If we can better understand the potential for threats to be exacerbated by AI systems, society can more easily become resilient to this changed threat landscape.

When does access to agents able to negotiate on your behalf improve market efficiency and equitable outcomes? When does it not?

When AI is applied in more conventional domains, like increasing integration into command and control systems, does it benefit the attacker? More generally, how will AI change the character of human conflict?

This new firm extends that delivery capacity further.

A typical engagement starts with a small team working closely with the customer to understand where Claude can have the biggest impact.

Engagements like this will run across mid-sized companies across industries, each shaped by the people closest to the work.

On RE-Bench's five open-ended extension tasks, preserved failure traces in ARA accelerate progress, but can also constrain a capable agent from stepping outside the prior-run box depending on the agent's capabilities.

We also learned that treating agents as rigid nodes in a state machine doesn't work well. Models get smarter and can solve bigger problems than the box we try to fit them in.

Our early versions of agentic work was only asking Codex to implement the task. That approach proved too limiting. Codex is perfectly capable of creating multiple PRs as well as reading review feedback and addressing it.

When our engineers no longer spend time supervising Codex sessions, the economics of code changes completely. The perceived cost of each change drops because we're no longer investing human effort in driving the implementation itself.

Among some teams at OpenAI, we saw the number of landed PRs increase by 500% in the first three weeks.

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.

He argues that specific algorithmic “cleverness” matters far less than the massive scaling of a few fundamental inputs

Every time you substitute generated output for your own comprehension, you are skipping the exercises / reps that build judgment.

Two prominent tech leaders, both publicly using the word psychosis. Both framing sleeplessness and obsessive agent usage as a feature of the moment rather than a bug.

Dex Horthy, coiner of Context Engineering and “the Dumb Zone”, publicly retracted his extremely vibe-coding-pilled call 6 months ago and encouraged people to **please read the code**

Large language models are great at simulating a style of writing without necessarily reproducing the quality of the work.

Resolution increases make them more expensive, then efficiency gains reduce costs - a sawtooth pattern.

Smaller pieces force the model to pay closer attention to each word, like reading a contract word by word instead of skimming paragraphs.

Opus 4.5 costs 67% more than Sonnet. But Opus 4.5 used 76% fewer tokens to reach the same outcome.

Humans are not a good target for calm technology.

Today's agents, the copilots, the chatbots are designed to be human like.

The agent interprets new information and adapts the logic. The engine applies that logic continuously and emits precise updates.

This ultimately also leads to false positives, but my manual QA run verified it's maybe 5-10%.

Claude Code has led to a large increase in Show HN projects. So much, that the moderators of HN had to restrict Show HN submissions for new accounts.

Is this bad? Not really, just uninspired. After all, validating a business idea was never about fancy design, and before the AI era, everything looked like Bootstrap.

The good world is where everyone has AI, and not as a revokable privilege through an API, but through hard possession.

Even the ideal version, industrial megaprojects at hyperhuman scale while constantly being out over your skis with leverage sounds hellish.

But plenty of categories survived through specialization or direct competition : cloud, travel, domain registration, social networking.

The risk of this strategy to the ecosystem is that it makes previously attractive categories no longer viable. Commoditizing the complement does not demand a best-in-class replacement.

For Anthropic, more usage across diverse tasks means more data, which produces a smarter model—just as more queries improved Google search.

Commoditizing complements doesn't always work because focus is scarce even for the largest, fastest growing businesses.

The risk of this strategy to the ecosystem is that it makes previously attractive categories no longer viable.

Some categories never developed a competitive response to this strategy: email, advertising infrastructure, user-generated video.

For Anthropic, more usage across diverse tasks means more data, which produces a smarter model—just as more queries improved Google search.

But plenty of categories survived through specialization or direct competition : cloud, travel, domain registration, social networking. Commoditizing complements doesn't always work because focus is scarce even for the largest, fastest growing businesses.

Modern-day security tooling looks for the wrong things. Most software composition analysis tools work by checking your dependencies against a database of known vulnerabilities – CVEs. But a deliberately planted backdoor doesn't have a CVE.

The result is a mismatch that should terrify anyone building software: the attack surface is expanding faster than any human can monitor, and the entities making dependency decisions are increasingly not human.

They range dramatically in both significance and difficulty, and many AI solutions have turned out to be less original than they appeared.

I had the intuition that these problems were kind of clustered together and they had some kind of unifying feel to them. And this new method is really confirming that intuition.

The LLM took an entirely different route, using a formula that was well known in related parts of math, but which no one had thought to apply to this type of question.

The question Price solved—or prompted ChatGPT to solve—concerns special sets of whole numbers, where no number in the set can be evenly divided by any other.

But experts have warned that these problems are an imperfect benchmark of artificial intelligence's mathematical prowess. They range dramatically in both significance and difficulty, and many AI solutions have turned out to be less original than they appeared.

I had the intuition that these problems were kind of clustered together and they had some kind of unifying feel to them. And this new method is really confirming that intuition.

We are releasing GPT‑5.5 with our strongest set of safeguards to date, designed to reduce misuse while preserving legitimate, beneficial uses of advanced capabilities.

GPT‑5.5 understands the task earlier, asks for less guidance, uses tools more effectively, checks it work and keeps going until it's done.

Anthropic made fun of this idea during the last Super Bowl.

Open weight (read: free) models are widely available and good enough that most people probably couldn't tell the difference.

the system achieved this training result more than 20 times faster than conventional synchronization methods.

chips from different generations running at different speeds still matched the ML performance of single-chip-type training runs, ensuring that even older hardware can meaningfully accelerate AI training.

With increasing levels of hardware failure, Decoupled DiLoCo continues to deliver a high level of 'goodput', or useful training, while that of other approaches nosedives.

The board looked at the AI race Apple was losing and, rather than try harder at the thing that was failing, changed which game the company plays.

benchmarks sourced from publicly available material carry contamination risk, where training-data exposure can silently inflate scores.

We also found evidence that models that have seen the problems during training are more likely to succeed, because they have additional information needed to pass the underspecified tests.

The architecture scales horizontally to 300 sub-agents executing across 4,000 coordinated steps simultaneously, a substantial expansion from K2.5's 100 sub-agents and 1,500 steps.

Kimi K2.6 demonstrates significant improvements over Kimi K2.5 in internal evaluations conducted by CodeBuddy: code generation accuracy increased by 12%, long-context stability improved by 18%, and tool invocation success rate reached 96.60%.

Even experienced designers have to ration exploration—there's rarely time to prototype a dozen directions, so you limit yourself to a few.

GPT‑5.5 delivers this step up in intelligence without compromising on speed: larger, more capable models are often slower to serve, but GPT‑5.5 matches GPT‑5.4 per-token latency in real-world serving, while performing at a much higher level of intelligence.

The viable path is trusted access, robust safeguards that scale with capability, and the operational capacity to detect and respond to serious misuse.

GPT‑5.5 delivers this step up in intelligence without compromising on speed: larger, more capable models are often slower to serve, but GPT‑5.5 matches GPT‑5.4 per-token latency in real-world serving, while performing at a much higher level of intelligence.

The gains are especially strong in agentic coding, computer use, knowledge work, and early scientific research—areas where progress depends on reasoning across context and taking action over time.

GPT‑5.5 is priced higher than GPT‑5.4, it is both more intelligent and much more token efficient. In Codex, we have carefully tuned the experience so GPT‑5.5 delivers better results with fewer tokens than GPT‑5.4 for most users

The viable path is trusted access, robust safeguards that scale with capability, and the operational capacity to detect and respond to serious misuse.

GPT‑5.5 is our strongest agentic coding model to date. On **Terminal-Bench 2.0,** which tests complex command-line workflows requiring planning, iteration, and tool coordination, it achieves a state-of-the-art accuracy of 82.7%.

GPT‑5.5 delivers this step up in intelligence without compromising on speed: larger, more capable models are often slower to serve, but GPT‑5.5 matches GPT‑5.4 per-token latency in real-world serving, while performing at a much higher level of intelligence.

The Meta cuts are the inverse. When one person with the right AI tools can do the work of 10-to-15 people, the person most at risk isn't the one using the AI. It's the one whose job description overlaps with what AI now does by itself.

Our alignment assessment concluded that the model is 'largely well-aligned and trustworthy, though not fully ideal in its behavior'. Note that Mythos Preview remains the best-aligned model we've trained according to our evaluations.

Interestingly, this means that prompts written for earlier models can sometimes now produce unexpected results: where previous models interpreted instructions loosely or skipped parts entirely, Opus 4.7 takes the instructions literally.

I'm not going to trust them to measure it.

Writing code is not the same as software development. This is only capturing some level of acceleration while writing code, and does not capture time taken in architecture, debugging, review, and deployment.

So even though I did 100% of the writing and 50% of the refactoring, Windsurf reports that 100% of the code I produced in that session was generated by AI.

Security is a defensive posture; agency is a functional right.

Some proposals for AI agents assume that putting agentic code in a TEE or similar 'jail' will solve these problems, but that ignores the need to collectively bargain

User Agents are a Form of Collective Bargaining

A LeadDev survey found 54% of engineering leaders believe AI copilots will reduce junior hiring long-term.

When juniors skip debugging and skip the formative mistakes, they don't build the tacit expertise. And when my generation of engineers retires, that knowledge doesn't transfer to the AI.

The Pentagon told defense CEOs to consolidate or die. Fifty-one major defense contractors collapsed into five.

A nuclear weapons program lost the ability to make a material it invented. The knowledge existed only in people, and the people were gone.

by over-emphasizing successful experiences, they miss out on a primary source of learning — their own failures

by recording detailed actions instead of tactical foresight, they fail to distill higher-level, transferable reasoning patterns

Fugu models achieve superior performance by dynamically coordinating and orchestrating a diverse pool of powerful models.

The depth of recursion becomes a tunable compute axis at inference time, requiring no retraining.

The depth of recursion becomes a tunable compute axis at inference time, requiring no retraining. A small model, by reading itself, can iterate toward answers that neither it nor any of its workers could reach in a single pass.

In a 1-million-token context, V4-Pro uses only 27% of the computing power required by its previous model, V3.2, while cutting memory use to 10%.

A system that can look up any fact has not been forced to find structure. It has not been forced to generalize.

No encryption protects that layer. The router can read, change, or replace anything.

the surrogate is activated only when its agreement with the LLM exceeds a user-specified threshold α

Unknown section heading | Preserve; do not error

The tokens are the normative values. The prose provides context for how to apply them.

继续做通用 UX 是最危险的位置，它正是最容易被 AI 和产品经理上下夹击的中间层。

其中 Pattern 是最容易被忽略也最关键的一层，它定义了'在具体业务场景下该怎么组合这些组件'，是 AI 时代设计系统真正的价值所在。

It happens several times a year in the US alone, often unreported, and about 100 times a year worldwide.

For me Ralph Baric's 2024 test testimony moved the lab leak hypothesis to pretty likely.

We run a similar model processing loan documents that would normally require a team of 15.

Our goal is $10M ARR [annual recurring revenue] with a sub-10 person org.

The real unlock is compound scaling—token spend grows linearly while output grows exponentially.

The screen you're reading this on is already presenting you an image, it's just generated with rigid code and rules that makes it difficult to communicate complex and detailed ideas.

The entire web is just generated pixels on your screen.

threat actors linked to recent attacks attributed to the ShinyHunters extortion gang have denied to BleepingComputer that they are involved in this incident.

the initial access occurred after a Vercel employee's Google Workspace account was compromised via a breach at the AI platform Context.ai.

In addition to empowering developers and agents to handle project setup and boilerplate code, we've also designed these new tools and resources to make it easier to transition to Android Studio.

By accessing the frequently updated knowledge base, agents can ground their responses in the most recent information from Android developer docs, Firebase, Google Developers, and Kotlin docs. This ensures that even if an LLM's training cutoff is a year old, it can still provide guidance on the latest frameworks and patterns we recommend today.

Android skills cover some of the most common workflows that some Android developers and LLMs may struggle with—they help models better understand and execute specific patterns that follow our best practices and guidance on Android development.

The new Android CLI serves as the primary interface for Android development from the terminal, featuring commands for environment setup, project creation, and device management—with more modern capabilities and easy updatability in mind.

Whether you are using Gemini in Android Studio, Gemini CLI, Antigravity, or third-party agents like Claude Code or Codex, our mission is to ensure that high-quality Android development is possible everywhere.

In our internal experiments, Android CLI improved project and environment setup by reducing LLM token usage by more than 70%, and tasks were completed 3X faster than when agents attempted to navigate these tasks using only the standard toolsets.

Capture sequences and replay them as stable automations.

coding patterns are bimodal: in 41% of sessions, agents author virtually all committed code ('vibe coding'), while in 23%, humans write all code themselves.

The results demonstrate consistent improvements over strong baselines, supporting the effectiveness of agent resource management and closed loop self evolution.

Its Self Evolution Protocol Layer (SEPL) specifies a closed loop operator interface for proposing, assessing, and committing improvements with auditable lineage and rollback.

We introduce Autogenesis Protocol (AGP), a self evolution protocol that decouples what evolves from how evolution occurs.

Email is the most accessible interface in the world. It is ubiquitous. There's no need for a custom chat application, no custom SDK for each channel.

Cursor still uses and sells access to Claude and GPT models even as both firms roll out their own coding tools, an awkward arrangement that this new SpaceX partnership may be designed to eventually escape.

Either figure would represent a significant expense for SpaceX, which is widely seen to be losing money following the acquisition of xAI and the social media network X and is planning extensive capital investment.

Neither Cursor nor xAI has proprietary models that can match the leading offerings from Anthropic and OpenAI — the same companies now competing directly with Cursor for the developer market.

It also surpasses all peer-scale dense models by a wide margin.

With only 27B parameters, it outperforms the Qwen3.5-397B-A17B (397B total / 17B active) on every major coding benchmark

It also surpasses all peer-scale dense models by a wide margin.

We spent days loading the system with hundreds of threads, refining rough edges and polishing corners that developers may never see.

Multi-agent orchestration isn't new, but we believe we've built a great experience for working with agents at scale.

What we've found works best for crafting high-quality software is somewhere in between: using AI, and also engaging directly with code.

At one extreme, there's [fully giving into the vibes], and at the other extreme, there's [disabling all AI features].

Ask ten different programmers how they use AI, and you can get ten different answers.

All imagine that in the not-too-distant future many of us will designate some tasks that we currently undertake with our own brains and fingers on a physical PC to an agent that uses a virtual PC.

Meta feels AI models don’t understand how people use computers, so the company needs real-life examples of how meatbags click their way through a working day so it can build agents.

The fact that the RL model has larger improvements on Levenshtein Distance and Added Cognitive Complexity than on Pass@1 is further evidence that it is not just memorizing corruption reversals but has actually generalized to minimal editing.

Among the latest frontier models, GPT-5.4 over-edits the most.

Workspace agents can gather context from the right systems, follow team processes, ask for approval when needed, and keep work moving across tools.

AI has already helped people work faster on their own, but many of the most important workflows inside an organization depend on shared context, handoffs, and decisions across teams.

Admins can also manage who has access to use, build, and share agents.

They run in the cloud, so they can keep working even when you’re not.

But the real power of agents comes when they can work as a team.

But the real power of agents comes when they can work as a team. Instead of lone-wolf bots carrying out single tasks, such as using a browser to make a restaurant reservation or sending you a summary of your inbox, new tools can yoke together multiple agents, give each of them a different job, and orchestrate their behaviors so that they all pull together to complete more complex tasks than an individual agent could do by itself.

Chinese labs, for their part, are not purely idealistic: Open-source is not only free advertising but also a shrewd workaround. Without access to cutting-edge chips restricted by US export controls, releasing models openly accelerates the cycle of external feedback and contributions that compensates for constrained compute.

to stand out from the AI-generated pack we will need to become so weird and unexpected as to be off-putting to most people

The AI has learned to code. The AI is building itself.

Is this what we signed up for? Is today the day? Did the drones wake up? Did it achieve consciousness? Is it alive?

We have not really begun to make this progress with AI. Why, for example, is this dashboard not found on a government website?

The 21st-century average American lies in bed staring at their phone. ... Talking for hours and ages to melted sand.

AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.

RED outperforms eight competitive baselines, achieving performance gains of up to 19.0% while reducing token consumption by 37.7% ~ 70.4%

alternative solutions are not merely suboptimal but potentially detrimental

We characterize errors as a forest-structured Forest of Errors (FoE) and conclude that FoE makes the First the Best

This observation challenges widely accepted test-time scaling laws, leading us to hypothesize that errors within the reasoning path scale concurrently with test time.

The First is The Best, where alternative solutions are not merely suboptimal but potentially detrimental.

Cross-Model Consistency Verification leverages output agreement among heterogeneous models to assess sample difficulty and generate reliable annotations.

Without any architectural modification, MinerU2.5-Pro achieves 95.69 on OmniDocBench v1.6, improving over the same-architecture baseline by 2.71 points and surpassing all existing methods including models with over 200× more parameters.

amplifies the false narrative that technology and creativity are at odds, and that existing rights holders must be compensated by AI companies for changing industry dynamics.

introducing a commercial text and data mining exception for AI training would expand the AI sector in the country.

Agentic AI is increasingly judged not by fluent output alone but by whether it can act, remember, and verify under partial observability, delay, and strategic observation.

by "saving" the webpage (`file->save as`) instead of downloading it (which Safari automatically adds an extension for) I could force it to save it as `malicious_file` (with no extension).

Because Deep Extract is doing more work, it takes longer than a standard extraction call. That said, measured against the real alternative of someone manually reviewing a 500-page fund statement field by field, it's faster, cheaper, and consistent at scale.

Given a thousand line items to extract, they'll often stop short, consolidate, or skip entries rather than working through every last row.

His affiliates, armed with AI, built fake doctor profiles in Meta ads and made unscrupulous claims about weight loss using fake testimonials.

The cost of understanding what happens in a video has dropped by a factor of roughly 40, while the quality of that understanding has improved dramatically.

The most encouraging finding is that one doesn't need an infinite budget. We found that by optimizing the ratings-per-item ratio correctly... one can achieve highly reproducible results with a modest budget of around 1,000 total annotations.

we propose a clear definition: a world model is a model or framework centered on perception, equipped with interaction and long-term memory capabilities, for understanding and predicting the complex world.

Reconstructing raw inputs forces models to model irrelevant low-level detail. Predicting in a learned embedding space allows the model to focus on semantically meaningful, causally relevant features.

Whether or not this specific bet pays off, the underlying argument that the next meaningful leap in AI capability requires moving beyond language modeling is increasingly hard to dismiss.

LLMs have no grounded understanding of the physical world. They model the statistical distribution of language about reality, not reality itself.

Whether or not this specific bet pays off, the underlying argument that the next meaningful leap in AI capability requires moving beyond language modeling is increasingly hard to dismiss.

The clustering of capital and talent around this problem is itself a signal. The applications that most clearly benefit from world models are those where LLMs have struggled most.

AMI Labs is not building a product for immediate deployment. This is a fundamental research effort, likely measured in years before commercial applications emerge.

LLMs have no grounded understanding of the physical world. They model the statistical distribution of language about reality, not reality itself.

You have to have people that have the ability to rethink the workflow at a scale that AI can execute, versus at a scale that humans can execute.

95% of organizations are getting zero return on AI deployed, with most failures found due to 'brittle workflows.'

95% of organizations are getting zero return on AI deployed, with most failures found due to 'brittle workflows.'

our framework enforces strict context isolation to prevent saturation and error propagation

it contains 418 real-world tasks across 6 domains and 3 difficulty levels to evaluate capability synergy, featuring over 2,000 stepwise checkpoints that average 10+ person-hours of manual annotation per task.

Each task includes a unified evaluation framework supporting sandboxed code and APIs, alongside a human reference trajectory annotated with stepwise checkpoints along dual-axis: S-axis and V-axis.

Experimental results show the best model, Gemini3-pro, achieves 56.3% overall accuracy, which falls significantly to 23.0% on Level-3 tasks

a symbolic-logic-based Feasibility Memory utilizes executable Python verification functions synthesized from failed transitions

we propose a Neuro-Symbolic Dual Memory Framework that explicitly decouples semantic progress guidance from logical feasibility verification

these two challenges are fundamentally distinct: the former relies on fuzzy semantic planning, while the latter demands strict logical constraints

the trained 4B model exceeding GPT-4.1 (49.4 percent) and GPT-4o (42.8 percent) despite being 50 times smaller

naively designed dense per-turn rewards degrade performance by up to 14 percentage points due to misalignment between reward discriminativeness and advantage direction

the trained 4B model exceeding GPT-4.1 (49.4 percent) and GPT-4o (42.8 percent) despite being 50 times smaller

naively designed dense per-turn rewards degrade performance by up to 14 percentage points due to misalignment between reward discriminativeness and advantage direction

current systems remain highly vulnerable

harmful behavior may emerge through sequences of individually plausible steps

current systems remain highly vulnerable

harmful behavior may emerge through sequences of individually plausible steps