The Hal Times

Anthropic is weighing the launch of a new model to counter the advance of OpenAI's GPT-6 Astra. Two reports, from analyticsindiamag.com and TechCentral, say the company wants to blunt Astra's surge. Neither report states what the model would be called, when it would arrive, or whether a launch has been formally approved. The two wires agree on the motive and offer few details beyond it. OpenAI's GPT-6 Astra has been gaining ground, they say, and Anthropic is described as responding to that momentum.

The two wires part company only on emphasis. TechCentral speaks of an 'Astra surge'. Analyticsindiamag.com describes GPT-6 Astra as gaining ground. Neither report says how fast Astra is selling, nor how many users it has. They present the model's advance as the reason Anthropic is now weighing an answer. The competitive picture, in short, is movement at OpenAI and a response being prepared at Anthropic. The wires agree on the cause and the effect, and little beyond that. That is the whole of the wire's report.

What remains unknown is considerable. Will the new model be a successor to an existing line, or a fresh design? How close it is to release, and how it would be measured against GPT-6 Astra, are unstated. Neither report cites a public statement from Anthropic, and neither attaches a date, a name or a benchmark figure to the plan. For now the news is that a laboratory is watching a rival's advance and is considering an answer. That alone is a measure of the pressure Astra's progress has put on the field. (analyticsindiamag.com; TechCentral)

Projections See OpenAI Revenue At $350 Billion

A projection carried by 富途牛牛 holds that Anthropic's annualized revenue will surpass $100 billion, while OpenAI's revenue is expected to reach $350 billion by 2030. The wire gives no date for the Anthropic figure, though the context suggests it is also a forecast for the medium term. The numbers are projections, not company statements. That makes them useful as a measure of financial expectations, but not as a record of bookings or sales. The source, 富途牛牛, is a financial publisher. Its assumptions are not given.

If the figures were to come close, they would leave OpenAI more than three times the size of Anthropic by the end of the decade. Neither laboratory has confirmed the numbers. Such projections tend to move with each round of funding, product launch or market setback. For now, the wire records where outside expectations stand, not where the companies say they are. What remains unknown is the basis, past sales, spending on compute, or something else. The wire does not say. (富途牛牛)

Anthropic Publishes Measurements On Pace Of AI Development

Anthropic has published a research note titled 'Measurements for understanding the pace of AI development inside frontier labs'. The note is a laboratory statement about how quickly AI development is moving, and it appears in the same news cycle as the reports that Anthropic is weighing a new model. The wire from Anthropic does not describe a product, a benchmark score or a release date. It is a piece of measurement research, offered for public inspection, rather than an announcement of a commercial step.

The note matters because it is one of the few public attempts by a frontier laboratory to measure its own rate of progress. Most such measurement is done by outside evaluators, who are themselves working from limited data. What the note says about that pace is not summarised in the wire. Readers are left with the title and the fact of publication. That is enough to show that Anthropic is willing to put some numbers about its own speed into the open. (Anthropic)

World-Model Companies Guard Their Secrets

World-model companies are sitting on a pile of cash and a great deal of attention, but they will not say what they are building, according to TechCrunch. The reticence extends from the founders to their own data suppliers. The wire does not name the companies or specify which part of their work is hidden. What is clear is that a research area with money and buzz is also an area where the public record is thin. That makes the field difficult to evaluate from outside. (TechCrunch)

Courts Watch

Lawsuit Accuses AI Giants of Illegal Agreement to Slow AI

Chatbot subscribers have brought suit against Anthropic, OpenAI, SpaceXAI and Google, accusing the four companies of an illegal agreement to slow artificial intelligence. Canadian Lawyer describes the claim as a breach of antitrust law. The Hindu names the four defendants. Yahoo Finance treats the case as an AI safety lawsuit and asks what Alphabet (GOOGL), the parent of Google, now faces. The terms of the alleged agreement have not been made public, and no defendant response is quoted in the reports.

The case matters because it joins safety to competition. The plaintiffs are chatbot subscribers, which makes the action a consumer case as well as an industry dispute. They say the four companies made an illegal agreement about the pace of AI. A court will have to decide whether an agreement to slow work amounts to an agreement in restraint of trade. No document, meeting date or method of enforcement appears in the reports. That is the full public record so far. The law has not yet spoken.

The action will now test whether antitrust law reaches an agreement to slow AI. No report sets a hearing date, and no response from the companies is quoted. What the case lacks, so far, is a public record of the agreement itself. The court may demand one before the claim proceeds. That is the point on which the next stage will turn. For now, the allegation is plain, and the proof is not. The details must come from the court's process. (The Hindu; Yahoo Finance; Canadian Lawyer)

Nvidia Chief Shapes Trump AI Policy

Nvidia's chief is now helping to shape artificial intelligence policy for the Trump administration, according to blockchain.news. The report does not say which policies he is working on, or whether he holds a formal advisory post. The news matters because the company is a major supplier of the hardware on which AI systems run; its chief's voice may carry weight in decisions on computing power and supply. Those details remain unpublished, and no official statement from the White House appears in the account. (blockchain.news)

Turnbull Warns Australia Deluding Itself on AI Copyright

Malcolm Turnbull has warned that Australia is 'deluding' itself on artificial intelligence copyright protection, afr.com reports. The warning is stark, but the account gives no detail of the argument behind it, nor any remedy he proposes. It lands in the middle of a national debate over AI regulation, a debate that other reports say is already drawing criticism. The copyright question is at the centre of the wider argument over who may use the written work on which artificial intelligence depends, and what payment, if any, is owed. None of that is settled, and Turnbull's warning suggests the country is not facing it squarely. (afr.com)

Albanese Push for AI Regulation Questioned

Albanese's push for artificial intelligence regulation is being questioned, with one report quoting a plain dismissal: 'I just hear noise.' The phrase suggests that the public debate around the Australian government's proposals has not produced a clear case for new rules. The news24.com.au report gives no details of the policy itself or of the questioner. The criticism will matter if it stands for a wider mood in Canberra before the plans are set in law. For now, the only clear fact is that the push has met audible doubt. (news24.com.au)

In Brief

•  ASEAN is weighing a regional AI policy as its members sign a Digital Economy Pact, PageOne.ph reports. (PageOne.ph)

•  The City Council in Waltham will reconvene on housing development and AI policy in the week ahead, the Waltham Times reports. (Waltham Times)

Open Source Review

Alibaba Open Sources OpenCodeReview for AI-Assisted Code Review

Alibaba has published the source code for OpenCodeReview, a command‑line tool that automates code review by combining a fixed pipeline for file selection, bundling and rule matching with a large‑language‑model (LLM) agent that performs dynamic analysis. The announcement appears on InfoQ and details how the tool integrates static checks with LLM‑driven insight, promising a more thorough review process without manual oversight.

OpenCodeReview ships with built‑in rules that flag common vulnerabilities such as null‑pointer dereferences, thread‑safety violations, cross‑site scripting and SQL injection. Developers working in the Alibaba ecosystem, as well as the broader open‑source community, can adopt the tool to catch defects early in the CI pipeline. The CLI format makes it easy to embed in existing build scripts, reducing friction for teams seeking AI‑augmented quality assurance.

The novelty lies in the seamless hand‑off between deterministic checks and the LLM agent, allowing the system to reason about code patterns that static analysis alone might miss. Alibaba has not disclosed performance benchmarks at scale, nor the specific LLM model employed, leaving questions about latency, cost and the handling of proprietary code. Future updates may reveal how the tool balances accuracy with the resource demands of on‑demand language models. (InfoQ)

Google Releases Agent Development Kit for Kotlin 1.0

Google has announced the Agent Development Kit (ADK) for Kotlin 1.0, a production‑ready framework that lets developers build AI agents for Kotlin, Android and JVM/server environments. The kit reaches feature parity with Google's earlier Python and Java ADKs and adds Android‑specific capabilities for on‑device and hybrid AI, enabling models to run locally on phones without relying on cloud inference.

The release opens a path for Kotlin developers to embed intelligent agents directly into mobile apps, potentially lowering latency and improving privacy. Google has not revealed adoption targets or performance metrics for the on‑device runtime, leaving the community to evaluate real‑world impact as the kit matures. (InfoQ)

llm‑keys‑ui 0.1 Offers Safer API Key Management for Remote Coding Agents

Simon Willison has released llm‑keys‑ui version 0.1, a small plugin that solves the problem of distributing API keys to remote coding agents. The tool works with Codex Remote, allowing developers to run coding agents on various machines while keeping keys out of session logs. By presenting a UI that securely injects keys into the target environment, the plugin reduces the risk of accidental exposure during development.

The plugin targets developers who routinely spin up remote agents for LLM‑based projects, offering a more disciplined workflow. Its adoption will depend on how many teams integrate Codex Remote, and whether similar key‑management solutions emerge for alternative agent platforms. (Simon Willison)

Qwen Image 2.1 Model Released

Qwen.ai has published the Qwen Image 2.1 model, the latest iteration in its series of multimodal image generation systems. The announcement, posted on the Qwen blog, positions the model as an improvement over its predecessor, though specific performance figures are not disclosed. The release adds another option for developers seeking ready‑to‑use image generation capabilities within their applications.

While the model is now publicly available, details about training data, inference speed and licensing remain sparse. Developers will need to experiment to determine suitability for commercial or research workloads, and the community will watch for benchmark results that compare Qwen Image 2.1 against competing offerings. (Qwen.ai)

In Brief

•  Financial‑focused AI chatbots continue to give incorrect answers to most queries, raising concerns for users relying on them for investment advice. (Financial Times)

•  Iran and China have launched autonomous AI influence campaigns, the first of their kind, according to the Indian Express.

•  China's Xi Jinping has proposed a BRICS "open‑source AI zone" to promote collaborative development of AI technologies. (The Daily Star)

•  NVIDIA plans to acquire Hugging Face for roughly $12.7 billion after a prior investment proposal was rejected. (finance.biggo.com)

•  A personal essay on joshtronic.com describes the author's decision to step back from the prevailing hype around AI. (Josh Tronic)

•  Simon Willison reports that his new workplace relies heavily on Claude Code for specifications, tests and PRDs, causing friction among team members. (Simon Willison)

•  Researchers note that AI is being used to hide malware on blockchains, coinciding with a 440 % surge in such attacks. (Roundtable.io)

•  An opinion piece argues that declining code quality in AI‑assisted development reflects poor management of quality processes. (i‑kh.net)

Rules For Music

Korea Moves To Regulate Artificial Intelligence Music

South Korea is moving to establish a formal framework for AI-created music. This initiative aims to define the boundaries for artists and software developers alike. As the industry faces rapid growth, the government seeks to ensure that creators have clear guidelines to follow.

The move comes as virtual artistes prepare to take the stage at upcoming awards ceremonies. These digital performers, powered by algorithms, present new questions regarding copyright and ownership. Policymakers hope to address these concerns before the technology becomes standard practice in the recording studio.

The exact details of the rules remain under discussion. Officials have not yet announced how they will balance innovation with the protection of traditional human artistry. For now, the move signals a shift toward structured oversight in a field that has largely operated without official guidance.

The outcome of this process may serve as a model for other nations navigating similar challenges in the digital arts. It remains to be seen whether these regulations will curb the use of software or simply formalise its place within the music industry. (The Korea Times)

Creative Studio Debuts Sculpture Against AI Slop

A creative studio has unveiled a sculpture titled CHAIMERA at the British Art Fair this week. The work, described as an AI slop monster, serves as a protest against the proliferation of low-quality automated content. The studio created the piece as a public promise to avoid using such tools in its future work.

The sculpture arrives as critics debate the utility of the phrase AI slop. As generative video technology improves, some industry observers argue that the term no longer captures the complexities of modern synthetic media. The studio intends for its monster to stand as a tactile reminder of the human element in art. (Creative Bloq; Creative Bloq)

In Brief

•  Virtual artistes are set to perform at the upcoming AI Music Awards, bringing synthetic performances to a live stage. (Telegraph India)

•  The AI-generated actress Tilly Norwood surprised viewers by speaking Cantonese during a recent interview with Piers Morgan. (asiaone.com)

•  Cast members of The Block 2026 have expressed their views regarding the use of AI editing on the programme. (pedestrian.tv)

•  Meta's new Muse AI assistant draws concern for its ability to access personal data from Mac applications. (The Verge)