Anerli wrote: “When the interface does eventually change, that's where it becomes non-deterministic again by necessity, since the planner will need to generatively update the test and continue building the new cache from there.”
But what determines that the UI has changed for a specific URL? Your software independent of the planner LLM or do you require the visual LLM to make a determination of change?
You should also stop saying 100% open source when test plan generation and execution depend on non-open source AI components. It just doesn’t make sense.
The small VLM (Moondream) decides when interface changes / its actions no longer line up.
We say 100% open source because all of our code (test runner and AI agents) is completely open source. It’s also completely possible to run an entire OSS stack because you can configure with an open source planner LLM, and Moondream is open source. You could run it all locally even if you have solid hardware.
I have never heard anyone think this way: “The main mistake people usually make is thinking Newton or Einstein were just scaled-up good students, that a genius comes to life when you linearly extrapolate a top-10% student.
The reason such people are widely lauded as geniuses is precisely because people can’t envision smart students producing paradigm-shifting work as they did.
Yes, people may be talking about AI performance as genius-level but any comparison to these minds is just for marketing purposes.
we kinda think too much of them though. each is also a product of their surroundings, and had contemporaries who could or did come to the same revelations.
Good note. Your examples suggest thinking of 'property' as a sort of discontinuous indicator function on the subsets. I'm thinking about the interdependence between the function values or across subsets, regardless of continuity, in the context of universal computation. How to localize or define intelligence?
Take the example of IQ, as a platonic ideal for measuring intelligence, vs all possible groups you could make with those people. Hard to define intelligence https://news.ycombinator.com/item?id=39977664
You should check out the HyWiki part of the Emacs Hyperbole hypertext package available from melpa or elpa-devel. It utilizes Org mode for notes and WikiWords to automatically interlink them as you type. The WikiWords work in any kind of text buffer as well, even in programming file comments. It even lets you link to many other kinds of Emacs entities like bookmarks, Org IDs and headings, etc. One command to build a web-based wiki. With a little practice, it might amaze you.
The same thing that has happened forever. People practice and gain skills on their own time before being welcomed into the professional community. Skilled new college graduate programmers have often put in a decade or more of learning before they graduate because they start young. They may have taught themselves many programming languages, studied many open source libraries, learned about profiling and efficiency tradeoffs, etc. Companies with serious engineering needs simply don’t want and haven’t wanted those who lack this natural desire to build and iterate. The only change is that the bar continues to rise a bit and all people are now competing at one level or another with machines.
It's worth reading right through. I particularly liked this bit:
"Up until recently (say, Google exists but you don't yet need to be transgender to work there), computers ran what were known as programs. These so-called programs were made of machine code, consisting of microscopic numbers that tell your CPU how to arrange particles of electricity into pornography. Ask your parents. The last program ever written was the V8 engine in 2008, after which programming was over and we all set about writing Javascript engines in Javascript for the next Javascript years.
Machine code was outlawed: in order to display pop-up advertisements, everyone agreed that it was best if we blindly ran whatever code was sent to us by strangers on the internet. We all got hacked for a while, but as a result, we now have a long list of extremely specific things that code from strangers shouldn't be allowed to do. We've only had to amend the list a few thousand times in the past; and—thanks to our collective willful ignorance of statistics, history, computer science, economics, crime, psychology, and of how lists even work—we're pretty sure that the list is complete this time."
I met Mr. Engelbart once as a child when my mother worked at SRI in the 90s/early 2000s. At the time I had seen the mouse prototype while wandering the halls and on some open house day I was introduced to him. He listened to me, a young kid, talk excitedly about technology for what must've been awhile and was very encouraging and kind. While I was excited about computers at the time I had no idea the significance of his ideas until I was an adult.
Looking back on it, I am awed by the kindness he showed to some random kid.
This is a trait of many great people. I guess they realise at some level that inspiring another generation helps carry the work forward than being dismissive and focusing on one's own work. This is also why I admire Andrew Ng.
There are so many systems thinkers from that time that basically understood how the world works and how it could or should work that basically no one today knows of or considers. It's a shame.
This demo us one of the most mindblowing things I've seen and especially so if you consider its date of arrival. I truly think the audience didn't really understand it.
My point was more that these people wrote about things that predicted where we were headed but no one listened to them then, and no one listens to them now.
Why can’t you keep the calling interfaces of functions as close to the Py libraries as possible, simplifying the transition for everyone? Will that really destroy the performance increase? Common calling interfaces make everything much simpler. Even in this simple example, the calls differ significantly.
We've tried our best to match python as well as we can, or falling back to matlab-style if Python doesn't have it. Many of our unit tests are verified against python, so the conversion is typically very easy. The one thing that python has that makes this much easier is keyword arguments. We've tried to use overloads to mimic this as best as we could.
That being said, in the example on the home page there are notable differences:
1) we have the run() method. The reason is that the expression before the run is lazily evaluated for performance and does not execute anything. Having the run() method allows you to run the same line of code on either a CPU or GPU by changing the argument to run()
2) in MatX memory allocation is explicit. Python does it as-needed, but this causes a performance penalty with allocations and deallocations that are not under your control. Specifically in the FFT example, numPy will allocate an ndarray prior to calling it, but on the same line. In MatX the allocation is (typically) done before the operation so you can control the performance of the hot path of code.
If you have any specific suggestions, we would love to hear it
Probably because many people struggle to read with any reasonable level of comprehension. People reading this site are not representative of the public at large, as you know.
GNU Hyperbole for rapid, implicit linking of all your programming artifacts and management support for handling technical complexity. One of those you can get for free.
Actually, using Org mode and the Hyperbole package in Emacs solves all the problems listed in easy to use ways except you get much more editing dunctionality rather than just the features requested. Don’t reinvent the wheel; just learn something that runs everywhere and solves your problems. You start Emacs once and then rapidly edit files from the command-line by calling emacsclient, not restarting Emacs.
But what determines that the UI has changed for a specific URL? Your software independent of the planner LLM or do you require the visual LLM to make a determination of change?
You should also stop saying 100% open source when test plan generation and execution depend on non-open source AI components. It just doesn’t make sense.