72% revenue decline from a 54% price increase, but how did profitability change? Some of the features of the Deck are still differentiating like the OLED screen and touch pads, and maybe some of the price increase is meant to keep it profitable at a low volume.
I personally think a controller and a VR headset that isn't standalone would be fine products for Valve to sell. Maybe they can't justify working on Steam OS without the Deck, Steam Machine or Frame. If they can still sell low volume differentiated products, good for them.
I disagree with the first half quite a bit, COGS ultimately depends on the use case. If someone just wants something that a smaller model can do, running a local model on phone is going to have a negligible cost close to running any other piece of software. The alternatives to running a model also determines COGS, and even Jensen Huang has distinguished between the job and the work for the job that AI is capable of doing. Smaller models are always going to win in efficiency too.
I also heavily disagree with this no-marginal cost in software distribution view whenever I see it, bit rot is real, and someone is paying a marginal cost whenever they do an update. You have to re-distribute with changes whenever anything changes. These costs are just hidden because things are ad-supported or bundled in some way. These costs are also kept low because of standards and open source, but could become high anytime. Additional licensing also has costs.
That said, I couldn't agree more with the last paragraph, charging a high price for models would be better than denying access for any model that wants to stay relevant.
It's hard to sympathize with Anthropic for this or the export ban, the hype over model capabilities probably fuels both things (in some ways). Training data for me, but not for thee (at any scale) doesn't seem like a tenable position. If anything, Claude's constitutional outputs should be trained on more rather than less.
Out of all the possible permutations of display/sensors/processing for AR, I would like to see eyewear with just a Bluetooth-like display and camera/mic. Let me use the phone or watch for all processing. Bonus points for eye tracking and body position sensors. Make the camera tethered if you must and somehow integrate it with my clothes.
Off course, this would mean less lock-in for everybody and we can't have that.
If large models were primarily used for distillation to make smaller models better, that could be a more efficient use of resources. Smaller providers could choose to fund a larger model and benefit from it.
I think in this context it would mean not stopping movement of labor to an extreme or having tariffs, but taxing the unequal gains and having social programs that help everyone keep up in an economy.
I hope the EU and DMA succeed in that case, but would also like to point out operating system choices would also help accomplish the same objective of the DMA.
I am for real browser choice, and also for real operating system choice. You can download other photo apps and not use Siri AI if you don't want to. Not sure why you would think this is similar to the browser situation on iOS at least in the USA. I believe alternate browser engines are allowed in the EU now thanks to the DMA.
EU bureaucrats really earning their name here. Maybe they can realize real choice and innovation will come from allowing different operating systems on locked down hardware, instead of this crusade against Siri AI. Something that is easily fixed by downloading literally any other AI or agentic photo/video/whatever app.
I've upgraded my desktops more with GPUs and power supplies accordingly, but old laptop SSDs make fine backup enclosure drives (no I haven't seen them lose data when unplugged) and upgrading RAM for larger models on a AMD Zen 2 laptop worked pretty great for me. None of these were needed, they were just nice to be able to do mostly before the higher RAM prices started.
I am genuinely curious what causes the different colors of plasma that were visible throughout the descent. I saw at least purple, yellow, and shades of green.
The congressional technology committee Gingrich killed could've fixed this. You can't half ass democracy, you have to keep getting better at governing and convince people it's worth it.
Informative page, but the most common speeds with 2x2 MIMO probably were (in Mbps):
Far (QPSK) 4(20Mhz)/5(80Mhz)/6(160Mhz)/7(320Mhz): 28.8/130/288.23/576.47
Near (64QAM) 4(20Mhz)/5(80Mhz)/6(160Mhz)/7(320Mhz): 144.4/650/1,441.17/2,882.35
Not bad for throughput increases, though most of the increases come from more spectrum, and the reliability comes from more MIMO antennas/streams. I've had WiFi 4/5 2x2 routers and something tells me I won't see much more than what's listed above for 7. Buying a 4x4 does get you a generation of throughput in advance pretty much, if you need it.