Humans don't use command line lsp tools - why should agents use them?
I use lsp when it is integrated in the IDE - but I haven't seen yet any usable command line tools - even if theoretically it would be very useful to type 'lsp-python -replace-function-name foo bar'.
Here is a realisation from my recent work: it is not if the data is relational - because if you need you can always push the data into relations - just like you can push it into hierarchies if all you have is files and directories. It is about how much churn there is in the relations. You can model relations with just links or something and if there is not much churn you can keep it in git and it will work OK - the problem starts when your relations start churning (like when you have reviews that need to be updated on both review instruction change and document change) this is when you see yourself start building a relational db (badly).
I am still impressed by QubesOS track and I use it for my dedicated 'financials' laptop.
IMHO the thing that is holding back QubesOS is the lack of hardware acceleration for graphics - maybe now when dual monitor setups are getting popular this could be a workaround for the security considerations?
I dropped QubesOS once exactly for that reason in the past but now I’m running it again on a separate computer.
Even with all its drawbacks, there is something really nice about being able to run different applications over Tor, VPN or plain internet simultaneously, the ability to isolate non-safe binaries and being able to backup your VMs easily.
I wish a similar distro would be made based on KVM so that the standard kernel could be used. It would be great for compatibility.
> being able to run different applications over Tor, VPN or plain internet simultaneously, the ability to isolate non-safe binaries and being able to backup your VMs easily.
These are all possible using light containers. For example, on FreeBSD I will spin up a jail which runs wireguard, and then I'll bridge that to another a jail. That 2nd jail is running entirely off wireguard without any other way to access the network. Since it is a jail, it is isolated. And backing up is as simple as a zfs snapshot and zfs send. I assume the same is possible on Linux.
I would argue non-safe binaries are not safe in a container either. And even with full VMs, there have been an embarrassing number of escape exploits over the years.
This is sort of a solved problem with virtio native context. You can just run mesa driver in the virtual machine and proxy it's ioctl calls to host kernel gpu driver.
I'm pretty sure all the major GPU drivers in mesa now supports this feature.
They use Xens pci passthrough, you can setup a pci device to be assigned to a vm.
The vm runs it's own driver and has direct access to the device, you typically have to hide the device from dom0 for it to work flawlessly and hardware support is required. But I believe the kernels do not interact with each other in a direct way.
I used to have crashes on trying to restart a vm with a pci device attached probably related to resetting the device. But they seem to have resolved now days.
Hardware graphics acceleration can work, it just needs a bit of fiddling and opens a few attack vectors. I don't legitimately know if there has ever been a viable attack through a shared pci device though.
Problem is not that nobody wants to do it, it is that GPU stack is exactly the kind of enormous driver surface Qubes exists to keep away from dom0. Second monitor does not really change that, somebody still has to trust the driver.
In the age of LLMs it is important to have vocabulary that survives when taken out of context - German nouns fit that very well. In my own project I decided to use phrases of two words (often with a hyphen) for all technical terms (terms that have special meaning in my project).
And by the way I also started using Auftragstaktik as an anchor for a methodology of writing plans and prompts.
Wow, thank you for that, must've been quite a bit of work to put that together and I feel sorry for not digging through the archives. I hope this brought up a few new thoughts, perhaps, I know it's mostly ramblings and us-against-the-ones-on-top banter, but here and there, I see and was hoping for some insights. I'll go through the list in a bit, thank you!
Honestly I asked in google (I remembered the main sources) - then I iterated with Gemini to format it well. But my first post in this thread was flagged - I don't know what was the reason but it is possible that it was for using AI, as it also contained some more narrative text from Gemini.
Advertising has a positive function - currently we don't have better (market based) was to make it working other than via paid advertising. The first step is to find the new business models.
I know it is overused on tweeter - but all this really seems to be about status. Like Robin Hanson said - people read stuff because it seems prestigious.
IMHO training weights has peaked and now it is time for a training paradigm for prompts and code. We don't have the gradient descent here - but I think it can be more sample efficient because causal theories can be better than just correlations.
I am working on a unified theory in https://zby.github.io/commonplace/ - it is all agent edited so it might be hard to read, but hopefully we are catching most logical errors. Some day the llm prose will improve.
Is there any reason to think that training weights has peaked rather than is accelerating? It feels like now they are increasingly able to pick some low hanging fruit by using the models in order to improve themselves and test optimizations.
I think it is now or close to now - and not because there is no more to be gained - but rather because it is now clear that you can gain more with the right agent environment.
But benchmark saturation is also something to account for.
It's a pity that the Perl6 saga killed the new regex ideas - I wish the raku regexes were adopted elsewhere (like the Perl5 regexes were): https://docs.raku.org/language/regexes
I use lsp when it is integrated in the IDE - but I haven't seen yet any usable command line tools - even if theoretically it would be very useful to type 'lsp-python -replace-function-name foo bar'.
reply