Congratulations! I’ll be interested to see the next steps in alignment. Do you plan to start selling access, or collect more data to train bigger & better? What tasks or benchmarks are your biggest guide stars, or what was unexpectedly tricky—a few are hinted in the post.
It would be pretty interesting to see activation maps for the encoder on video, confidence building to see the compression derived from so much training.
we have an alignment blog post dropping soon! scaling up in the next couple of months, then hopefully opening up an API or licensing it.
Benchmarks are really fun—lots of secret ones. Our main thesis is that you should be using the same benchmarks to measure human ability to use a computer, as you would an AI model. Definitely a suite of continuous long term planning tasks (games) and things such as marking emails as spam etc.
definitely! we are looking into more interp + visualizations in general as we scale up.
I would appreciate it. It’s such a joy to use, to share, and (admitting I haven’t) nice to see & share the code & configuration to feed algolia, on a more serious scale of us than I’m likely to use in a hobby project or find/grok/share another public example with code + data.
It's good FUD. You re-iterate their talking points. (Also, no CTA, no takeaway, just "worry!")
As others have said, the data has to be publishable to be useful. We do have data export laws. The format is known to be ready to use interoperably, not some private schema--atop the PBC commitment, which will at least have moderate legal costs if not a guarantee. It has unequivocally set a new high bar.
They seem pretty locked in to doing what they committed to. The day may come when they turn. It may come first by friction, but the turn has to be pretty complete, because the data is pretty open. What's needed to view it, use it at all, is pretty close to what's needed to host it.
"The site whose value prop is sharing your posts and data with other apps may stop sharing your posts and data with other apps." Yeah, it's possible. It's also possible they just close.
Well, Twitter/X gets this wrong too. Pretty often jumps away from what you're viewing, especially on the nav-in to a thread or nav-out from a thread actions.
aiui "distrobox" is built to support these setups and experimentation, even more readily, including defaults to support:
> The created container will be tightly integrated with the host, allowing sharing of the HOME directory of the user, external storage, external USB devices and graphical apps (X11/Wayland), and audio.
> Why
* Provide a mutable environment on an immutable OS, like ChromeOS, Fedora Silverblue, OpenSUSE Aeon/Kalpa, or SteamOS3 ...
* Provide a locally privileged environment for sudoless setups (eg. company-provided laptops, security reasons, etc…)
* To mix and match a stable base system (eg. Debian Stable, Ubuntu LTS, RedHat) with a bleeding-edge environment for development or gaming (eg. Arch, OpenSUSE Tumbleweed, or Fedora with the latest Mesa)
* Leverage a high abundance of curated distro images for docker/podman to manage multiple environments.
> Aims
This project aims to bring any distro userland to any other distro supporting podman, docker, or lilipod. It has been written in POSIX shell to be as portable as possible and it does not have problems with dependencies and glibc version’s compatibility.
> It also aims to enter the container as fast as possible, every millisecond adds up if you use the container as your default environment for your terminal:
> Security implications
Isolation and sandboxing are not the main aims of the project, on the contrary it aims to tightly integrate the container with the host. The container will have complete access to your home, pen drive, and so on, so do not expect it to be highly sandboxed like a plain docker/podman container or a Flatpak.
distrobox create -n test
> Create a new distrobox with Systemd (acts similar to an LXC):
distrobox create --name test --init --image debian:latest --additional-packages "systemd libpam-systemd pipewire-audio-client-libraries"
distrobox enter test
Unfortunately it looks like sandbox mode [0] is not a goal, so it doesn't solve the main problem I have - running semi-trusted apps (e.g. Android Studio) and minimising their impact. Currently I just share X11 socket and run it in Docker.
Yeah, I'm bouncing for now on the localhost requirement. Or, on a related issue of not parsing my .ssh/config, a Match directive, and not wanting it to parse it yet. I grep'ed for an env var to override, but only USER and SSH_AUTH_SOCK are pulled in.
I did go get install ...nerdlog/cmd/nerdlog-tui@latest just fine.
Thanks for hacking in the open, and releasing early.
Hey mcint, fyi both of these issues are addressed: the localhost one is addressed for real, and a Match issue is worked around: while it's still not properly implemented, at least it doesn't prevent Nerdlog from starting now. Just in case you wanted to give it another try.
I would say that their thanks is sincere, and that they're applauding you for releasing a new tool to a public/critical audience while also taking feedback in very constructive manner.
Excellent descriptive analysis. Wrong, misleading title, perhaps "technically correct," but at best with a "backdoored" meaning.
It points out the need and use for build-manager tools that go a step beyond union file system layers, but track then enforce that e.g. tests cannot pollute build artifacts. Take a causal trace graph of files affecting files, in the build process, make that trace graph explicit, and then build a way to enforce that graph, or report on deviations from previous trace graphs.
In defense of the author: nobody reads your article if the name is boring (that is my experience at least), which it would've been if they titled it more accurately. That gives incentive to authors to use click-bait titles.
In defense of the bank robber: no clerk simply gives you money if you aren't threatening them (that is my experience at least), which it would have been if they acted like a respectable citizen. This gives people the incentive to become bank robbers.
It would be pretty interesting to see activation maps for the encoder on video, confidence building to see the compression derived from so much training.
reply