jam

I'm really enjoying https://handy.computer/

Runs local and has several models that should support Polish, though I can not personally validate this.

As followup to this... I tried half a dozen air mice, and they were all awful. The TV-remote style mice all send pre-defined keyboard presses that can not be modified or intercepted.

That's why I went with the single-hand game controllers instead. I'm looking forward to the new Steam Frame VR controllers, because they will probably be the best option once they're out.

A few months ago I put together a library that lets you use a Nintendo Switch JoyCon 1 (the JoyCon 2s can't connect to bluetooth) or a Playstation VR controller to a Mac in order to control the mouse, use a radial menu, etc: https://github.com/jturnshek/JamCon

I map one of the keys to voice input push-to-talk for handy: https://handy.computer/

This involved reverse engineering and writing new drivers for both (with no existing reference for the PSVR). It's all in that library, so if that sounds interesting to you, you can probably get Claude/Codex/whatever to convert it to a Windows app pretty quickly.

I haven't been maintaining this tool actively, but maybe it's useful to someone.

Are there good texture generators that export in KTX2? I use iliad.ai mainly, but for PNGs.

Would love a "parachute" that works like this...

We managed to successfully pass lighting regulations in Pittsburgh last year [0] which will result in about 40k city lights being replaced with much more dark-sky-friendly hardware.

My mother [1], an astronomer at CMU, has been very involved with this effort, and is tracking it with "before" and "after" photography courtesy of the ISS crew.

It is one area of environmental regulation that seems to be pretty easy to make progress on because it's not a partisan issue and the "right way" ends up being cheaper for cities as well as more pleasant.

[0] https://skyandtelescope.org/astronomy-news/pittsburgh-goes-d...

[1] https://dianeturnshek.com/dark-skies/

Formant | Multiple Roles | Remote | Full-time | https://formant.io

Formant supports the most advanced robotics and autonomy companies in the world by providing open cloud infrastructure, remote monitoring and teleoperation tools, and collaborative support and operations workflows.

We are hiring for roles across the business including in design, engineering, sales, and customer success.

If you are interested in the future of autonomy (especially as a generalist engineer), this is the way to get the best vantage point in the entire world. We work with everything from fleets of quadrupeds inspecting nuclear power plants to gigantic automated farm vehicles to delivery drones, and have customers in every domain imaginable.

Apply for open positions here: https://jobs.lever.co/formant

Agreed - only Apple standalone trackpads are really worth using as far as I can tell.

If you happen to like the ThinkPad TrackPoint mouse nub, these keyboards work great for that type of setup: https://www.lenovo.com/us/en/accessories-and-monitors/keyboa...

Luckily I've managed to associate it with more of a "deep focus" state, but a lot of people seem to feel like they'd get tired and fall asleep if they used a setup like this.

I actually think the biggest downside to the multi-station dynamic setup is that the ideal monitor configuration for that is a single curved ultrawide, and there are essentially no high resolution options.

Yes, keyboard just above the knees. It's similar to the angle you'd get if you were slouching and your forearms were resting on the desk. Generally quite comfortable.

I think raising my hands up higher (90 degrees as you say) would actually be a lot worse. Never understood that aspect of all the more "real" laying desks.

I should note that I purposely don't use a mouse with this setup, because it would end up being quite far away and would need to be on a solid surface.

I've had a setup like this for years, but it's much simpler: monitor arm [1], ultrawide display [2], keyboard + trackpad lap tray [3], any couch. Been through many iterations and like this gear, but you can do it for cheap just as easily.

Mount the arm to something (like a desk) and swing it over the couch to lay down. In my case, I can actually pivot it between the couch and the chair on the other side of the desk.

Great for a few hours a day when you want to sink into focus mode. Super highly recommended it.

[1] https://www.humanscale.com/products/monitor-arms/m-81

[2] https://www.lg.com/us/business/desktop-monitors/lg-34BK95U-W

[3] https://www.hekseskudd.com/products/ambidextrous-slim-keyboa...

Good time to share Kane’s MachinePix project, which is made for this crowd: https://twitter.com/machinepix

A few they mention are production systems at fairly high volume. The floor scrubbers from Brain Corp and shelf scanners from Bossa Nova are running in hundreds of Walmart stores, and the delivery robots from Starship are running in dozens of cities.

The call center example is actually more relevant than you might imagine. The jobs often look very similar to remote tech support in practice. In fact, some companies running call centers have begun offering remote support to autonomous systems as an expansion of their market.

This type of labor falls pretty squarely in the "information work" category, and often (as in your Kiwi example) looks mostly like operators performing a high-level task like specifying waypoints on a map.

I'd be interested in how you'd propose to make that work when operators may be jumping in to a robot anywhere on the planet from one minute to the next. Require logging and reporting on when remote labor was used to support the tax jurisdiction you're operating in? It seems that would also require reporting to every locality you operate robots in individually, which is currently not necessary.

Would you propose to use the same method for other remote services such as call center support?

It's important to keep in mind that teleoperation is always paired with an (initially quite immature) autonomy system. Just as blakesterz called out in the sibling comment, the strategy here is to start with high volume of teleoperation and then scale up autonomy over time, essentially increasing the productive output of one teleoperator.

We see companies all across the board, from 1:1 (or even multiple people always watching at once) to over 1:100 in operator / robot ratio.

In practical terms, this usually takes the form of operators being notified of situations that they have to resolve, and switching between systems frequently.

Once you're above a certain level of productivity, the cost of the teleoperator is negligible... so it doesn't really matter where you locate the labor as long as you have a reasonable path to reaching a high ratio using autonomy.

Fortunately, getting robots out in the world lets you gather the data needed to improve your autonomy system, and so it's worth it to get out there early as long as it's safe.

Our company [1] works with a large number of autonomous robotics companies of varying scales, tasks, and morphologies. Teleoperation is a regular part of the workflow for many.

It is an exceptionally good time to be building specialized hardware & autonomy systems for a task and then providing that system as a service to customers. This article went into a few examples, but by our count there are ~800-1000 companies currently operating with this model.

Happy to answer any questions about how people doing this deal with the myriad of issues (both business and technological). It’s a fairly new thing, but extremely well suited to the future.

[1] https://formant.io

Wouldn't this require the Ethereum community adopting a modification to allow paying for gas with ERC20 tokens, and thus purposefully making ether worthless?

If they were to do that, how would platform development be incentivized? Sounds pretty far fetched.

Pretty sure you managed to understate it.

I was sure from the title that this was a reinforcement learning agent set up to train itself to play Doom...

Really nice eye-tracking compilation of an experienced rider on track: https://youtu.be/4gNbuMQTRaU

Tires affect a vehicle's performance more than any other component. It's worth buying good ones. A shame more people don't put the thought and money in.

macbook air

home theater projector: http://goo.gl/S9siD

apple wireless keyboard and trackpad with tray: http://www.hekseskudd.com/

nice couch, good speakers

the iphone app "nomis" does this well: http://www.nomisapp.com/

I think desks in general stopped being ideal after the switch away from writing on paper. Nothing new has caught on and it's quite unfortunate.

I like to work with a keyboard on my lap, and there are so few good options for it that it's incredible. My requirements are:

- some access to the mouse pointer. 1% of my usage still requires it. - shallow keys (like the apple keyboard). This is just preference. - apple cmd and option keys. - no numpad. - wireless and usable at up to 10 feet or so.

Doesn't sound crazy right? The only decent quality keyboard that meets those requirements is the Logitech diNovo Edge... the design of which I find just cheezy enough that I'm not into it.

I'd easily pay $200 for the right keyboard, but it just doesn't exist. In the meantime, I'm stuck using one of these: http://www.hekseskudd.com/products/abidextrous-wooden-tray-f...

K&R, but technically more like 1TBS (as described on Wikipedia).

I'm not sure I agree with the author, but Apple doesn't really need to make easy-to-use desktop operating systems any more. That's what the iPad is for.

Try Vimium: http://vimium.github.com/

motorcycle tires