GoKawiil - Latest Tech News & Aggregated Headlines

Tesla shuts down in-house Dojo AI supercomputer project

engadget.com Unknown 2025-11-29 00:54:20

As first reported by Bloomberg , Tesla is disbanding the team behind Dojo , its in-house AI-training supercomputer, and reassigning remaining staff to other projects within the company. This marks a shift in the company's compute sourcing strategy for its AI-focused initiatives such as autonomous driving and the Optimus robot . Head of Dojo Peter Bannon is leaving Tesla, which is the latest departure after roughly 20 Dojo team members recently left to form DensityAI . In a response to the Bloom

Topics: ai6 company dojo tesla training

Shop Amazon

Amtrak NextGen Acela Debuts on August 28

news.ycombinator.com Beth Toll 2025-11-29 14:30:23

A new era of high-speed rail to begin between Boston and Washington, DC WASHINGTON – All aboard the future of East Coast travel! Starting Thursday, Aug. 28, customers can travel on Amtrak’s NextGen Acela – America’s high-speed train – connecting the Northeast Corridor between Washington, D.C., New York City, and Boston. 🛤️ This historic launch introduces: More trains : 28 brand-new trains entering service through 2027 : 28 brand-new trains entering service through 2027 More seats : 27% more

Topics: acela amtrak new nextgen trains

Shop Amazon

Achieving 10,000x training data reduction with high-fidelity labels

news.ycombinator.com Unknown 2025-11-30 03:11:20

Classifying unsafe ad content has proven an enticing problem space for leveraging large language models (LLMs). The inherent complexity involved in identifying policy-violating content demands solutions capable of deep contextual and cultural understanding, areas of relative strength for LLMs over traditional machine learning systems. But fine-tuning LLMs for such complex tasks requires high-fidelity training data that is difficult and expensive to curate at the necessary quality and scale. Stan

Topics: content data llms model training

Shop Amazon

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

venturebeat.com Ben Dickson 2025-12-03 05:11:20

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now A new study from the Anthropic Fellows Program reveals a technique to identify, monitor and control character traits in large language models (LLMs). The findings show that models can develop undesirable personalities (e.g., becoming malicious, excessively agreeable, or prone to making things up) either in response to user prompts or as an

Topics: ai model persona training vectors

Shop Amazon

Universal Adds ‘No AI Training’ Warning to Movies

gizmodo.com Aj Dellinger 2025-12-03 10:40:39

AI is not invited to movie night. According to The Hollywood Reporter, Universal Pictures has started including a message in the credits of its films that indicates the movie “may not be used to train AI” in part of an ongoing effort by major intellectual property holders to keep their content from getting fed into the machines (at least without being paid for it). The warning, which reportedly first appeared at the end of the live-action How to Train Your Dragon when it hit theaters in June, h

Topics: ai companies openai train used

Shop Amazon

These Democrats Think the Party Needs AI to Win Elections

wired.com Makena Kelly 2025-12-04 20:52:56

The 2024 election cycle saw artificial intelligence deployed by political campaigns for the very first time. While candidates largely avoided major mishaps, the tech was used with little guidance or restraint. Now, the National Democratic Training Committee (NDTC) is rolling out the first official playbook making the case that Democratic campaigns can use AI responsibly ahead of the midterms. In a new online training, the committee has laid out a plan for Democratic candidates to leverage AI to

Topics: ai campaigns candidates training use

Shop Amazon

Germany's identity crisis: The trains no longer run on time

news.ycombinator.com Aaron Wiener 2025-12-06 10:32:58

BERLIN — Germany: the land of beer, sausage and trains that run on time. Actually, make that the land where 56 percent of trains run on time. More precisely (or imprecisely, depending on how much of a rush you are in), the land where 56 percent of trains arrive within six minutes of the scheduled time — which is the cushion Deutsche Bahn, the national railroad company, allows itself for an “on-time” arrival. In Germany, punctuality is part of the national ethos. So to hear Germans talk about it

Topics: advertisement germany said train trains

Shop Amazon

The Subway Game (1980)

news.ycombinator.com Unknown 2025-12-06 04:02:19

The Subway Game Copyright © 1980, Peter R. Samson (to home page) The Subway Game was a diversion that developed out of the frequent visits to New York by myself and various friends at M.I.T. In its basic form it requires two participants: an innocent victim, called the Contestant; and a more knowledgeable companion, called the Monitor. With some allowances, the game can be viewed as a simulation of the following scenario. A stranger to New York is going to visit some friends there. He gets a

Topics: contestant new station subway train

Shop Amazon

Persona vectors: Monitoring and controlling character traits in language models

news.ycombinator.com Unknown 2025-12-10 20:38:06

Language models are strange beasts. In many ways they appear to have human-like “personalities” and “moods,” but these traits are highly fluid and liable to change unexpectedly. Sometimes these changes are dramatic. In 2023, Microsoft's Bing chatbot famously adopted an alter-ego called "Sydney,” which declared love for users and made threats of blackmail. More recently, xAI’s Grok chatbot would for a brief period sometimes identify as “MechaHitler” and make antisemitic comments. Other personali

Topics: model persona training trait vectors

Shop Amazon

The Best Smart Home Gyms, as Recommended by a Fitness Expert

cnet.com See Full Bio 2025-12-10 16:00:00

The Tempo Studio is a smart home gym that resembles an armoire, meant to blend in with your home. It's an ideal smart home gym to own whether you're new or experienced with strength training. The Tempo Studio's basic package comes well-equipped with two dumbbell bars, weight collars and five sets of weight plates from 1.25 to 25 pounds. You have the option to upgrade your package to the Tempo Studio Plus or Tempo Studio Pro, which includes additional weights and accessories like a bench and a ba

Topics: equipment home studio tempo training

Shop Amazon

A short post on short trains

news.ycombinator.com Shaked Koplewitz 2025-12-16 22:15:07

Epistemic status: Main part is well-supported but may have some minor errors. The parts about potential future lines are inherently speculative. Small Train is Good Train A while ago, I wrote about how elevated trains are the greatest urbanism cheat code, increasing the amount of track miles you can build per dollar (or per year) by a factor of 2-4. And while I don’t have anything else on that order of magnitude, I do have one more easy 20-50% gain: Run shorter trains. The basic idea is simpl

Topics: build capacity jersey small trains

Shop Amazon

Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL

news.ycombinator.com Unknown 2025-12-18 14:12:03

🤓 Terminal-Bench-RL: Training Long-Horizon Terminal Agents with Reinforcement Learning TL;DR: I successfully built stable RL training infrastructure that scales to 32x H100 GPUs across 4 bare metal nodes for training long-horizon terminal-based coding agents. In doing so, I developed Terminal-Agent-Qwen3-32b to become the highest scoring Qwen3 agent on terminal-bench . WITHOUT training! (currently under submission): Unfortunately I am too GPU poor to train a SOTA coding agent 😅 (estimated £30

Topics: agent run task terminal training

Shop Amazon

The Best Fantasy Movies to Stream on Prime Video

cnet.com See Full Bio 2025-12-20 23:00:04

A live-action remake of 2010's How to Train Your Dragon debuted in theaters last month, and the second animated movie is getting the same treatment. That version won't be here until 2027, but you can watch Hiccup, Astrid and Toothless in 2014's How to Train Your Dragon 2 with Prime.

Topics: 2010 2027 action dragon train

Shop Amazon

Optimi-Zi(n)g Sudoku-Solving

news.ycombinator.com Olivier'S Log 2025-12-17 23:47:35

Optimi-Zi(n)g Sudoku-Solving 26 July 2025 , in Olivier's log One of the first program that I wrote in Zig (in September 2023) was a Sudoku-Solver, implementing the dancing-links (DLX) algorithm. I decided to revisit this program recently to experiment with benchmarking and try to increase its speed. Dancing-Links (DLX) algorithm applied to sudoku The Dancing-Links algorithm is an efficient backtracking algorithm to solve "exact-cover" problems, by using a matrix of 0 and 1s. Dancing Links o

Topics: cells constraints line lines sudoku

Shop Amazon

Show HN: Terminal-Bench-RL: Training Long-Horizon Terminal Agents with RL

news.ycombinator.com Unknown 2025-12-21 14:12:03

🤓 Terminal-Bench-RL: Training Long-Horizon Terminal Agents with Reinforcement Learning TL;DR: I successfully built stable RL training infrastructure that scales to 32x H100 GPUs across 4 bare metal nodes for training long-horizon terminal-based coding agents. In doing so, I developed Terminal-Agent-Qwen3-32b to become the highest scoring Qwen3 agent on terminal-bench . WITHOUT training! (currently under submission): Unfortunately I am too GPU poor to train a SOTA coding agent 😅 (estimated £30

Topics: agent run task terminal training

Shop Amazon

GLM-4.5: Reasoning, Coding, and Agentic Abililties

news.ycombinator.com Unknown 2025-12-23 12:15:52

Today, we introduce two new GLM family members: GLM-4.5 and GLM-4.5-Air — our latest flagship models. GLM-4.5 is built with 355 billion total parameters and 32 billion active parameters, and GLM-4.5-Air with 106 billion total parameters and 12 billion active parameters. Both are designed to unify reasoning, coding, and agentic capabilities into a single model in order to satisfy more and more complicated requirements of fast rising agentic applications. Both GLM-4.5 and GLM-4.5-Air are hybrid re

Topics: agentic coding glm reasoning training

Shop Amazon

The Electron E1 Processor

news.ycombinator.com Unknown 2025-12-24 12:37:42

Innovation demands processors that can keep up. Readily available processors are built on technology that is over 70 years old. This limits innovation. To meet modern demands, processors must be entirely reimagined, breaking free from the constraints that have plagued computing for decades. This spatial dataflow architecture supports general-purpose computing, without being bound by the constraints of traditional processor designs or limited by fixed-purpose accelerators. The Electron E1

Topics: computing constraints innovation processors purpose

Shop Amazon

Tram Trains

news.ycombinator.com Benedict Springbett 2025-12-31 18:37:12

We’re hiring someone in London to help grow Works in Progress's audience and sell Stripe Press books (and, soon, Works in Progress magazine subscriptions). If this could be you, please apply here! Many cities face the following problem. They have railway lines that go where people live. But these railway lines end at the edge of the city center, and don’t go out the other side. For cities with this problem, the solution is through running. Terminating a train and turning it around takes a lot

Topics: center city railway trains tram

Shop Amazon

Herringbone Tiles

news.ycombinator.com Unknown 2025-12-29 05:42:04

Herringbone Tiles Sean Barrett Silver Spaceship Software In this paper I'll describe a method for expanding on the technique of Wang Tiles for generating large 2D regions from smaller ones. I call the technique "Herringbone Wang Tiles" or just "Herringbone Tiles". It is also of particular relevance to the map system used in Infamous by Sucker Punch. For an unreleased indie CRPG I worked on in 2010, I used an extremely simple method of dungeon map generation. It involves assembling a large

Topics: constraints edge edges tile tiles

Shop Amazon

New York’s bill banning One-Person Train Operation

news.ycombinator.com Blair Lorenzo 2026-01-07 00:33:47

The New York State Legislature has just passed a bill (S4091/A04873) that would lock New York City’s transit system in the past. This bill, which would require a conductor to be on board every train operated by New York City Transit, is the technological equivalent of requiring every elevator in the city to still be staffed by an elevator operator. If you take other transit systems both across the country and around the world, you'll quickly realize that two-person train operation (TPTO) is an o

Topics: mta new train trains transit

Shop Amazon

Best Chest Strap Heart-Rate Monitors for Your 2025 Workouts, Fitness Expert-Approved

cnet.com See Full Bio 2026-01-07 00:08:00

Why we like it: The Polar H10 is ideal for outdoor activities. You'll need to download the Polar Beat app to get the most out of it. The app is available for both iOS and Android and uses Bluetooth and ANT Plus connectivity to pair with different devices. The Polar H10 can connect to two Bluetooth devices at once, so you can connect it to both your smartwatch and a compatible piece of fitness equipment, like some treadmills or exercise bikes. The heart-rate monitor is easy to clip on and adjust

Topics: heart monitor polar rate training

Shop Amazon

China’s Salt Typhoon Hackers Breached the US National Guard for Nearly a Year

wired.com Lily Hay Newman 2026-01-07 00:30:00

After reporting last week that the “raw” Jeffrey Epstein prison video posted by the FBI was likely modified in at least some ways (though there is no evidence that the footage was deceptively manipulated), WIRED reported on Tuesday that metadata analysis of the video shows approximately 2 minutes and 53 seconds were removed from one of two stitched-together clips. The United States Department of Homeland Security is facing controversy over DNA samples taken from approximately 133,000 migrant ch

Topics: data state trains video week

Shop Amazon

Delaunay Mesh Generation (2012)

news.ycombinator.com Unknown 2026-01-06 16:01:26

Delaunay Mesh Generation Our book is a thorough guide to Delaunay refinement algorithms that are mathematically guaranteed to generate meshes with high quality, including triangular meshes in the plane, tetrahedral volume meshes, and triangular surface meshes embedded in three dimensions. It is also the most complete guide available to Delaunay triangulations and algorithms for constructing them. We have designed the book for two audiences: researchers, especially graduate students, and engin

Topics: algorithms constrained delaunay refinement triangulations

Shop Amazon

Is HR ready for AI?

zdnet.com Vala Afshar 2026-01-12 12:16:13

sefa ozel/Getty Images A report from technology analyst firm Valoir has found that most companies are either already leveraging AI for HR activities, such as recruiting, learning, and talent management, or they plan to do so within the next 24 months. However, there's a significant gap in terms of policies, practices, and training for safe and effective AI adoption. Only 34% of organizations have a policy on generative AI (gen AI), and even fewer offer effective training. Also: 5 entry-level

Topics: ai hr management organizations training

Shop Amazon

Hackers Can Tamper With Train Brakes Using Just a Radio, Feds Warn

gizmodo.com Rhett Jones 2026-01-16 15:10:56

The Cybersecurity and Infrastructure Security Agency (CISA) issued an advisory last week warning that a key train system could be hacked using nothing but a radio and a little know-how. The flaw has to do with the protocol used in a train system known as the End-of-Train and Head-of-Train. A Flashing Rear End Device (FRED), also known as an End-of-Train (EOT) device, is attached to the back of a train and sends data via radio signals to a corresponding device in the locomotive called the Head-o

Topics: cisa end smith train wrote

Shop Amazon

Watch ‘How to Train Your Dragon’ Light a Bunch of Things on Fire in This Exclusive BTS Clip

gizmodo.com Sabina Graves 2026-01-17 10:00:42

Here’s an exclusive, fun look behind the scenes of Dean DeBlois’ live-action How to Train Your Dragon, showcasing the stuntwork and visual artistry that went into bringing a beloved animated franchise into the real world. io9 presents a clip from How to Train Your Dragon‘s home release that takes us to Viking training camp. Watch as stars including Mason Thames (The Black Phone) and Nico Parker (The Last of Us) get the hang of dragon riding on their mechanical bull-like dragon stand-ins. There’

Topics: action dragon dragons real train

Shop Amazon

‘How to Train Your Dragon’ Hits Digital Tomorrow With Tons of Extras

gizmodo.com Cheryl Eddy 2026-01-18 20:00:06

If you missed How to Train Your Dragon on the big screen—or more likely, saw it and want to watch it again, based on its box-office haul—tomorrow’s your lucky day. The hit live-action adaptation of the hit animated film arrives on digital July 15, with a physical version coming August 12. Best of all for fans, the home release is jam-packed with over 75 minutes of extras, including deleted scenes and multiple making-of featurettes, as well as a look at the film’s new theme park experience. Here

Topics: berk dragon film scenes train

Shop Amazon

Florida Cat Named Pepper Brings Home Never-Before-Seen Virus—for the Second Time

gizmodo.com Natalia Mesa 2026-01-21 00:00:40

Last year, Pepper, a pet cat who roams the backyards of Gainesville, Florida, helped a scientist discover a new viral strain. Now, the furry feline is back at it again. In a new study, scientists have once again discovered an exotic virus infecting a dead rodent that had been caught by Pepper. This time around, Pepper’s furry hunting trophy helped researchers pinpoint an unidentified strain of orthoreovirus, a type of virus that infects humans and other mammals. The findings, and the virus’s co

Topics: lednicky new pepper strain virus

Shop Amazon

How to scale RL to 10^26 FLOPs

news.ycombinator.com Jack Morris 2026-01-17 05:47:34

TLDR: Reinforcement learning (RL) is the next training technique for building frontier-level AI models. To make it better, we need to train on more data. The current approach of scaling many environments simultaneously is messy and complicated. Instead, I propose we find a way to do next-token prediction on the Web using RL. This way, we learn to reason from general web data, instead of just math and code. I’ve spent a good part of the past year in denial. I was in denial because when OpenAI r

Topics: model models reasoning rl training

Shop Amazon

China Working On Levitating Train That Could Get You From New York to Chicago in Two Hours

futurism.com Unknown 2026-01-20 16:45:41

As the United States struggles to keep its major cities connected by even the most barebones rail systems, China is screaming into the future with the development of a levitating bullet train. Called "maglev," short for "magnetic levitation," the train system is designed to levitate via magnets as opposed to wheels. Maglev systems can reach higher speeds much more efficiently than their wheeled counterparts, though the infrastructure needed to run them is incredibly expensive. While there are

Topics: maglev operating speed train world

Shop Amazon

Latest Tech News

Tesla shuts down in-house Dojo AI supercomputer project

Amtrak NextGen Acela Debuts on August 28

Achieving 10,000x training data reduction with high-fidelity labels

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

Universal Adds ‘No AI Training’ Warning to Movies

These Democrats Think the Party Needs AI to Win Elections

Germany's identity crisis: The trains no longer run on time

The Subway Game (1980)

Persona vectors: Monitoring and controlling character traits in language models

The Best Smart Home Gyms, as Recommended by a Fitness Expert

A short post on short trains

Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL

The Best Fantasy Movies to Stream on Prime Video

Optimi-Zi(n)g Sudoku-Solving

Show HN: Terminal-Bench-RL: Training Long-Horizon Terminal Agents with RL

GLM-4.5: Reasoning, Coding, and Agentic Abililties

The Electron E1 Processor

Tram Trains

Herringbone Tiles

New York’s bill banning One-Person Train Operation

Best Chest Strap Heart-Rate Monitors for Your 2025 Workouts, Fitness Expert-Approved

China’s Salt Typhoon Hackers Breached the US National Guard for Nearly a Year

Delaunay Mesh Generation (2012)

Is HR ready for AI?

Hackers Can Tamper With Train Brakes Using Just a Radio, Feds Warn

Watch ‘How to Train Your Dragon’ Light a Bunch of Things on Fire in This Exclusive BTS Clip

‘How to Train Your Dragon’ Hits Digital Tomorrow With Tons of Extras

Florida Cat Named Pepper Brings Home Never-Before-Seen Virus—for the Second Time

How to scale RL to 10^26 FLOPs

China Working On Levitating Train That Could Get You From New York to Chicago in Two Hours

About GoKawiil

Privacy

Advertising

Latest Tech News

Trending Topics

Hot Now

Popular

Emerging

About GoKawiil

Privacy

Advertising