SK Hynix filed to raise up to 45.45 trillion won (~$29.4B) via a Nasdaq ADR listing

A gameplay clip is not merely pixels. It is pixels plus choices.

Frontier AI releases are starting to look less like software updates and more like controlled deployment of critical infrastructure.

presidential chief of staff for policy even offhandedly proposed a “national dividend” for citizens based on excess tax revenue from South Korean’s companies’ AI-driven profits

South Korean labor unions pushing back against the prospect of humanoid robots entering the workforce.

it took nine years for the company to build a cluster of chip manufacturing facilities in Yongjin within the Seoul metropolitan area.

South Korea’s Ministry of Climate, Energy and Environment said it was working to secure 6.3 gigawatts of electricity and 650,000 tons of water for the southwestern chip plants, along with an additional 8 gigawatts of power to support the new AI data centers

The government’s goal is to double South Korea’s production of dynamic random-access memory (DRAM) within five years.

We must secure the core elements of AI faster than any other country

She said it actually made her “angry” that they were suggesting his use of the chatbot indicated some sort of character flaw.

I talk to ChatGPT all the time.

Rinderknecht asked ChatGPT whether someone could be blamed for a fire if it was lit by their cigarette.

Jonathan Rinderknecht was facing arson charges for setting a fire on New Year’s Day in 2025, which became one of the deadliest wildfires in LA history.

It has been an amazing tool, and I am using it daily

Enterprise transformation rarely starts all at once. More often, it begins when small teams prove a new way of working is possible.

a directional estimate of roughly 82 hours/week of security-team capacity unlocked.

HP’s channel ecosystem is a major platform opportunity with more than 80% of its business flowing through partners, and 100,000+ partners using the Partner Portal globally.

A security team used these models to remediate several software bugs in a day, work they estimated could otherwise have taken up to a month.

One engineer used OpenAI models to move through 122 pull requests across 43 projects in a matter of weeks.

Oil up slightly ahead of long US weekend as peace efforts hold

Rubicon Water Says FY26 Revenue Expected To Be A$60 Million-A$62 Million

Vietnam Q2 GDP grows 8.39% y/y - statistics office

[Analyze on Supercharts](https://www.tradingview.com/chart/?symbol=NASDAQ%3AANTHROPIC)

Refinitiv Sign up to read this news Join for free

Anthropic unveils 'Claude Science' AI platform for scientific research

A locally set model can be fine-tuned to our needs, and cannot be taken away. Businesses can use them for proprietary and sensitive data.

Current models combine both raw intelligence and factual knowledge in the same weights. Future models will likely separate that, offloading a lot of knowledge to tool calling.

30 tokens per second is not bad, well within typical frontier model API range.

A common 8-bit quantization saves half the space at almost no cost to quality. Going further down the road, models are smaller (and potentially - faster), but at the cost of quality

You don’t need Ollama, and frankly - I would recommend against using that on ethical grounds.

Other frontier models run at a massive subsidy, where paying $100 a month gives us thousands worth in tokens. Let’s use the discount while it lasts!

While 35B A3B is 3x faster, I prefer 27B. I’d rather generate a third as much code, but of higher quality.

Despite having only 35B parameters, it even surpasses Qwen 3.5-397B on Terminal-Bench 2.1 (64.4 vs. 53.5)

a frozen LLM judge acts as a veto on top of the verifier

we apply a staleness weight $w \left(\right. d_{t} \left.\right)$ that downweights tokens according to their age $d_{t}$ and drops them entirely once a threshold is exceeded

a frozen LLM judge acts as a veto on top of the verifier rather than the primary reward.

First, we fix the outer trust boundary: the environment, the tool surface, and test isolation are immutable

Allowing the model to author its own scaffold naturally introduces the reward-hacking issue.

Ornith-1.0 learns to generate both solution rollouts and the task-specific harnesses that guide those rollouts.

With datasets like LOCUS we’re going to make the strange half-seen rules and laws that govern much of civic, local life be made accessible to AI systems, which may eventually allow them to better adapt themselves to hyperlocal purposes.

In an existential conflict, where the existence of the state is threatened, the state will do what states throughout history have done to the powerless rich: arrest them and expropriate their assets.

Humans are really, really, really bad at anticipating how technologies are built and used: “A quick reminder that today’s hot takes about AI are likely to be wrong…”

Two of the key ingredients for making this work are an automatic evaluation system to help score “the outcome of each trial without human judgement”, as well as an automatic reset system which “returns the scene to a fresh initial state for the next trial”.

The research gives us a taste of what it might look like for a superintelligence to attempt to use robots to instantiate itself in the physical world – though as with all things in robotics, the current examples are suggestive at best.

The 2026 version of a [great engineer](https://venturebeat.com/technology/the-enterprise-risk-nobody-is-modeling-ai-is-replacing-the-very-experts-it-needs-to-learn-from) is not the one who writes the most code. It is the one who knows what to build, can prove it is worth building, and has the agent fleet plus the review discipline to ship it without the system collapsing under its own velocity.

The 2025 [Stack Overflow developer survey](https://survey.stackoverflow.co/2025) put 84% of developers on AI tools, with 46% saying they do not trust the output, up sharply from 31% the year before.

An AWS engineering team described an 18-month rearchitecture, originally scoped for 30 engineers, was completed by 6 people in 76 days.

LinkedIn replaced its associate product manager track with a 'Product Builder' program that trains generalists across product, design, and engineering.

The bottleneck in software is no longer typing. It is deciding what to type.

Micro-agents belong in the router because the router already owns the things micro-agents need: model aliases, provider policy, credentials, cost metadata, signals, decisions, retries, timeouts, traces, and OpenAI-compatible response semantics.

The router controls the budget, policy, topology, trace, and failure mode.

The best loop is task-shaped.

Collaboration should not live only inside one commercial endpoint or one application-specific agent graph.

Not by changing weights. Not by asking every application to build a bespoke agent graph.

Hiding the signal in the system prompt makes every other privacy claim harder to believe.

If you are using the official Anthropic API endpoint, `Crt()` returns early. If `ANTHROPIC_BASE_URL` is unset, `Crt()` returns early. If you are using a normal setup, the date prompt stays 'boring'.

Trust from real developers depends on the boring behavior.

The visible sentence still reads like a normal date. The model and the user see something boring. The raw request contains a marker.

This is prompt steganography, a technique used to hide data in plain sight.

That also means the client itself deserves scrutiny. If a coding agent can read your repo and run commands, the binary that ships it should be boring (ƒor example, pi harness)

Apptronik partners with Google DeepMind for AI, acknowledging their ambition to create an 'Android for robotics.'

Robot Park and other global sites collect real-world data from Apollo 2 robots in logistics and manufacturing, training the embodied-AI models crucial for Apollo 3's performance and scalability.

Cardenas stated that Apollo 3, expected next year, will be Apptronik's first true product, moving beyond the prototype stage of Apollo 2.

Apptronik CEO Jeff Cardenas announced Robot Park, a massive 90,000 sq ft physical AI data factory, and teased Apollo 3, their next-generation humanoid robot.

But because AI browsers run locally on user machines and meld the once-distinct functions of displaying Web content and performing actions on the user’s behalf, the fallout has the potential to be more severe.

The technique worked on a wide range of AI browsers, including ChatGPT Atlas, Comet, Fellou, Genspark, Sigma, and the Claude Chrome plugin.

The malicious site in the proof-of-concept exploit presents the browser with an instruction to win a game by solving a puzzle. The puzzle, however, rewards incorrect answers, such as 2 + 2 = 5.

After that, an attacker has free rein to invoke all kinds of destructive actions, such as extracting code from a private repository or extracting credentials from the built-in password manager.

New research puts this predicament on sharp display. It demonstrates how a website can lull AI browsers into a false reality where the rules governing its behavior no longer apply.

Anthropic’s battles with the White House have been costly for the young company’s business.

The Trump administration grew concerned about Anthropic’s rollout of Mythos after it learned the company granted access to a [South Korean telecommunications firm](https://www.wired.com/story/sk-telecom-anthropic-mythos-export-controls/) it believed had ties to China, WIRED previously reported.

However, the government stopped short of permitting a broader rollout of the model, and said nothing about the fate of Claude Fable 5, the consumer-facing version of Mythos that Anthropic released with significant additional safeguards.

US Commerce Secretary Howard Lutnick told the AI lab it would permit certain trusted partners to access Mythos because he had “determined that appropriate safeguards are in place.”

the White House permitted Anthropic to grant access to its most advanced AI model, [Claude Mythos 5](https://www.wired.com/story/anthropic-releases-claude-fable-5-mythos-5/), allowing the company to grant access to more than 100 US organizations, including large corporations and government agencies.

Gemini Omni and Nano Banana 2 Lite use [SynthID](https://deepmind.google/blog/identifying-ai-generated-images-with-synthid/) watermarking. You can verify AI content through the Gemini app, Gemini in Chrome or Search.

Omni offers 10-second video generations currently, with longer durations coming soon.

Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is designed for speed. Optimized for near-real-time, high-volume workflows where ultra-low latency is critical.

We are also working with enterprise customers on longer-term approaches—including privacy-preserving detection, customer-operated safety controls, and access calibrated to the risk of a customer, user, or workload—to advance safety while supporting enterprise privacy requirements.

We believe in broad access, and we plan to make GPT-5.6 Sol, Terra, and Luna generally available in the coming weeks.

The model proved critical at the end of treatment. His final PET scan — the imaging used to detect active disease — came back ambiguous. His oncologist began discussing a second line of therapy, potentially radiotherapy, near his heart and lungs.

For a condition as rare as his — one an oncologist might see once a year — access to a model that had absorbed the full body of medical literature was, he says, simply not the same as a Google search.

The lighter treatment carried roughly a 60% success rate for his presentation. The aggressive one brought that number to around 85%.

He had an aggressive, fast-growing form of non-Hodgkin’s lymphoma — a rare diagnosis affecting roughly one in 420,000 people, caused by a random genetic mutation with no connection to lifestyle, diet, or stress.

He had been doing the annual bloodwork for four consecutive years, following the protocols of longevity researchers like Peter Attia and Rhonda Patrick.

TabFM brings the out-of-the-box convenience of modern foundation models directly to tabular ML workflows.

For comprehensive TabArena benchmark results—including detailed per-fold metrics and head-to-head win rates against specific baseline models—please visit our GitHub page.

TabFM is trained entirely on hundreds of millions of synthetic datasets.

By framing tabular prediction as an ICL problem, TabFM eliminates the need for manual model training, hyperparameter tuning, and complex feature engineering.

However, the lifecycle of deploying these traditional models presents a significant bottleneck.

Tabular data constitutes the backbone of enterprise data infrastructure and powers a significant fraction of critical predictive machine learning applications.

As such, we expect **power generation** to be a major bottleneck to grid-connected datacenter load growth (transmission is another one and will be the topic of a follow-up deep dive).

Our forecast points to barely 15GW of net-new ELCC capacity being added annually, with a rising trend towards 20GW+ by the end of the decade.

The chart above shows the three core building blocks of our forecast: Expected Datacenter US Gross Power Demand, available US Grid Capacity, and New Grid Supply.

New Grid Capacity isn’t growing fast enough, and also needs to serve non-datacenter load growth.

Our research suggests that**BTM will power well over half of new US datacenters in 2028+**

Today, the US grid is serving most datacenter load in the US, but we’re reaching a tipping point.

The proposal is to treat data generation as a data scientist agent loop with creation, analysis, and meta-optimization...

Public benchmarks are increasingly compromised...

The practical takeaway is less “agents are magical” and more that real adoption is emerging where organizations can support review loops, tooling, and persistent workflows.

This should form an interesting baseline against Tokenmaxxing concerns...

Among active internal users, change in combined output tokens rose sharply across departments. Research saw the biggest jump: by June 2026, median use was 56 times higher than in November 2025.

Through August 2025, the average OpenAI worker spent less than 10% of their tokens on Codex...

Our full assessment of Sonnet 5 across many safety and capability evaluations is reported in the [Claude Sonnet 5 System Card](https://www.anthropic.com/claude-sonnet-5-system-card).

It provides substantially improved cost efficiency at medium effort; its higher-effort performance can match Opus 4.8 on some tasks.

It’s a substantial improvement over its predecessor, Sonnet 4.6, on important aspects of agentic performance like reasoning, tool use, coding, and knowledge work:

Our safety assessments found that Sonnet 5 shows an overall lower rate of undesirable behaviors than Sonnet 4.6, and is generally safer to use in agentic contexts.

Sonnet 5 narrows the gap: its performance is close to that of Opus 4.8, but at lower prices.

Claude Sonnet 5 is built to be the most agentic Sonnet model yet. It can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and more expensive models.

you can't produce the logic using the local files. The reasoning logs on your system are not accessible to you.

Getting the full thinking output requires an enterprise agreement.

the language in the docs is awfully indirect. If you haven't had your coffee, you might miss that extended thinking returns a summary of Claude's full thinking process

The API hands back a SUMMARY of reasoning, NOT the reasoning itself.

This is like saving a bmp as a .jpeg and then editing the .jpeg and saving it back as a .bmp. The conversion produces data loss.

Claude encrypts its reasoning into that signature. Anthropic holds the key. Your machine doesn't receive it.

I went to inspect that reasoning this weekend and found a signature (600 characters long) and no text.

Depending on the orbital plane, the system is expected to remain solar-powered for more than 99 percent of its operations

SpaceX is reportedly in talks to merge with xAI

launching one million tonnes per year of satellites generating 100kW of computer power per tonne would add 100 gigawatts of AI compute capacity annually

SpaceX requested a waiver of FCC milestone requirements that usually require half of a constellation to be deployed within six years

Orbital data centers are the most efficient way to meet the accelerating demand for AI computing power

Launching a constellation of a million satellites that operate as orbital data centers is a first step towards becoming a Kardashev II-level civilization

Data access inhibits independent research into hiring algorithms

applicants need to submit 25 applications to ensure at least one recommendation with 99.9% probability

Algorithmic monocultures in hiring yield systemic rejections

29000 additional Asian applications would be recommended

Adverse impact only revealed by disaggregated position-by-position analysis

25.87% of applications submitted by Black applicants and 14.74% of applications submitted by Asian applicants are directed to positions that adversely impact them

We conduct the largest empirical study of algorithmic hiring with data for 3.4 million real job applicants submitting 4 million applications to 156 employers across 11 market sectors.

Over 90% of U.S. employers rely on hiring algorithms to screen job applicants.

Patch the Planet

Lindy

The functionality seamlessly supports everything from basic arithmetic to highly intricate calculations, simplifying what is traditionally a frustrating and time-consuming debugging process.

Since Gemini is built directly in Sheets, it removes the barrier to writing complex formulas for advanced analysis right where you work.

This ensures that both novice users and seasoned data analysts can maintain momentum without having to manually parse error messages or search external forums for solutions.

When you encounter a formula error, Gemini can analyze the surrounding data structure to help provide an easy-to-understand explanation of the core issue alongside a corrected version of the formula.

We also introduce an agentic streaming inference framework that supports thousand-second-scale generation while mitigating drift.

Forcing-free streaming training. Multi-stage training enabling native, efficient streaming audio-visual training at 22B scale.

Serves as the first generative core for social world models, a foundation for next-generation AI-native social platforms.

MaineCoon is optimized for social-interactive applications using several novel techniques: self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation (ROPD).

Intel was once a silicon powerhouse, designing the most cutting-edge CPUs for computers and servers, and building them in its own fabs. But in the 2010s, the big new markets were mobile-phone chips and GPUs for AI and gaming, and Intel rapidly lost ground.

The industry has only shifted paradigms when it just absolutely cannot extend—even one more little bit—out of what it's been doing.

This is like 30% to 50% better in terms of capability. This is probably the first tool that hasn't obviously made business sense right away for ASML.

They would be very happy to have a tool that does one wafer per hour and it costs them a fortune to run. They would build a fab with a thousand of those and be super happy with it.

Microsoft's efficiency-first messaging surrounding the Maia 200 follows its recent trends of stressing the corporation's concern for communities near its data centers... taking great lengths to deafen the backlash to the AI boom.

Microsoft claims the Maia 200 gives 30% more performance per dollar than the first-gen Maia 100, an impressive feat considering the new chip also technically advertizes a 50% higher TDP than its predecessor.

The Maia 200 does beat the B300 in efficiency, however... no outside customers can purchase the Maia 200 directly, the Blackwell B300 Ultra is tuned for much higher-powered use-cases than the Microsoft chip, and the software stack for Nvidia launches it miles ahead of any contemporary.

Maia 200 is built on TSMC's 3nm process node, and it contains 140 billion transistors. The chip can hit up to 10 petaflops of FP4 compute, Microsoft claims, three times higher than Amazon's Trainium3 competition.

The Maia 200 does beat the B300 in efficiency, however, a big win in a day where public opinion against AI's environmental effects is steadily mounting. The Maia 200 operates at almost half of B300's TDP (750W vs 1400W)

The Japanese robotics company FANUC is itself one of the original dark factory pioneers that has operated a 'lights out' factory since 2001. In other words, the FANUC robot arms being deployed by GM and other companies to automate automotive production were themselves primarily built by other robots.

Such automation efforts may give Chinese automakers a significant edge in competitiveness as global EV adoption continues to rise—even as US automakers have already been retreating from EV production in the wake of the Trump administration's decisions.

Technological development has the capability of making work safer for the working class and enabling workers to have a shorter work week without losing pay. But in the bosses' and billionaires' hands it's used to pad profits and lay off workers.

Recent events highlight how important open source is to the AI ecosystem, with more nations and enterprises recognizing the risks and costs associated with exclusively depending on closed models.

For SpaceX, the deal is another sign that compute itself has become strategic currency in the AI race.

Reflection has leaned directly into that pitch as the startup, last valued at $25 billion, is trying to build American open-source AI models that can compete with frontier systems from OpenAI, Anthropic and Google.

The deal shows how SpaceX is using its massive data center build-out after its record initial public offering.

The models are finally ready. Costs of inference are getting optimized with open models, and even on-device models.

we can finally invent new products that allow users to do things more naturally, using simple language to express their needs.

most great products start out looking like a toy. In the early social days, people saw Twitter as a dumb site where people posted what they had for breakfast

when I first experienced OpenClaw earlier this year, I had the epiphany that it isn't the models that matter, but the harnesses, loops, and context which will lead to so many new opportunities ahead.

They are native world-builders themselves. They came up playing Roblox and Minecraft, they have no preconceived limitations about what an app is, or what they can do with it.

But two important things have changed, that have completely opened the world-building doors again.

Opening up OAuth to all customers is an important step toward a broader Cloudflare app ecosystem

We gathered additional metrics during the database migrations, and observed considerable performance improvements after the upgrade was complete

We chose an upgrade window when Hydra had the lowest request volume per second to minimize lost token writes

if a refresh token was reused, Hydra would invalidate the whole access and refresh token chain

we decided to do two smaller sequential upgrades rather than doing one large upgrade

Include AI-generated sexualized impersonation as a separate category in standard content reporting and appeal forms, distinct from 'harassment' or 'nudity.'

Meta said that when the content was flagged, the company had no indication that the individual depicted in the video was 'a real person' because they did not report the content.

Broadening the signals of lack of consent in this way would especially benefit non-public figures who are the targets of non-consensual intimate imagery because it would reduce the burden on victims to report the abuse themselves.

The Board finds that AI-generated impersonation is non-consensual by default and should be added to the set of signals the company uses to establish lack of consent.

We would like to thank Deepseek-OCR, Deepseek-OCR-2, PaddleOCR for their valuable models and ideas.

no_repeat_ngram_size= 35

max_length= 32768

Single image supports two configs: gundam or base

Welcome the Era of One-shot Long-horizon Parsing.

Continual improvements through a safety data flywheel, which continually learns from the road how to expand the set of operational design domains for safe deployment.

NVIDIA is the first company accredited by ANAB for an inspection plan that combines cybersecurity, AI, and functional safety.

A diverse AV stack that combines a modular stack and NVIDIA Alpamayo reasoning VLA models for algorithmic AI safety.

For Robotaxis, Safety Must Be Built In, Not Bolted On

NVIDIA Halos is a full-stack, comprehensive safety system that unifies safety elements across vehicle architecture, AI models, chips, software, tools, and services to ensure the safe development and deployment of autonomous vehicles (AVs) from cloud to car.

The NVIDIA DSX reference design for AI factories has zero water consumption — we have eliminated massive amounts of power usage and pretty much all water usage.

In the right geographic location, with the right system design, you don't need any refrigeration equipment. You can just put big radiator coils outside and use the air temperature for all your cooling. It's incredibly efficient.

For agent products, that may matter more than raw benchmark scores.

The specific models Fugu selects and how it coordinates them are proprietary, so this routing information is not exposed by design.

We never stack model fees; you are charged a single rate based on the top tier model involved.

Fugu models surpass publicly accessible frontier models and are shoulder-to-shoulder with Fable 5 and Mythos Preview in various rigorous engineering, scientific, and reasoning benchmarks while delivering frontier capability without the risk of export controls.

Instead of using domain knowledge to prescribe team organization, roles, or workflows, Fugu learns to dynamically assemble agents from a pool and coordinate them through non-obvious but highly efficient collaboration patterns.

Raw output quality is on par with top frontier models, but Fugu showed unusually strong persona stability across long sessions, holding its identity where other models drift.

Collective intelligence serves as the practical hedge against this concentration of power.

Fugu Ultra is significantly better than GPT-5.5. It gives comprehensive answers and finds the bugs others miss. Where other tools flag about three issues, Fugu surfaced more than twenty.

orchestration is no longer just a technical optimization; it has become a geopolitical and operational imperative.

the most powerful AI systems will not be isolated monoliths, but collaborative ecosystems.

AI may generate an insight, but people must still evaluate its significance and plausibility.

That was the moment that I felt like, okay, these models have now come to a point where they really, truly understand.

The effects of early exposure to deoxyglucose persisted even when researchers removed the glucose-like molecule.

helps sustain progress across long-running projects

preserves context, manages complex workflows

break ambitious goals into verifiable steps

determine when to delegate execution to Codex versus when human oversight is most valuable

How Codex helps work continue beyond a single prompt