• 0 Posts
  • 15 Comments
Joined 6 months ago
cake
Cake day: June 14th, 2025

help-circle

  • If the government procurement person doesn’t really understand the deep technical requirements, they are likely to choose the bidder who also doesn’t really understand the deep technical requirements, and is the low bidder because they don’t realize what they are getting themselves into.

    By the time everyone realizes how much more is really required, they are already halfway through the project. The government could have saved money by choosing a more realistic higher bidder to start with. But once they have half a program from the low bidder, throwing that away and starting over doesn’t save any money. Better to just finish with the team that’s invested with the project.


  • I work at a large company that is critically dependent on VAX software written in the 1980s for almost every aspect of functioning. This was recognized as a problem. A replacement coding and testing team was established. It included a full-time team of contractors - a handful US based and I believe dozens located in India - along with a few full-time dedicated employees and maybe a dozen each of people brought part time out of retirement (the people with the 1980s knowledge!) and people with other main jobs who had to start dedicating significant time to support.

    It ran for two years, then two more years, then another year. Very much a case of “the more you know, the more you know you don’t know” in that the more functions were programmed and tested, the more edge cases and sub-function requirements were uncovered. This program has been upgraded in pieces by so many people for so many decades that no one realized how hugely complex it had become, and what an enormous undertaking it would be to replace it. But after five years - more than double the original two-year projection - it was coming together, more things being really finalized than new needs being uncovered.

    And then the software that the replacement program was being written with lost support. It was too old. Documents were written to try to give some future team a better chance of success, and everything was disbanded and shut down.

    Being peripherally involved in that really made me more sympathetic to fiasco large tech projects.






  • Sometimes regular stimulation is enough. I saw one case where a man took on an orphaned infant where there wasn’t even any animal milk available to hack together formula, and the starving infant attempting to get milk out of his nipple every hour for multiple days was enough to get the one breast to start making milk, and the infant lived thanks to it.

    The need for suckling to stimulate milk production is a catch-22 for women who don’t make enough milk. They have to supplement with formula to prevent “failure to thrive”, but the infant spending some of its sucking time on a bottle instead of a breast reduces their supply even more, so then they have to feed even more formula… There are devices that run a tube to the nipple so the infant can get formula from the tube while at the same time stimulating breast milk production, and they work, but look like a huge pain.





  • I started using in-shower lotion because I needed something (had a small wound that didn’t heal for more than a year), water-based lotions weren’t cutting it, and I couldn’t get the quantity right with oil-based (too little didn’t work, too much was uncomfortably greasy). Once I found it and it worked for me, I used the Olay in-shower lotion for many years, but it recently seems to have been discontinued, so now I’m using Nivea. The smell is too strong for my taste (pleasant, just too much of it), but at this point I am too hooked on the in-shower convenience to try anything else.




  • The current iteration of agentic AI technology used by Logitech is little more than a glorified note-taking bot capable of summarizing meetings and “generating” the occasional idea.

    Given that most humans hate note-taking and avoid it, but it has a lot of value as a meeting output, getting a machine to do it makes sense.

    I also heard a podcast where a consulting company couldn’t get their client contact to make any decisions because he wanted his CEO to review, but she had a busy schedule and was never available. The consultants trained an AI on this CEOs writings, and presented it to their client contact. The model was convincing enough the client felt comfortable making decisions. I thought that was interesting, and this article refers to something similar with models of stakeholders.