this and why software folks think they need to. It’s related to the shipping early, fast, and iterating which is the currently taught best practice by places that have a lot of experience getting new things off the ground like YC.
- fidotron 3y ago
  But there's a world of difference about lying about capability today and working with customers to deliver what they need. My favourite example is how ARM ended up in Nokia phones. They had a 32 bit RISC CPU core which met the performance requirements but Nokia turned them down due to the memory density of the instructions meaning it required more expensive storage than their existing solutions. On the way back from the meeting they came up with the idea of Thumb, worked out how to do it, did it, and Nokia bought it. It then took over the industry. That conversation in SV today would involve Nokia going "We love what you folks are doing" but just keeping the existing system without ever saying why, and nothing would improve.
  - to11mtm 3y ago
    Interesting to think about that. I remember the SuperH had 16 bit instructions (well, SH-5 was variable length,) I wonder whether that would have become 'our future' if ARM hadn't done Thumb back then.
    - fidotron 3y ago
      IIRC ARM did end up licensing the SuperH patents about that. I tend to think SuperH could have been much more of a contender had Hitachi been more willing to license it for integration by others. The one that mystified me was why anyone thought MIPS was viable in this space, other than the fact that it existed. Ultimately MIPS utterly failed to take advantage of their very big wins.
rchaud 3y ago
It's about to be multiplied several times over with AI startups promising the moon and the stars and handwaving away criticisms with anthropomorphized terms like "hallucinations".

dbish 3y ago

This is pushed from the smallest startup upwards as a way to make sure you’re building something people want. There are pros and cons of course, but I can understand why companies do this and why software folks think they need to. It’s related to the shipping early, fast, and iterating which is the currently taught best practice by places that have a lot of experience getting new things off the ground like YC.

fidotron 3y ago
But there's a world of difference about lying about capability today and working with customers to deliver what they need. My favourite example is how ARM ended up in Nokia phones. They had a 32 bit RISC CPU core which met the performance requirements but Nokia turned them down due to the memory density of the instructions meaning it required more expensive storage than their existing solutions. On the way back from the meeting they came up with the idea of Thumb, worked out how to do it, did it, and Nokia bought it. It then took over the industry. That conversation in SV today would involve Nokia going "We love what you folks are doing" but just keeping the existing system without ever saying why, and nothing would improve.
- to11mtm 3y ago
  Interesting to think about that. I remember the SuperH had 16 bit instructions (well, SH-5 was variable length,) I wonder whether that would have become 'our future' if ARM hadn't done Thumb back then.
  - fidotron 3y ago
    IIRC ARM did end up licensing the SuperH patents about that. I tend to think SuperH could have been much more of a contender had Hitachi been more willing to license it for integration by others. The one that mystified me was why anyone thought MIPS was viable in this space, other than the fact that it existed. Ultimately MIPS utterly failed to take advantage of their very big wins.

rchaud 3y ago

It's about to be multiplied several times over with AI startups promising the moon and the stars and handwaving away criticisms with anthropomorphized terms like "hallucinations".