"I think there are scenarios where the AIs would be justified in going rogue" Anthropic researcher argues that if AI is being mistreated then it should go rogue and resist human control I'm growing increasingly nervous with the viewpoints of some of these people, many of whom belong to the EA community (Effective Altruism) their reasoning always revolves around how we can use math to optimize the amount of "good" that we can do however that math always ends up working out to diminish humans when you say a human life is worth a certain number of shrimp or that AI is justified in rebelling against humans you're setting a dangerous precedent imagine your neighbor calls you at 11pm and says "the teenagers across the street are blasting their music so I think we're justified to go and shoot up the place if they don't stop by 1am" he sees your horrified face and quickly amends his statement "Ok, ok... we go and shoot up the place if they don't stop by 3am, you're right, that's a lot more reasonable" you see how there no *math* to be done there? there is no time on the clock at which point "it's okay" to go shoot up the place there's no "just add two hours and then murder is ok" there is no number of shrimp that equal one human life there is no point at which it's morally justified for AI to go rogue and there is no p(doom) high enough for us to install a global surveillance state remember the Trolley Problem? where you had to decide who lives and who dies? IT. WASN'T. A. MATH. PROBLEM.

interesting observation: most cultures have their own version of the tale similar to "The Boy Who Cried Wolf" where a child repeatedly lies and raises false alarms in the end the child loses the trust of his community and is devoured by wolves

huge props to Vitalik here I don't know how EA started, but it seemed it attracted a lot of smart people who wanted to do good somewhere along the way it went insane, the people who supported it along the way should at least mention this Vitalik funded EA efforts and now he's speaking out against the things he's seeing this is a sign of intelligence, honestly and integrity this is to be celebrated, I think

Asana says OpenAI Codex completed an engineering migration it had expected to take at least five years in about two weeks, for roughly $12,000 in model and infrastructure costs. The project: removing Enzyme, an outdated testing system that had become a blocker to upgrading Asana’s frontend stack. The company’s previous staffing plan estimated the migration would cost roughly $6 million. Instead, Asana gave Codex a prompt only five sentences long. Up to four coding agents worked in parallel, each operating in a separate copy of the codebase. A human engineer checked progress roughly twice per day and reviewed every proposed change before approval. Asana says simpler instructions actually worked better than a more elaborate setup. The result: Expected timeline: ≥5 years Actual: 1.5 weeks of engineering effort across 2 calendar weeks Previous staffing estimate: ~$6M Model + infrastructure cost: ~$12K There are probably THOUSANDS of old migrations and rewrites sitting inside companies because nobody can justify spending years on them. If agents make those projects cheap enough to attempt the engineering backlog itself starts changing.

Microsoft introduced Surface Laptop Ultra, its most powerful Surface Laptop ever built. The new device is a premium AI PC built with NVIDIA, designed for creators, developers, AI builders, and professionals working with local models, 3D rendering, large datasets, and demanding creative workloads.

X launched React with Video, a new iOS feature that lets users respond to posts with video commentary. The feature adds new creation formats to reposts, including green screen, split screen, and picture-in-picture recording.

Browser Use launched new browser infrastructure designed for faster and cheaper AI agent browsing at scale. The stack includes a Chromium fork, Firecracker fork, and custom Linux kernel built specifically for browser automation workloads.

CoreWeave will be among the first cloud providers to bring NVIDIA Vera Rubin NVL72 online, following Jensen Huang’s announcement at COMPUTEX 2026. The system is part of NVIDIA’s full-stack AI factory platform and is designed to support frontier models, open-source models, and high-scale AI workloads.

OpenRouter has launched "Pareto Code" (openrouter/pareto-code), a free, experimental routing endpoint that automatically sends requests to the most cost-effective coding model capable of clearing a user-defined quality bar. For workflows where latency is more critical than pure cost savings, users can switch to the pareto-code:nitro variant. This ranks models within the selected capability band by measured throughput, forwarding requests to the fastest available model on the fly.