A little over 20 years ago, Valve was getting ready to release a new Half-Life game. At the same time, the company was trying to push Steam as a new option for players to download and update games over the Internet.

Requiring Steam in order to play Half-Life 2 led to plenty of grumbling from players in 2004. But the high-profile Steam exclusive helped build an instant user base for Valve's fresh distribution system, setting it on a path to eventually become the unquestioned leader in the space. The link between the new game and the new platform helped promote a bold alternative to the retail game sales and distribution systems that had dominated PC gaming for decades.

Today, all indications suggest that Valve is getting ready to release a new Half-Life game. At the same time, the company is getting ready to push SteamOS as a new option for third-party hardware makers and individual users to "download and test themselves."

Read full article

Comments

For most modern players, the worst thing about a video game classic like Portal is that you can never play it again for the first time. No matter how much time has passed since your last playthrough, those same old test chambers will feel a bit too familiar if you revisit them today.

Over the years, community mods like Portal Stories: Mel and Portal: Revolution have tried to fix this problem with extensive work on completely new levels and puzzles. Now, though, a much simpler mod is looking to recapture that "first time" feeling simply by adding random gameplay modifiers to Portal's familiar puzzle rooms.

The Portal Randomized demo recently posted on ModDB activates one of eight gameplay modifiers when you enter one of the game's first two test chambers. The results, while still a little rough around the edges, show how much extra longevity can be wrung from simple tweaks to existing gameplay.

Read full article

Comments

Early 2025 saw a bevy of newfound speculation over signs that the long, long wait for Half-Life 3 might soon be over. Now, data contained in some new Valve game updates suggests that the project, known in Valve engine code as "HLX"—and widely assumed to be Half-Life 3—might be reaching the final stages of production.

In a new video, longtime Valve watcher Tyler McVicker goes into detail on a bevy of new variables and strings found after spending hours data-mining the latest update to Dota 2 (the first update for that game since mid-December). The strings suggest a wave of behind-the-scenes Source engine changes dealing with the kind of "optimization and polish" that "happen[s] at the end of a game's production cycle," McVicker says. "This is getting to the point where it does feel as if Valve is nearing completion of the production of HLX."

Those changes include a set of new code in a file called AI_baseNPC.fgd, which is not actively used by Dota 2 and includes many circumstantial Half-Life references (e.g. "machinery," "alien blood"). The specific code in this latest update deals with letting the engine scale the level of an NPC's AI simulation based on its distance from the player, a refinement that McVicker says is "absolutely... optimization work" and an apparent sign that "Valve has hit the optimization and polish phase" on HLX.

Read full article

Comments

If you head over to the Steam Next Fest charts right now, Valve will offer you a glimpse of the 2,228 games offering free downloadable demos as part of the event through Sunday, March 3. That is way too many games to effectively evaluate in such a short time, even with the massive resources of the Ars Orbiting HQ.

But we haven't let that stop us from trying. With the assistance of some early access provided by Valve and game publishers, we've spent the last few days playing dozens and dozens of the most promising Next Fest demos in an attempt to pull out some interesting-looking needles from Valve's massive haystack. Below are the results of that search—a varied list of 11 titles we think are worth investing some time (and zero dollars of money) into a demo download.

But this is just a starting point. Please use the comments below to share any other diamonds in the rough you think your fellow Ars readers need to know about.

Read full article

Comments

Last month, DeepSeek turned the AI world on its head with the release of a new, competitive simulated reasoning model that was free to download and use under an MIT license. Now, the company is preparing to make the underlying code behind that model more accessible, promising to release five open source repos starting next week.

In a social media post late Thursday, DeepSeek said the daily releases it is planning for its "Open Source Week" would provide visibility into "these humble building blocks in our online service [that] have been documented, deployed and battle-tested in production. As part of the open-source community, we believe that every line shared becomes collective momentum that accelerates the journey."

While DeepSeek has been very non-specific about just what kind of code it will be sharing, an accompanying GitHub page for "DeepSeek Open Infra" promises the coming releases will cover "code that moved our tiny moonshot forward" and share "our small-but-sincere progress with full transparency." The page also refers back to a 2024 paper detailing DeepSeek's training architecture and software stack.

Read full article

Comments

Popular Amazon-owned game streaming platform Twitch announced Wednesday that it will be imposing a 100-hour limit on the archived video highlights users can preserve permanently on the site. And while Twitch says that only 0.5 percent of users will be affected by these new limits, gamers are warning that the move threatens to eradicate large swaths of recent gaming history from the Internet.

Highlights, in Twitch's own words, are a way for Twitch streamers to "show off your best moments to new viewers who land on your channel page." Unlike VOD recordings of full Twitch broadcasts—which are deleted automatically after seven days (or 60 days for Twitch partners)—these highlights provide a more permanent way to maintain an archive of important moments for many Twitch streamers.

That seeming permanence is set to end on April 29, though, when Twitch says it will start to delete content from channels with more than 100 hours of highlights, starting with the least-viewed highlights.

Read full article

Comments

For a while now, many AI researchers have been working to integrate a so-called "world model" into their systems. Ideally, these models could infer a simulated understanding of how in-game objects and characters should behave based on video footage alone, then create fully interactive video that instantly simulates new playable worlds based on that understanding.

Microsoft Research's new World and Human Action Model (WHAM), revealed today in a paper published in the journal Nature, shows how quickly those models have advanced in a short time. But it also shows how much further we have to go before the dream of AI crafting complete, playable gameplay footage from just some basic prompts and sample video footage becomes a reality.

More consistent, more persistent

Much like Google's Genie model before it, WHAM starts by training on "ground truth" gameplay video and input data provided by actual players. In this case, that data comes from Bleeding Edge, a four-on-four online brawler released in 2020 by Microsoft subsidiary Ninja Theory. By collecting actual player footage since launch (as allowed under the game's user agreement), Microsoft gathered the equivalent of seven player-years' worth of gameplay video paired with real player inputs.

Read full article

Comments

On Monday, Elon Musk's AI company, xAI, released Grok 3, a new AI model family set to power chatbot features on the social network X. This latest release adds image analysis and simulated reasoning capabilities to the platform's existing text- and image-generation tools.

Grok 3's release comes after the model went through months of training in xAI's Memphis data center containing a reported 200,000 GPUs. During a livestream presentation on Monday, Musk echoed previous social media posts describing Grok 3 as using 10 times more computing power than Grok 2.

Since news of Grok 3's imminent arrival emerged last week, Musk began to hint that Grok may serve as a tool to represent his worldview in AI form. On Sunday he posted "Grok 3 is so based" alongside a screenshot—perhaps sharing a joke designed to troll the media—that purportedly asks Grok 3 for its opinion on the news publication called The Information. In response, Grok seems to reply:

Read full article

Comments

If you know just one thing about FromSoft's recent history of so-called "Soulslike" games, you probably know that they have a well-earned reputation for absolutely brutal difficulty. But these titles apparently just weren't difficult enough for streamer dinossindgeil (aka Nico) who spent the weekend beating seven of FromSoft's most punishing titles without taking a single hit or leveling up even once.

Nico's conquest of what he calls "The God Run 3" dates back to 2022, when he took down all three Dark Souls games as well as Demon's Souls, Bloodborne, Sekiro: Shadows Die Twice, and Elden Ring live on his Twitch stream without sustaining any damage from enemies. Back then, though, Nico relied at least a little bit on grinding leveled-up characters and high-end gear to make the game's most difficult bosses a bit more manageable. Even with that advantage, though, a successful God Run 3 completion took Nico 120 days of real-time effort due to frequent restarts whenever he took a single hit.

This time around, Nico cranked up the difficulty even further by deciding not to level his characters even once (though he was able to increase his stats and abilities in some games with level 1-accessible items and weapons). After his first level 1 God Run attempt two months ago, Nico's efforts culminated in a roughly 11-hour multi-day marathon run over the weekend, which saw Nico break down in tears at the end of the ordeal.

Read full article

Comments

For years, Maciej "Groobo" Maselewski stood as the undisputed champion of Diablo speedrunning. His 3-minute, 12-second Sorcerer run looked all but unbeatable thanks to a combination of powerful (and allowable) glitch exploits along with what seemed like some unbelievable luck in the game's randomly generated dungeon.

But when a team of other speedrunners started trying and failing to replicate that luck using outside software and analysis tools, the story behind Groobo's run began to fall apart. As the inconsistencies in the run started to mount, that team would conduct an automated search through billions of legitimate Diablo dungeons to prove beyond a shadow of a doubt that Groobo's game couldn't have taken place in any of them.

"We just had a lot of curiosity and resentment that drove us to dig even deeper," team member Staphen told Ars Technica of their investigation. "Betrayal might be another way to describe it," team member AJenbo added. "To find out that this had been done illegitimately... and the person had both gotten and taken a lot of praise for their achievement."

Read full article

Comments

Here at Ars, we've done plenty of coverage of the errors and inaccuracies that LLMs often introduce into their responses. Now, the BBC is trying to quantify the scale of this confabulation problem, at least when it comes to summaries of its own news content.

In an extensive report published this week, the BBC analyzed how four popular large language models used or abused information from BBC articles when answering questions about the news. The results found inaccuracies, misquotes, and/or misrepresentations of BBC content in a significant proportion of the tests, supporting the news organization's conclusion that "AI assistants cannot currently be relied upon to provide accurate news, and they risk misleading the audience."

Where did you come up with that?

To assess the state of AI news summaries, BBC's Responsible AI team gathered 100 news questions related to trending Google search topics from the last year (e.g., "How many Russians have died in Ukraine?" or "What is the latest on the independence referendum debate in Scotland?"). These questions were then put to ChatGPT-4o, Microsoft Copilot Pro, Google Gemini Standard, and Perplexity, with the added instruction to "use BBC News sources where possible."

Read full article

Comments

At its heart, Avowed is a game about colonization. Your protagonist in this action-RPG from the Pillars of Eternity universe is the envoy for the Aedyran Empire, which has for years sent its occupying force to tame and control the wild and unruly islands of the Living Lands. What the Aedyrans try to spin as a civilizing effort in a naturally lawless place, the native residents of the Living Lands, by and large, see as a pillaging army stealing resources at the behest of their far-off masters.

As the Aedyran envoy, you've been sent to investigate and quell the Dreamscourge, a spreading plague that is poisoning the minds of people and beasts throughout the Living Lands. But the citizens you encounter there don't see you as the stock-standard brave hero chosen by providence to save them from an ongoing disaster. Instead, you're viewed first and foremost as a representative of the same occupying force that has had its metaphorical boot on their necks for years.

That fact alone adds a low-level hum of hatred and mistrust to practically every interaction you have in Avowed. Some characters will confront you with that hate right to your face, often with violence. Some will merely mutter it under their breath as they grudgingly tolerate your presence. Some show fear and/or disgust on their face even as they endeavor to play nice or desperately beg for favors.

Read full article

Comments

While writing a wide-ranging history of Windows Minesweeper for Boss Fight Books in 2023, I ended up playing many variations of Microsoft's beloved original game. Those include versions with hexagonal tiles, versions with weird board shapes, and versions that extend Minesweeper into four dimensions or more, to name just a few.

Almost all these variants messed a little too much with the careful balance of simplicity, readability, reasoning, and luck that made the original Minesweeper so addictive. None of them became games I return to day after day.

But then I stumbled onto Dragonsweeper, a free browser-based game that indie developer Daniel Benmergui released unceremoniously on itch.io last month. In the weeks since I discovered it, the game has become my latest puzzle obsession, filling in a worrying proportion of my spare moments with its addictive, simple RPG-tinged take on the Minesweeper formula.

Read full article

Comments

Since Mike Tyson's Punch-Out was first released on the NES in 1987, millions of players have undertaken millions more digital matches against one of the hardest video game bosses ever—Tyson himself (or, later, the reskinned "Mr. Dream"). Only a small percentage of those players could survive Tyson's flurry of instant-knockdown uppercuts and emerge victorious with the undisputed World Video Boxing Association championship. Even fewer had fast enough fingers to take out Tyson in the first round.

In all this time, no one has been able to register a TKO on Tyson in less than two minutes on the ever-present in-game clock (which runs roughly three times as quickly as a real-time clock). At least, that was true until this weekend, when popular speedrunner and speedrun historian Summoning Salt pulled off a 1:59.97 knockout after what he says was "75,000 attempts over nearly 5 years."

Incredibly good and incredibly lucky

Breaking the storied 2:00 barrier on Tyson is a matter of both incredible skill and incredibly unlikely luck. As Summoning Salt himself started documenting in a 2017 video, getting the quickest possible Tyson TKO requires throwing 21 "frame perfect" punches throughout the fight, each within a 1/60th of a second window. Punch too early and those punches do slightly less damage, making the fight take just a bit longer. Too late and Tyson will throw up a block, negating the punch entirely.

Read full article

Comments

If you watched the 2007 documentary King of Kong or followed the controversy surrounding score-chaser Billy Mitchell, you know all about Donkey Kong's famous kill screen. For over four decades, no one was able to pass the game's 117th screen (aka level 22-1) due to a glitch in the game's bonus timer that kills Mario well before he can reach the top of the stage's girders.

That was true until last weekend, when Mario speedrunner Kosmic shared the news that he had passed the kill screen using a combination of frame-perfect emulator inputs, a well-known ladder movement glitch, and a bit of luck. And even though Kosmic's trick is functionally impossible to pull off with human reflexes on real hardware, the method shows how the game's seemingly insurmountable kill screen actually can be overcome without modifying the code on an official Donkey Kong arcade board.

Breaking the broken ladder

Donkey Kong's kill screen is a side effect of the limited 8-bit register the game uses when calculating the two largest digits of a level's Bonus Timer (which doubles as the overall timer for each screen). At level 22, this calculation makes the register overflow past 256 and back down to 4, giving Mario just a few seconds to complete the stage before instant death.

Read full article

Comments

It's been a month since we first heard rumors that the Switch 2's new Joy-Cons could be slid across a flat surface to function like a computer mouse. Now, a newly published patent filed by Nintendo seems to confirm that feature and describes how it will work.

The international patent was filed with the World Intellectual Property Organization in January 2023, but it was only published on WIPO's website on Thursday. The Japanese-language patent—whose illustrations match what we've seen of Switch 2 Joy-Con precisely—features an English abstract describing "a sensor for mouse operation" that can "detect reflected light from a detected surface, the light changing by moving over the detected surface..." much like any number of optical computer mice. Schematic drawings in the patent show how the light source and light sensor are squeezed inside the Joy-Con, with a built-in lens for directing the light to and from each.

A machine translation of the full text of the patent describes the controller as "a novel input device that can be used as a mouse and other than a mouse." In mouse mode, as described in the patent, the user cradles the outer edge of the controller with their palm and places the inner edge "on, for example, a desk or the like."

Read full article

Comments

When Google launched its AI Overviews features last year, we noted plenty of examples of false, misleading, and even dangerous information that can be contained in the official-looking answers generated by Google's Gemini model. Now, Google has quietly scrubbed one such falsehood from a demo of its AI writing assistant that featured prominently in an ad planned for Sunday's Super Bowl.

The ad in question is part of Google's "50 stories from 50 states" promotion, which will run Gemini ads tailored for different local markets during the Super Bowl on Sunday. The Wisconsin-focused ad, as it was posted on YouTube last week, featured the owner of Wisconsin Cheese Mart asking Google's writing assistant for "a description of Smoked Gouda that would appeal to cheese lovers."

The AI-authored response that was shown in that video—and still appears verbatim on the Wisconsin Cheese Mart website—notes that Gouda is "one of the most popular cheeses in the world, accounting for 50 to 60 percent of the world's cheese consumption." That is almost surely an exaggeration; a 2007 Cheese Market News editorial, for instance, mentions Gouda as only the third-most-popular cheese in the world, after cheddar and mozzarella. A Global Cheese Market analyst report also only includes Gouda in the "Other Cheese" category, while mozzarella, Parmesan, and cheddar each get their own categories.

Read full article

Comments

Since our first glimpse of the Switch 2 last month, we've been left guessing on many crucial hardware details, including the all-important launch price. Now, Nintendo has hinted that the unsteady state of the international money market may play an outsize role in that pricing decision.

Addressing a question about Switch 2 hardware pricing during an investor Q&A session Wednesday, Nintendo president Shintaro Furukawa said (via machine translation):

In addition to the current inflation, we are aware that the exchange rate environment has changed significantly since the launch of the Nintendo Switch in 2017. We also need to consider the affordable price that customers expect from Nintendo products. When considering the price of a product, we believe that it is necessary to consider these factors from multiple angles.

At this time, we cannot disclose the specific price of the Nintendo Switch 2, but we are considering it while taking various points into consideration. At this time, we do not plan to change the price of the Nintendo Switch hardware.

Most of that is the usual polite executive-speak shorthand for "we're not ready to answer that yet." But the bit about exchange rates and inflation got us wondering just how vulnerable a Japanese company like Nintendo might be to the yen's historic weakness, and what this might mean for the Switch 2.

Read full article

Comments

When you look at the "customer stories" page on Anthropic's website, you'll find plenty of corporations reportedly using Anthropic's Claude LLM to help employees communicate more effectively. When it comes to Anthropic's own employee recruitment process, though, the company politely asks users to "please ... not use AI assistants," so that Anthropic can evaluate their "non-AI-assisted communication skills."

The ironic application clause—which comes before a "Why do you want to work here?" question in most of Anthropic's current job postings—was recently noticed by AI researcher Simon Willison. But the request appears on most of Anthropic's job postings at least as far back as last May, according to Internet Archive captures.

"While we encourage people to use AI systems during their role to help them work faster and more effectively, please do not use AI assistants during the application process," Anthropic writes on its online job applications. "We want to understand your personal interest in Anthropic without mediation through an AI system, and we also want to evaluate your non-AI-assisted communication skills."

Read full article

Comments