Many years ago, Alan Carter wrote an online book, "The Programmer's Stone", part of his "Reciprocality" project. In it he distinguished two types of people, "mappers" and "packers". Mappers are nthe natural programmers, who have a unified representation of the world and the problem to be solved in their minds. Once the map is complete and accurate, they can read the solution off at a glance and implement the solution so concisely that there is no room for bugs. Packers are the manager types, scarcely human in any real sense, P-zombies, who think that understanding is all about stacking pre-approved knowledge packets. They try to turn software development into a factory job, specifying endless procedures, and always fail.
AI has no world model, no map, and until it does, it not only can't replace the mappers, it will reliably enshitify every project it touches in ways that will take many times the effort to fix than was saved.
Thanks for that reference -- it captures a point that occurred to me reading this article. In the previous millennium I was for a while a programmer, and worked briefly on an operating system analogous to the OS project that Brooks drew his lessons from. It was the VME/B operating system for the ICL "New Range" (2900), and it was explicitly designed in this way -- hundreds, maybe thousands, of modules with interfaces precisrly described and kept consistent in a specialised database (relational, because relational databases were the answer to everything). The code was, intellectually speaking, secondary, and the people who wrote it were a lot lower down the pecking order than the people who had designed it. It was a bureaucratic nightmare. It did work, but it was not a bit happy. In fact Brooks refers to this mode of design: in his example, the issue is which side of the interface owns the memory, that being the crucial resource at the time, and how conflicts over that crucial resource were, or were not, resolved.
What you describe as mappers chimes with my experience of working with a different system, VAX/VMS. It was written by a core team of, I think, three or four people. The lesson I drew from that was that the optimum team size was probably about three, and the harder the task, the smarter the people you had to employ. When the world ran out of smart people was when you had to switch paradigms.
But the overall lesson is, I think, still true. What AI is doing is shifting from the smart team paradigm to the zombie horde paradigm.
The ascendancy of the zombie managers isn't some eternal or inevitable thing. My grandfather started working on mechanical analog computers (Mark I naval gun control) for Sperry Gyro in 1942, basically designed by one genius who doesn't even have a Wikipedia biography. No room for bugs in that design. (Sure, some idiot admiral got a ring-sight added, but he wasn't allowed to dictate the internal design.) That experience, along with being on other small, smart teams such as the Sidewinder guidance package showed him how to run the F-14 avionics development.
About four years later, 1946, his wife’s great uncle died at age 98; he had never retired as president of the sewing machine company he built; in the 1870s, he had invented the machine that makes the stitch that holds everyone's pants together, with 3-D cams for motion control at 2000 stitches per minute. Not like today's managers.
Well most of them — John Walker founded Autodesk as a programmers’ co-op, served as president for many years while continuing to work on Autocad programming, took the company to IPO without a dime of outside financing then eventually handed over the reins so he could get back to programming. He tells the story in the “Autodesk Files” at his fourmilab. ch site. (I interacted a bit with him on his “scanalyst” forum in the year up to his untimely death from a fall. Great fellow.)
Even more recently, Chris Doran, the top man in applying Clifford Algebras to physics (very clear papers, too) founded and ran Geomerics, which brought real-time lighting to games. After its acquisition by ARM, he has been doing much of the administration at Sidney Sussex College, Cambridge. Loss for physics, gain for not having zombies in charge.
It's crucial to not allow “packers” to be in charge of “mappers”, especially at the very top and in strategy-setting positions. Unfortunately, the educational bureaucracy and the other zombie bureaucracies are dedicated to turning children into zombies like themselves.
Alan Carter had some interesting ideas on the origin of the packer pathology (e.g. dopamine/boredom/ritual addiction and particularly “the ghost not”) and for how to convert packers into mappers. (His brain cooling by information compression idea seems quantitatively unlikely to me, though.)
“The tractor did not make the farmer obsolete,” he says, but the number of farmers now sits at a tiny fraction of what it once was. They still exist, but so do blacksmiths and farriers and fletchers. Just a handful, with second jobs, living in backwaters, practicing their craft for the love of the game, not because it is financially remunerated.
Looks like there will be increased need for specialization that is yet to be well understood. That’s what happened during previous times of industrialization.
Ahh, the old mythical man month. If 1 woman can have a baby in 9 months the 9 women can have a baby in one month.😅😅😅😅
Many years ago, Alan Carter wrote an online book, "The Programmer's Stone", part of his "Reciprocality" project. In it he distinguished two types of people, "mappers" and "packers". Mappers are nthe natural programmers, who have a unified representation of the world and the problem to be solved in their minds. Once the map is complete and accurate, they can read the solution off at a glance and implement the solution so concisely that there is no room for bugs. Packers are the manager types, scarcely human in any real sense, P-zombies, who think that understanding is all about stacking pre-approved knowledge packets. They try to turn software development into a factory job, specifying endless procedures, and always fail.
AI has no world model, no map, and until it does, it not only can't replace the mappers, it will reliably enshitify every project it touches in ways that will take many times the effort to fix than was saved.
Thanks for that reference -- it captures a point that occurred to me reading this article. In the previous millennium I was for a while a programmer, and worked briefly on an operating system analogous to the OS project that Brooks drew his lessons from. It was the VME/B operating system for the ICL "New Range" (2900), and it was explicitly designed in this way -- hundreds, maybe thousands, of modules with interfaces precisrly described and kept consistent in a specialised database (relational, because relational databases were the answer to everything). The code was, intellectually speaking, secondary, and the people who wrote it were a lot lower down the pecking order than the people who had designed it. It was a bureaucratic nightmare. It did work, but it was not a bit happy. In fact Brooks refers to this mode of design: in his example, the issue is which side of the interface owns the memory, that being the crucial resource at the time, and how conflicts over that crucial resource were, or were not, resolved.
What you describe as mappers chimes with my experience of working with a different system, VAX/VMS. It was written by a core team of, I think, three or four people. The lesson I drew from that was that the optimum team size was probably about three, and the harder the task, the smarter the people you had to employ. When the world ran out of smart people was when you had to switch paradigms.
But the overall lesson is, I think, still true. What AI is doing is shifting from the smart team paradigm to the zombie horde paradigm.
Thanks for sharing those recollections.
The ascendancy of the zombie managers isn't some eternal or inevitable thing. My grandfather started working on mechanical analog computers (Mark I naval gun control) for Sperry Gyro in 1942, basically designed by one genius who doesn't even have a Wikipedia biography. No room for bugs in that design. (Sure, some idiot admiral got a ring-sight added, but he wasn't allowed to dictate the internal design.) That experience, along with being on other small, smart teams such as the Sidewinder guidance package showed him how to run the F-14 avionics development.
About four years later, 1946, his wife’s great uncle died at age 98; he had never retired as president of the sewing machine company he built; in the 1870s, he had invented the machine that makes the stitch that holds everyone's pants together, with 3-D cams for motion control at 2000 stitches per minute. Not like today's managers.
Well most of them — John Walker founded Autodesk as a programmers’ co-op, served as president for many years while continuing to work on Autocad programming, took the company to IPO without a dime of outside financing then eventually handed over the reins so he could get back to programming. He tells the story in the “Autodesk Files” at his fourmilab. ch site. (I interacted a bit with him on his “scanalyst” forum in the year up to his untimely death from a fall. Great fellow.)
Even more recently, Chris Doran, the top man in applying Clifford Algebras to physics (very clear papers, too) founded and ran Geomerics, which brought real-time lighting to games. After its acquisition by ARM, he has been doing much of the administration at Sidney Sussex College, Cambridge. Loss for physics, gain for not having zombies in charge.
It's crucial to not allow “packers” to be in charge of “mappers”, especially at the very top and in strategy-setting positions. Unfortunately, the educational bureaucracy and the other zombie bureaucracies are dedicated to turning children into zombies like themselves.
Alan Carter had some interesting ideas on the origin of the packer pathology (e.g. dopamine/boredom/ritual addiction and particularly “the ghost not”) and for how to convert packers into mappers. (His brain cooling by information compression idea seems quantitatively unlikely to me, though.)
“The tractor did not make the farmer obsolete,” he says, but the number of farmers now sits at a tiny fraction of what it once was. They still exist, but so do blacksmiths and farriers and fletchers. Just a handful, with second jobs, living in backwaters, practicing their craft for the love of the game, not because it is financially remunerated.
Looks like there will be increased need for specialization that is yet to be well understood. That’s what happened during previous times of industrialization.
Did see a headline today that Ford has rehired 350 gray beard engineers because of an over reliance on AI that hadn’t gone so well.