Author: admin

CD Projekt Red’s latest update for Cyberpunk 2077 is here. The update is packed with some exciting new features. Update 2.3 has introduced some major customization, vehicles, and immersion improvements, right alongside technical upgrades for smoother gameplay. Here is the breakdown of the biggest changes that are coming to Night City.

Top 10 changes in Cyberpunk 2077 update 2.3

Landing on July 17, 2025, on PlayStation 5, Xbox Series X|S and PC, the update delivers a significant package. While it’s not a massive narrative expansion, it does inject fresh life into highly requested features, which are aimed at making navigation of the dystopian metropolis more stylish and smoother than ever before.

AutoDrive: For a hands-free experience

AutoDrive feature Cyberpunk 2077 the most anticipated addition with the update. AutoDrive allows players to sit back while the car navigates Night City. Just set the destination or enable the free-roam mode. The vehicle would follow traffic rules, avoid collisions and even stop at the red lights. The feature disengages during the combat to ensure you remain in control as things start to get heated up.

Cinematic cruising camera

Alongside AutoDrive is a paired cinematic cruising camera mode that dynamically cycles through all cinematic angles as the car moves. It captures stunning vistas. It allows players to create their own Night City music videos, as the car drives itself.

Delamin Taxi Service

This feature can be unlocked by completing the ” Don’t Lose Your Mind quest. The feature allows players to summon the Delamin cab through the vehicle call menu. Just input the destination and sit back to enjoy the ride. It’s a seamless way to explore around.

Four new vehicles

Update 2.3 adds 3 new cars and a motorcycle. Each of these are coming with new vehicles Cyberpunk 2077 2.3 update, and they are tied to the unique side jobs.

Expanded vehicle customization

TwinTone and CystalCoat now support a total of 32 vehicles, including the motorcycles. The hacked version allows lower-end cars to get a fresh paint job. There are over 370 new paint schemes—168 unique and 205 generics. Use vehicle customization Cyberpunk 2077 to make sure your ride stands out in Night City.Note: Partner vehicles and motorcycles can now use CrystalCoat.

Photo mode is revolutionized

Capture the Night City like you never did before with Photo mode in Cyberpunk 2077. Spawn the 27 new NPCs, including some fan favourites like Cassel twins and Brendan. You can even change character outfits—NPCs and V or alter time and weather instantly. Apart from it, use it to skip frames to have perfect action shots. Use the powerful new Look-At system and the craft custom filters along with advanced color balance tools. With it all, players now have more creative freedom in Cyberpunk 2077 than before.

Mac debut and cross-progression

Cyberpunk 2077 has finally and officially arrived on “Mac with Apple silicon”. The Mac version Cyberpunk 2077 via Mac App Store, Epic Games Store and Steam. The existing PC owners on the supported platforms will have free access. It would feature frame generation, ray tracing, HDR tuned for the Apple XDR displays and frame generation too.

AMD FSR 3.1 and Intel XeSS 2.0 support

The PC players would benefit from the AMD FSR 3.1 frame generation, apart from Intel XeSS 2.0. It will improve visuals and performance. FSR 4 support will arrive later with the AMD driver update.

HDR10+ and Console VRR upgrades

HDR10+ Gaming is used for compatible displays. Apart from it, VRR support is now available on consoles. It will reduce screen tears and will deliver smoother gameplay on Xbox Series X|S and PlayStation 5.

Quality of life improvements

Cyberpunk 2077 update 2.3 is more immersive than it’s ever been. Whether you are customizing cars, letting AutoDrive handle wheels or snapping pictures, the new patch arriving on July 17, 2025, will change the gaming experience. Enjoy on PlayStation 5, Xbox Series X|S or PC, as you hit the streets.

In a bold new step against online cheating, Epic Games is now making Fortnite cheaters issue public apologies as part of legal settlements. The move is part of the company’s broader effort to protect the integrity of the game and send a message to those who try to gain unfair advantages.

We took legal action against two people who cheated and broke our rules:

One sold and used cheats and the other carried out cyber attacks on content creators who were livestreaming gameplay (aka: DDoS attacks). Both have been ordered to stop these activities and are banned from…

— Epic Games Newsroom (@EpicNewsroom) July 14, 2025

According to a recent report by The Express Tribune, players who were caught using illegal software to cheat in Fortnite were not only banned but also legally compelled to admit guilt. These cheaters, mostly based in the U.S. and Europe, signed court-enforced agreements that included statements like:

“I regret cheating in Fortnite and will not do it again.”

These apologies have been posted publicly online. Sometimes these apologies also appear on forums such as Reddit or personal websites, making the embarrassment part of the punishment.

Legal Action Against Fortnite Cheaters Isn’t New for Epic

Epic Games has a history of taking legal action against cheaters and cheat developers. The company previously filed lawsuits against multiple individuals who created or sold aimbots, wall hacks, and other exploits for Fortnite. In the past, some cases have ended with heavy fines, while others required the offenders to promise never to cheat again or face future legal consequences.

The public apology approach is a new twist that combines legal deterrence with public accountability.

Why Epic Is Doing This

With over 400 million registered players and a massive in-game economy, cheating in Fortnite can ruin competitive balance and damage the experience for millions. By forcing cheaters to publicly own up to their actions, Epic hopes to:

A spokesperson for Epic Games said the company is “committed to fair play” and will continue pursuing cheaters legally when necessary.

Since the groundbreaking 2017 publication of “Attention Is All You Need,” the transformer architecture has fundamentally reshaped artificial intelligence research and development. This innovation laid the foundation for Large Language Models (LLMs) and Video Language Models (VLMs), fueling a wave of productization across the industry. A defining milestone was the public launch of ChatGPT in November 2022, which brought transformer-powered AI into mainstream use. Since then, LLMs have enabled a broad spectrum of applications, from conversational agents to advancements in medical research.

However, running these LLMs efficiently presents substantial challenges, particularly on edge computing devices and legacy hardware architectures that were designed before the widespread adoption of large language models.

One of the significant difficulties facing AI processors is the sheer size of LLMs compared to prior state-of-the-art CNNs (Convolutional Neural Networks), RNNs (Recurrent Neural Networks), and other network types.

Among these CNNs and RNNs, an 85 million-parameter model would have been considered large. In comparison, even a modestly sized LLM might have 1B parameters, while models with 8B parameters, and larger, are commonplace. Said plainly, there is no mass market, cost-effective method to load that many parameters on a single chip; thus, pre-existing solutions may not be effective.

Consider Llama 3.2, Meta’s latest generation of LLMs, which introduced significant advancements in both text and multimodal (text + vision) AI capabilities. This release expands on previous versions with new model variants and features designed for both enterprise and edge-device deployment. Llama 3.2 contains 1 billion parameters. Furthermore, attention operations (O) increase with context size (n). During prefill, where operations are predominantly done in parallel, the compute load is a function of the square of the context size, O(n2). The prefill phase is compute-bound, meaning its speed is limited by the raw computational power of the hardware. In contrast, decode is predominantly sequential, and the compute load is an order of magnitude smaller, O(n)*n per token. However, the decode phase is dominated by memory access speed rather than compute power, and the per-token compute cost can be orders of magnitude higher than during prefill.

Exploring LLM inference flow

The LLM inference flow begins with a user prompt, which is a sentence spoken or input by the user. In Figure 1 below, we use the example of “May the force.” The user prompt is first translated into what is called a token, a mathematical representation of the user prompt, using a processing mechanism aptly named the “tokenizer.” The token is then sent to the inference processing steps, where they are divided into two phases: the prefill phase and the decode phase.

In the prefill phase, all the tokens are sent at once through a series of transformer blocks. At each transformer layer, the Key (K) and Value (V) vectors for each token are stored in a cache — this is called the KV cache or attention cache. This cache is then used to make subsequent generations faster. In the decode phase, one response at a time is generated, as this part is sequential in nature.

In Figure 1, examples of this are the generation of “be, with” and “you”.

Fig. 1: LLM inference flow.

During the prefill stage, the model needs to compute attention over all previous tokens. However, during the decode stage, the processor only needs to compute attention over the new token because it reuses the cached values from prefill. This enables efficient autoregressive decoding — the processor doesn’t need to recompute everything from scratch each time. The prefill stage is compute-heavy because it processes the entire input prompt.

Compute cost scales linearly with prompt length – if the prompt is N tokens and the model has L layers, the processor must do N × L full forward passes. However, the decode stage is relatively fast after that because the system processes only one token at a time, reusing the cache. This raises an issue, though – as the phases are very different in compute, memory, and power, how can a solution be enabled which is optimal for both?

LLM inference runtime

Fig. 2: LLM inference runtime.

Let’s also examine the runtime changes between traditional AI networks and LLMs, which are illustrated in Figure 2.

Traditional CNNs have a simple, monolithic runtime with only two phases:

• Data loading phase
• Inference phase

LLMs introduce a multi-phase runtime system with five distinct phases, each with different computational and memory requirements.

Prefill Phase

• Processes the initial user prompt by embedding and tokenizing the entire input sequence
• Runs all transformer layers in full sequence mode (dense computation)
• Initializes and populates the Key-Value (KV) cache with attention values
• Generally has higher per-token latency due to sequence processing
• May use microbatching for lengthy inputs

Decode Phase

• Generates output tokens one at a time autoregressively
• Only processes the last generated token per step
• Retrieves past tokens from the KV cache for efficiency
• Computes self-attention only against past tokens
• Highly optimized with KV caching and batching

Inactive Phase

• No computation occurs, but the sequence remains “alive” in memory
• Occurs when waiting for new user input in streaming/chat interfaces
• KV cache remains in memory (costly resource usage)
• Can become a bottleneck in high-throughput systems with many cached sequences

Follow-up Prefill

• Triggered when new user input is appended to a partially generated sequence (multi-turn conversations)
• Processes new input as a short prefill segment, appending to existing cached context
• Updates KV cache with new tokens
• Distinct from initial prefill as it operates on shorter, appended segments

Retired Phase

• Sequence is terminated and removed from the active pool
• KV cache is freed and resources are released
• Triggered by conversation completion, user cancellation, or timeouts
• Frees memory and scheduling capacity for other sequences

This multi-phase complexity significantly exacerbates deployment difficulties for LLMs compared to traditional AI networks. Managing multiple phases simultaneously, maintaining expensive KV caches, and handling dynamic transitions between phases create substantial challenges for efficient LLM deployment and resource management.

New architectures for LLMs

Large Language Models introduce distinct challenges for inference hardware and software systems, including complex model architectures, demanding runtime computations, transformer-specific operations, and implementation considerations. To address these requirements, AI processing platforms must advance to accommodate diverse data representations, support multiple precision formats, deliver enhanced computational throughput, and enable efficient multi-core processing architectures.

Expedera explores this further in our white paper at: https://www.expedera.com/transformers-at-the-edge/

The result of the pioneering joint Psion and Nokia smartphone effort is still out there on GitHub.

Smartphones are everywhere. They are entirely commoditized now. Most of them run Android, which uses the Linux kernel. The rest run Apple’s iOS, which uses the same XNU kernel as macOS. As we’ve said before, they’re not Unix-like, they really are Unix™.

There have been a bunch of others. BlackBerry tried hard with BB10, but even a decade ago, it was over. It was based on QNX and Qt, and both of those are doing fine. We reported last year that QNX 8 is free to use again. Palm’s WebOS ended up with HP and now runs in LG smart TVs – but it’s Linux underneath.

The most radical, though, was probably Symbian. The Register covered it at length back in the day, notably the epic Psion: the Last Computer feature, followed by the two-part Symbian, The Secret History, and Symbian UI Wars features.

Built from scratch in the late 1990s in the then-relatively new C++, it evolved into a real-time microkernel OS for handhelds, with the radical EKA2 microkernel designed by Dennis May and documented in detail in the book Symbian OS Internals. There’s also The Symbian OS Architecture Sourcebook [PDF]. An official version of the source code is on GitHub, and other copies are out there.

We liked this description from CHERI Project boffin David Chisnall:

Before Nokia was assimilated and digested by Microsoft, it open sourced the OS, and despite some licensing concerns, it’s still there.

It strikes this vulture as odd that while work continues on some ground-up FOSS OS projects in C++, such as the Genode OS, or Serenity OS, which we looked at in 2022,the more complete Symbian, which shipped on millions of devices and for a while had a thriving third-party application market, languishes ignored.

(Incidentally, the Serenity OS project lead has moved on to the independent Ladybird browser, which we looked at in 2023. Work on the OS continues, now community-led.)

Symbian’s progenitor, Psion EPOC32, predates much of the standardization of C++ – much as BeOS did. We’ve seen comments that it was not easy to program, but tools such as P.I.P.S. made it easier. Nokia wasted vast effort on multiple incompatible UIs, which have been blamed for tearing Symbian apart, but none of that matters now: adapt some existing FOSS stuff, and forget backwards compatibility. Relatively few of the apps were FOSS, and who needs touchscreen phone apps on a Raspberry Pi anyway? Qt would be ideal – it’s a native C++ tool too.

Fans of all manner of 20th century proprietary OSes from AmigaOS to OS/2 bemoan that these never went open source. Some of BeOS made it into PalmOS Cobalt but that sank. Palm even mulled basing an Arm version of PalmOS on Symbian, but the deal fell through.

Some of those OSes have been rebuilt from scratch, including AmigaOS as AROS and BeOS as Haiku. But they run on Intel. Neither runs natively on Arm, and yet Symbian sits there ignored. Sometimes you can’t even give the good stuff away. ®

Samsung files trade protection for ‘Galaxy Z TriFold’ in Korea