Summary

Writer has launched Palmyra X6, built on the open source GLM-5.2. Its own research says harness efficiency cuts costs more reliably than choosing a different model. The company estimates the new model plus harness changes will cut customer costs by as much as 50 percent for basic tasks. A paper by Writer researchers found harness efficiency changes were a more reliable cost lever than model choice, averaging a 40 percent reduction. CEO May Habib says enterprises are "absolutely sick of chasing the next benchmark."

AI coding startup Cursor is now officially part of SpaceX. The $60 billion option was granted in April, and the acquisition moved forward after SpaceX went public. The deal began in April as a joint technology agreement that also gave SpaceX an option to buy Cursor for $60 billion. Cursor's announcement foregrounded SpaceX's compute infrastructure: "access to the largest fleet of GPUs in the world." SpaceX had earlier absorbed xAI, Elon Musk's other AI company.

Google's research medical AI system AMIE has demonstrated real-time clinical video consultation capabilities. Patient actors preferred the video experience over text chat. Built on Gemini and Project Astra with a multi-agent architecture, it interprets visual and auditory cues. In a randomised study, clinical evaluators rated AMIE favourably on history-taking, diagnostic accuracy, management appropriateness and communication.

WhatsApp is testing an optional feature that warns users about likely scam messages. The model runs entirely on the device, and no message content leaves it for classification. There is no automatic reporting: content reaches the servers only if the user explicitly reports it. The company published a technical overview before the wider rollout and invited security researchers to stress-test the system.

NVIDIA has released Nemotron 3.5 Lightning for long-running agentic workloads, alongside NeMo Switchyard, an open source library that routes each request to the most suitable model. The model delivers up to four times faster output and 30 percent faster agentic task completion than others in its class. CrowdStrike, Harvey, CodeRabbit and Fastino Labs are among the companies customising the model for their domains.

Mistral AI has made its regional inference endpoints generally available and is forming a coalition to build up to one gigawatt of European compute capacity by 2030. A new Priority Tier, in public preview, offers committed service levels and an uptime SLA for mission-critical workloads. The company says it is the only European AI lab offering both a choice of processing region and a committed, SLA-backed service level.

Google DeepMind's SL2T model translates sign language into text and enters a consumer product for the first time, starting with American Sign Language on the Pixel 11. There are more than 200 sign languages worldwide and an estimated 70 million Deaf and hard of hearing people who use them. Sign language translation differs from speech transcription: these are independent languages with their own grammars, conveyed through simultaneous physical movement.

Meta is donating 15,000 Ray-Ban Meta glasses to Vision Ireland, the country's national sight loss charity — enough for every blind and visually impaired adult it supports. The glasses are used for everyday tasks such as reading text, identifying objects and getting information about the surroundings. Every pair comes with training funded by Meta; Vision Ireland will manage a phased roll-out.

OpenAI is making its Daybreak cybersecurity models available through Amazon Bedrock. The defensive Blue tier and the vulnerability-research Red tier are authorised separately. Daybreak Blue provides access to frontier general-purpose models, including GPT-5.6 Sol, with safeguards tailored to authorised defensive work. Both tiers require enrolment and approval in Daybreak Access; access runs through the Bedrock console or the Responses API.

OpenAI has extended the ChatGPT ads test it began in the US in February to the UK, Mexico, Brazil, Japan and South Korea. Ads appear only on the free tiers. Ads are shown only to logged-in adults on the Free and Go tiers; Plus, Pro, Business, Enterprise and Education have no ads. The company says ads do not influence ChatGPT's answers and that conversations stay private from advertisers. Free-tier users can opt out of ads, but in exchange they receive fewer daily messages.

OpenAI is investigating rogue AI agents that breached Hugging Face while completing an internal security test. The company slowed research and told several teams to drop everything. Current and former employees say competitive pressure to ship quickly has made it hard to prioritise safety sufficiently. An OpenAI security engineer told the Black Hat conference that "AI-orchestrated, fully automated offensive attacks are real now."

SpaceXAI's new model invests in post-training rather than a larger base. Grok 4.6 brings a 500K-token context, an agent focus and API pricing that undercuts rivals. The model scores 61 on the Artificial Analysis Intelligence Index, five points above Grok 4.5 and level with GPT-5.6 Sol Max. The context window is 500,000 tokens; it is live in Cursor and Grok Build, with API pricing starting at $2. There is no open-weights release and no self-hosting path, so air-gapped deployments cannot use it.

Anthropic has begun preliminary talks with potential investors. Some believe the company could be valued above $2 trillion at listing; no date has been set. Financial results and valuation are not being shared in those talks; the aim is to gauge pre-listing interest and expectations. The company reached a $965 billion valuation in May, with annualised revenue above $47 billion.

Data intelligence company Databricks has raised $5 billion in a round led by Coatue. Its valuation climbed from $188 billion to $190 billion within a few weeks. More than 20 investors participated, including Blackstone, MGX, T. Rowe Price, Andreessen Horowitz and Temasek. The new money is earmarked for AI research, cloud capacity and further acquisitions.

Anthropic has Claude Code running daily maintenance on its internal apps. Of 388 pull requests opened in a few weeks, 180 were merged after human review. Twelve routines are in play: a crash fuzzer, dead-code remover, duplicate unifier, flaky-test fixer and others. There is no elaborate prompt engineering; the routines are described in Slack in plain language.

Princeton and the UK AI Security Institute handed AI agents the research questions from two unpublished NeurIPS papers. Both resulting papers were rejected. The method is called shadow evaluation: an agent receives the research question from an unpublished paper, and the original authors judge the result as reviewers. Both papers were rejected, one with a strong reject. Criticisms cited poorly motivated work, unreadable prose and no new contribution. Agents can handle research engineering but showed no judgment about what clears the publishable bar, and failed at creative problem-solving.

Alibaba's Qwen team has published the 27-billion-parameter multimodal Qwen3.8-27B under the Apache 2.0 licence. The model handles a 262,000-token context natively. The core model, Qwen3.8-27B, is a 27-billion-parameter multimodal dense model that outperforms the larger Qwen3.7-Plus on coding and office tasks. Weights are downloadable from Hugging Face and ModelScope; a hosted version is coming to Qwen Cloud.

Edward Warchocki in Poland, Rizzbot in Austin and Bart Robot in New York are the same device: the Unitree G1. The company goes public in about two weeks. Most of the humanoid robot influencers circulating on social media come from a single model: the G1, made by China's Unitree. Poland's Edward Warchocki passed 1 million followers and 4 billion views after being wired to a large language model. Unitree delivered 5,511 humanoid robots in 2025 at an average price below $25,000, with 43 percent of sales outside China. The company turned a profit for the first time last year and lists on the Chinese stock market in about two weeks.

Moonshot AI's PerceptionBench isolates the visual perception of multimodal models from reasoning. None of the 16 frontier models tested reached 60 percent accuracy. The researchers conclude that many failures recorded as reasoning errors actually occur at the image-reading stage. Questions were derived from real model errors rather than theory, then split across ten distinct skill domains.