OpenAI选择砍掉视频应用，把算力集中到GPT-5.5的Agent架构和Codex代码工具上

Ilya Sutskever的SSI获20亿美元融资押注新范式，Yann LeCun离职Meta创办AMI Labs，融资10.3亿美元，估值35亿。

20亿参数对比同体量自回归模型、千亿参数LLaDA2.0，连续路线的scaling曲线健康有效。

ELF用Flow Matching完成生成，仅32个采样步生成质量就超过离散模型1024步结果

训练数据约450亿token，仅为主流方法的十分之一。

KPMG and UT Austin's research helps clarify what that human should be doing

KPMG becomes a preferred consultant for deploying Claude and Anthropic's agents into those portfolio companies

Anthropic raises $65B in Series H funding at $965B post-money valuation

Building an AI agent to help clients adjust to changing tax regulations used to take weeks and required teams to switch between multiple tools and chat windows

every one of KPMG's 276,000+ employees globally will gain access to Claude

KPMG—one of the world's largest professional services firms for audit, tax, legal, and advisory services across 138 countries and territories

Currently, the US only fully manufactures about 10 percent of the chips it requires

Tech giants collectively plan to spend $750 billion on AI infrastructure this year, with "a significant portion" of that expected to "go towards chips for data centers"

Four years ago, five years ago, Nvidia was spending about 10, 15 billion dollars a year in Taiwan. Now we're spending 100, going to 150 billion dollars in Taiwan each year.

Nvidia will invest $150 billion a year to make Taiwan an AI "epicenter."

Adoption differences extend beyond discipline and career stage. We classify researcher names according to gender and find that those with typically male names have adopted coding agents at more than twice the rate of respondents with typically female names.

Claude Code is the most common coding agent tool reported, with 86% of users reporting Claude Code use (31% report using Codex, the next most common tool).

On a 1 to 10 scale, 88% of respondents were above a 5, and half were at 8 or above. Figure 6 shows that these ratings vary strongly with AI use. The left side of the plot shows researchers that use AI for more types of tasks are more optimistic.

Coding agent users are starting projects at a pace of around a quarter of a paper more and posting around a half of a working paper more than non agent users. In percentage terms, coding agent users look around 10% (empirical projects started) to 75% (working papers posted) more productive than others in their discipline and career stage.

There are sharp disparities in use of coding agents. Twice as many researchers with typically male names use coding agents as those with female names. Researchers at top universities are 40% more likely than others to use coding agents.

The vast majority of respondents (81%) have tried using AI chatbots in research, particularly for writing code and editing prose. But only 20% have adopted coding agents—tools like Claude Code that autonomously write and execute analysis code—into their work.

We present results from a survey of 1,260 social scientists about AI and coding agent use, fielded in February and March 2026.

The time from business to production workflow drops from months to days.

McKinsey predicts that by 2030, three-quarters of current jobs will require redesign, upskilling, or redeployment

Although 85% of organizations say they want to be agentic within the next three years, 76% say their current operations and infrastructure can't support that change.

the unemployment rate for recent college graduates rose to 5.6%, while the underemployment rate (the share of graduates working in jobs that typically do not require a college degree) reached 42.5%, its highest level since the covid pandemic

workers aged 22 to 25 in the most AI-exposed occupations experienced a 16% relative decline in employment after the spread of generative AI

the unemployment rate for recent college graduates rose to 5.6%, while the underemployment rate (the share of graduates working in jobs that typically do not require a college degree) reached 42.5%

workers aged 22 to 25 in the most AI-exposed occupations experienced a 16% relative decline in employment after the spread of generative AI

Vibe drafts the deliverable using the Canvas tool, from a one-page brief to a report, an RFP response, or a board deck

Sessions can run in parallel, can persist while your machine is off, and can be triggered from third-party apps, such as Slack (coming in June)

Mistral Vibe extension for VS Code; the coding agent working across your whole project, inside your IDE.

Team, $24.99/user/month: a shared workspace with admin controls and more storage.

Pro, $14.99/month: complex tasks, deeper reasoning, and all-day coding.

Over the last 20 years the fastest-growing occupation in the US was manicurists and pedicurists. But following close behind? Compliance Officers.

annual employment growth for coders has slowed significantly—by about 3%—since the introduction of ChatGPT

16% decline in entry-level jobs in AI-exposed occupations

a little over 40% of workers but adoption varies by sectors

US Census data showing that only one in five companies are using AI in any business function.

Crete practitioners prepare tens of thousands of tax returns each season which requires working through millions of underlying documents.

Over the past six months, OpenAI forward deployed engineers and researchers along with Thrive Holdings' engineers collaborated to build Tax AI

One senior accountant who spent 180 hours on tax prep last year spent only 15 hours on it this year.

Rental properties took about six weeks and substantial engineering oversight to reach 90% precision and recall

At launch, only a quarter of returns were at 75% correct field completion, but within six weeks, 86% hit that mark.

Even today, colonialism assumes new forms. It no longer dominates only bodies, but appropriates data, transforming personal lives into exploitable information.

In practical terms, in the age of AI and robotics, ensuring that the economy favors human dignity means adopting certain criteria for firm action. First, transparency and accountability: when data and algorithms influence credit distribution, personnel selection or access to services and opportunities, it is necessary that decisions be understandable, contestable and subject to oversight, so that individuals are not reduced to mere profiles. Second, inclusion and access: the benefits of innovation must be paired with investments in skills, infrastructure and essential services to ensure that technology does not widen the gap between those who have and those who have not. Finally, measures to ensure equity: taxation, social protection and industrial policies must correct the imbalances created by the concentration of wealth and power. Indeed, these criteria do not constitute a curb on innovation; instead they make it civilized and humane.

Moreover, ownership of data cannot be left solely in private hands but must be appropriately regulated. Data is the product of many contributors and should not be treated as something to be sold off or entrusted to a select few. It is necessary to think creatively in order to manage data as a common or shared good, in a spirit of participation, as Saint John Paul II already suggested regarding collective goods. [128]

It claims 8 million global users and 100 trillion tokens processed per month

after raising $40 million in Series A funding in June 2025

it landed at about $1.3 billion post-money

Claude Opus 4.7 has been used to patch over 2,100 vulnerabilities

on average, a high- or critical-severity bug found by Mythos Preview takes two weeks to patch

90.6% (1,587) have proved to be valid true positives, and 62.4% (1,094) were confirmed as either high- or critical-severity

Mythos Preview has found what it estimates are 6,202 high- or critical-severity vulnerabilities in these projects (out of 23,019 in total)

their rate of bug-finding has increased by more than a factor of ten

we and our approximately 50 partners have used Claude Mythos Preview to find more than ten thousand high- or critical-severity vulnerabilities

Claude Opus 4.7 has been used to patch over 2,100 vulnerabilities

on average, a high- or critical-severity bug found by Mythos Preview takes two weeks to patch

90.6% (1,587) have proved to be valid true positives, and 62.4% (1,094) were confirmed as either high- or critical-severity

Mythos Preview has found what it estimates are 6,202 high- or critical-severity vulnerabilities in these projects (out of 23,019 in total)

their rate of bug-finding has increased by more than a factor of ten

we and our approximately 50 partners have used Claude Mythos Preview to find more than ten thousand high- or critical-severity vulnerabilities

V4-Flash by default for cheap iteration; /pro lifts a single turn to V4-Pro

Node ≥ 22 on macOS / Linux / Windows

In long sessions the bill typically lands at ~1/3 of comparable generic tooling.

$0.07 /Mtok in · $0.014 /Mtok cached

long sessions hold 90%+ cache hit and input-token cost collapses to ~1/5

Perceptual BD-rates are based on human ratings from a large-scale subjective study

search over millions of model configurations to jointly optimize over perceptual quality and on-device runtime

Based on large-scale subjective user studies

faster than most top ML-based codecs run on a V100 GPU

on an iPhone 17 Pro Max, it encodes 12MP images as fast as 230ms, and decodes them in 150ms

error analysis identifies data-layer defects (e.g., incorrect query composition and ORM runtime violations) as the leading root causes.

the best data filter may be **no filter**, with projections suggesting the crossover for internet-scale pools lands around **1e30 FLOPs**

Another secondary summary gives Humanity’s Last Exam: 64.7% vs 53.1%, possibly under different setup/effort/tool conditions.

Domain-specific ECI scores can be used to compare performance relative to other model releases, but not to track the absolute performance or progress trends in different domains.

The SWE overperformance has been consistent across most generations, and remains in recent models.

The most extreme ratio observed is 4 math benchmarks to 2 SWE benchmarks.

All models included in our analysis have at least two scores in each domain, with an average of 3.2 SWE benchmark results and 3.4 math benchmark results.

Opus 4.6 and 4.7 both have Math-ECIs within 1 point of their general ECI, compared to larger gaps for earlier models.

On average Claude models have an SWE-ECI 2.7 points higher than their general ECI, and a Math-ECI 1.8 points lower.

We don't train on your data by default on our Team and Enterprise Plans.

We don't train on your data by default on our Team and Enterprise Plans.

We intend to publish our thinking and decision-making as we do

The first of these will be released publicly later this year

In sub-Saharan Africa and India, we are creating AI-powered apps that support foundational literacy and numeracy programs

PwC will roll out Claude Code and Cowork starting with U.S. teams and expanding toward a global workforce of hundreds of thousands of professionals, establish a joint Center of Excellence, and train and certify 30,000 PwC professionals on Claude

KPMG and Anthropic announce a global alliance, with Claude integrated into KPMG's Digital Gateway platform and available to all 276,000+ employees

the nearly two billion people whose incomes depend on smallholder farming

commit $200 million in grant funding, Claude usage credits, and technical support for programs in global health, life sciences, education, and economic mobility over the next four years

building toward full-scale deployment across its 167,000-person workforce

the $100 million investment we made this year to back the services firms helping enterprises actually deploy AI

more than 5,000 leaders saw the alliance up close, with hands-on training enabling a wave of early adopters

Security work that took hours now takes minutes

Insurance underwriting that took 10 weeks now takes 10 days

cutting delivery times by up to 70%

a program to train and certify 30,000 PwC professionals on Claude

PwC will roll out Claude Code and Cowork starting with U.S. teams and expanding toward a global workforce of hundreds of thousands of professionals

a drag that is estimated to be more than $2 trillion

AlphaEvolve has been used as a regular tool to optimize the design of the next generation of TPUs. It also helped discover more efficient cache replacement policies, achieving in two days what previously required a concerted, human-intensive effort spanning months.

AlphaEvolve began optimizing the lowest levels of hardware powering our AI stacks. It proposed a circuit design so counterintuitive yet efficient that it was integrated directly into the silicon of our next-generation TPUs.

This optimization reduced 'write amplification'—the ratio of data written to storage versus the original request—by 20%. It also provided insights for new compiler optimization strategies that reduced the storage footprint of software by nearly 9%.

achieving 10% accuracy gains over their competitive manual model optimizations

doubling its training speed whilst improving model quality

reduced 'write amplification'—the ratio of data written to storage versus the original request—by 20%

finding 10.4% improvement in routing efficiency over the previous heavily optimized solutions — saving over 15,000 kilometers of distance travelled annually.

suggesting quantum circuits with 10x lower error than previous conventionally optimized baselines

the overall accuracy of predicting the risk of natural disaster—aggregated across 20 categories such as wildfires, floods, and tornadoes—was increased by 5%.

increase the ability of our trained Graph Neural Network (GNN) model to find feasible solutions for the problem from 14% to over 88%

achieving a 30% reduction in variant detection errors.

achieving gold-medal-level performance on mathematical and physical olympiad competitions, including IMO 2025/USAMO 2026 and IPhO 2024/2025

Next-generation AI chips, such as Nvidia's Rubin, will shift to the 3nm process

of the roughly $30 billion year-over-year increase, around $20 billion came from HBM alone.

Total spending on components across the top four designers more than doubled from 2024 to 2025, rising from $22 billion to $52 billion.

The four designers consumed only ~11% of global leading-edge logic wafer capacity in 2024 and 2025.

The four designers still take roughly 80–85% of total CoWoS supply.

The top four designers collectively consumed nearly all of TSMC's CoWoS wafer output, leaving little headroom for other customers.

The main characteristic of how the data is transformed is that there will be a syntactic difference — in other words, very easy to see by the neural net — between most of the input statements, which will be tagged as 'communication acts.'

Collectively, this foundation represents an unmatched planetary-scale dataset for AI systems.

This integration marks the first time Earth AI imagery models have been deployed commercially against a dataset with the scale, accuracy, and temporal depth of Vantor's AI-ready spatial foundation.

Groundsource uses Gemini to analyze decades of public reports and identifies over 2.6 million historical flood events spanning more than 150 countries.

GPT-5.5 Pro still regularly gets my favorite GSM8K question wrong.

AI solutions were graded by the official judges, using the same criteria as were applied to human solutions.

software engineering tasks which may take humans weeks seem to be within reach for AI systems.

models climb close to the average human baseline over the past year and a half.

humans can do this in well under half an hour.

Top models scored around 40%.

When inference is expensive, teams limit usage, reduce context, or avoid certain applications altogether.

At 50 million tokens, the design space for AI applications changes fundamentally.

Subquadratic's team includes 11 PhD researchers and research engineers with backgrounds from Meta, Google, Oxford, Cambridge, ByteDance, Adobe and Microsoft.

Subquadratic has raised $29M in seed funding from investors including...

SubQ's research model performs on up to 12 million tokens, while other frontier models break down well before their stated 1M-token limit.

SWE-Bench Verified score of 81.8 compared to Opus 4.6 (80.8) and Deepseek 4.0 Pro (80.0).

Research result of 83 and a production model, third-party verified score of 65.9, SubQ 1M-Preview compares favorably with other SOTA models like Claude Opus 4.7 (32.2), GPT 5.5 (74), and Gemini 3.1 Pro (26.3).

SubQ Sparse Attention is 52× faster than FlashAttention in our architecture-level comparison, while requiring 63% less compute.

SubQ 1M-Preview scores 95% accuracy, compared to 94.8% for Claude Opus 4.6

With a research result at 12 million tokens, SubQ's architecture reduces attention compute by almost 1,000x compared to other frontier models.

compute requirements scale quadratically with context length

13K

1.3K

283 replies

2.5M Views

Apple accidentally left Claude.md files in today's Apple Support app update (v5.13)

19.3M Views

Leaders will own much more, with as many as 15+ direct reports

Over the past 13 years, we have weathered four crypto winters

We are flattening our org structure to 5 layers max below CEO/COO

US employees will receive a minimum of 16 weeks base pay (plus 2 weeks per year worked), their next equity vest, and 6 months of COBRA

reduce the size of Coinbase by ~14%

A Chinese court ruled that companies can't dump the costs of AI automation onto workers.

New Federal Reserve research confirms what private data already suggested, that AI is killing junior coding jobs first.

21 concrete protections drawn from 30+ studies on what AI does to your cognition.

The best AI models in the world score below 0.5% on ARC-AGI-3—is this what you call AGI, guys?

The price tag of the AI gold rush: $725 billion. Will it pay off?

Andrej Karpathy built a simple automation pipeline for AI agents to optimize training in 5-minute increments.

An engineer at Cloudflare used Claude with OpenCode to release vinext, a reimplementation of Next.js on Vite, for only ~$1,100 in API costs.

Nicholas Carlini at Anthropic tasked Claude with building a C compiler from scratch, spending roughly $20K in API costs.

Wilson Lin at Cursor coordinated hundreds of GPT-5.2 agents to build a web browser from scratch, running uninterrupted for one week. Over a million lines of Rust.

AI Village gives multiple AI agents their own computer environments and a shared group chat, then tasks them with open-ended real-world goals like fundraising, organizing events, making games, and gaining subscribers.

The volume of open-world evaluations has increased dramatically in recent months.

We plan to release new evaluations every 1–2 months.

Overall, it usually takes me about two hours to do this task. If only it were as simple as a single copy and paste, life would be so much easier — or so I thought.

For example, this could bring a five hour (300 minute) time horizon down to a three minute time horizon. But while the time horizons are much shorter, the growth rate is about the same as the METR's main results, with roughly two doublings each year.

By the end of the year, we expect AI to be able to do tasks roughly one day long with a 50% success rate. In comparison, I'd guess that this task would take several days for a person familiar with the paper and is able to play around with the web interface.

The benchmark tasks were meticulously constructed to be realistic, involving the hard work of hundreds of experts and likely millions of dollars — placing it among the most expensive economics papers of all time.

Researchers have been throwing tons of effort into getting better training data. For example, Surge AI had a revenue of over $1 billion last August, and Scale AI was probably in a similar boat.

The one real underlying asset, Workday's trillion-transaction dataset, is thinner than it sounds; what actually matters at runtime is how data connects to workflows, permissions, and integrations, and every layer of that stack is now a liability.

By late 2025, total AI data center power capacity had reached roughly tens of gigawatts, which puts AI's electricity consumption at a scale comparable to the peak electricity demand of the state of New York

Total AI computing capacity has been doubling approximately every seven months

Across leading AI companies where breakdowns are available, the chips and computing time to run them account for 54% to 62% of total spending

By the fourth quarter of 2025, the five largest chip designers had cumulatively shipped roughly 20 million AI chips

Stripe then sets a default limit of $100.00 USD/month as the maximum the agent can spend on any one provider.

Let’s say your product is a coding agent. You’d love for people to be able to take what they’ve built and get it deployed to production, using Cloudflare and other services.

We’ve seen remarkable adoption since its launch, with over 103,000 agents built and a total of more than 1.1 million agent sessions recorded

Military personnel and Defense Department civilians have used a version of Google Gemini’s [Agent Designer](https://docs.cloud.google.com/gemini/enterprise/docs/agent-designer) to create over 100,000 semi-autonomous AI agents in less than five weeks since the tool became available

We've spent five years building that surface area across Mac, Windows, and Linux, exceeding a million lines of code.

The feature can edit spreadsheets without a human-in-the-loop and was vulnerable to data exfiltration risks due to its ability to insert formulas that trigger external communication.

The average employee AI usage was 1.5 hours per week. The average CEO AI usage was less than one hour per week.

37,000 lines per day. And this was the output.

Worldwide IT spending is expected to reach $6.31 trillion in 2026, up 13.5% from 2025, according to Gartner.

IT budgets are getting blown out as some companies increasingly spend more on AI than on employees' salaries.

The Pentagon designated Anthropic a supply chain risk

some of the most reputable people in quantum hardware and quantum error-correction—people whose judgment I trust more than my own on those topics—are now telling me that a fault-tolerant quantum computer able to break deployed cryptosystems _ought_ to be possible by around 2029.

A search through GPT‑5.5’s SFT data found many datapoints containing “goblin” and “gremlin.”

When we looked, use of “goblin” in ChatGPT had risen by 175% after the launch of GPT‑5.1, while “gremlin” had risen by 52%.

As per The Information, Meta employees used a total of 60.2 trillion AI tokens (!!) in 30 days.

Claude Code used to be a feature of the $20/month Pro plan, but according to the new pricing page it is now exclusive to the $100/month or $200/month Max plans.

the numbers are mindboggling, they mostly serve to reinforce the sheer hardware advantage that a decade of investment has given to GDM and any models they train and serve.

Human advisors endorsed fraudulent investments at baseline rates of 13-14%, versus 0% across all LLMs, and suppressed warnings under pressure at two to four times the AI rate.

According to reporting from the _New York Times_ and the _Atlantic_, contract negotiations between Anthropic and the US Department of Defense fell apart in late February because Anthropic balked when the DOD demanded leeway to use the company’s models to analyze commercially available data on US citizens.

LLM agents could potentially do the work of intelligence analysts in a fraction of the time and for a fraction of the cost, which would enable the state to aim its all-seeing eye toward anyone, not just its highest-priority targets.

The release includes DeepSeek-V4-Pro (1.6T total / 49B active) and DeepSeek-V4-Flash (284B total / 13B active), both trained natively at 1M context length.

Claude skews high-income; Meta AI skews low-income

AI Startup Has Helped Reverse Thousands of Denied Health Insurance Claims

We will release all data, evaluation code, and model outputs to facilitate future research.

These papers suggest that strategic data engineering and inference-time optimization can substitute for raw parameter count.

The quality and structure of training data matters more than its volume.

This alignment ensures that human data seamlessly translates into enhanced action controllability for humanoid video generation.

By predicting these unified tokens, it effectively leverages diverse human data to achieve state-of-the-art data efficiency and robust out-of-distribution (OOD) generalization.

Scaling humanoid foundation models is bottlenecked by the scarcity of robotic data.