HN user

terramex

2,760 karma
Posts5
Comments295
View on HN
GPT-5.6 13 days ago

I am on Plus subscription and see Terra and Luna in Codex, but no sign of Sol. Will it be available only on Pro plans?

Unfortunately, console storage and memory prices have increased by more than 2.5x and we expect another doubling by the fall of 2027.

I have not seen such pessimistic future prediction from any of major consumer electronics producers yet.

Humanoidal robots make sense when they need to operate in spaces designed for humans bodies. Cars are designed and built to be used and serviced by humans, especially their interiors so you need humanoid robot to automate building them. Car exteriors are not built for humans to interact with so they are already being built by specialized robots.

I run Gemma 4 26B-A4B with 256k context (maximum) on Radeon 9070XT 16GB VRAM + 64GB RAM with partial GPU offload (with recommended LMStudio settings) at very reasonable 35 tokens per second, this model is similiar in size so I expect similar performance.

https://en.wikipedia.org/wiki/Thread_(network_protocol)

Thread is an IPv6-based, low-power mesh networking technology for Internet of things (IoT) products.

Often used as a transport for Matter (the combination being known as Matter over Thread), the protocol has seen increased use for connecting low-power and battery-operated smart-home devices.

Thread uses 6LoWPAN, which, in turn, uses the IEEE 802.15.4 wireless protocol with mesh communication (in the 2.4 GHz spectrum), as do Zigbee and other systems. However, Thread is IP-addressable, with cloud access and AES encryption. A BSD-licensed open-source implementation of Thread called OpenThread is available from and managed by Google.

There was unofficial Lapz app ( https://www.lapz.io ) on TestFlight but got taken down after complaint from F1: https://www.theverge.com/2024/11/20/24301420/apple-vision-pr...

I watched few races on it and it was pretty great. You logged in with your F1TV account and could set up streams around you plus there was 3D view of the track with current positions of cars on it, you could look at car and pinch and driver view opened in pop-up window attached to car model. Couple times I noticed a car stopped on track before it was shown on main broadcast.

It was also revealed that the Polish Internal Security Agency (ABW) had, in October 2022, submitted a case against Newag regarding the abovementioned software manipulation incidents to the prosecutor's office in Nowy Sącz, which initially downplayed the incident until said findings publicly came to light, after which, the investigation was taken over by the regional prosecutor's office in Kraków on suspicion of crimes committed under Article 269 §1 and Article 286 §1 of the Polish Penal Code.

The Sejm's [Sejm is "the lower house of the bicameral parliament of Poland"] Parliamentary Committee for Combating Transport Exclusion subsequently convened three hearings regarding the abovementioned allegations on 17 January, 27 February and 26 March 2024, whose participants included representatives of the Dragon Sector team, Newag, railway operators and members of the Sejm.[19]

https://en.wikipedia.org/wiki/Newag#2023_revelation_of_softw...

I could not find any later updates in polish media.

How good are modern external hard drives? Is it worth paying for more internal SSD storage or is it more reasonable to get high quality USB one?

Lvl 3 is supposed to support other models and providers in the future too. I hope it will support every server with simple, standard API so I can run self-hosted LLama 3 (or whatever will be released in next 6-12 months).

Why even open any other app?

This was my first thought when I saw Rabbit r1 - will all of us become backend developers just glueing various API between legacy services and LLMs? Today seems like another step in that direction.

They don't "train on log₂(3) bit". Gradients and activations are still calculated at full (8-bit) precision and weights are quantised after every update.

This makes network minimise loss not only with regard to expected outcome but also minimises loss resulting from quantisation. With big networks their "knowledge" is encoded in relationships between weights, not in their absolute values so lower precision work well as long as network is big enough.

If onboard computer would sense that Ship is going somewhere it should not be it would trigger FTS (Flight Termination System) and destroy it with debris falling into Atlantic Ocean (exclusion zone was there as well).

Once Ship finished burn it simply had to land in predicted landing area in Indian Ocean, it did not have enough fuel on board to significantly change trajectory even if computer went crazy.