Reading Time: 5 minutesOpenAI and Broadcom have unveiled Jalapeño, OpenAI’s first custom AI inference chip, built in just nine months and already running GPT-5.3-Codex-Spark internally. If its performance-per-watt claims hold up, it could dramatically lower the cost of running ChatGPT and give OpenAI a vertically integrated hardware advantage over rivals still renting Nvidia GPUs.
Reading Time: 4 minutesClaude Code 2.1.191 ships a sweeping set of reliability, safety, and performance improvements — including permanent agent stop controls, conversation rewind, 37% lower CPU usage during streaming, and guardrails blocking destructive commands in auto mode. Non-technical professionals benefit most from understanding the governance and oversight changes this release introduces.
Reading Time: 5 minutesOpenAI has unveiled Jalapeño, its first custom AI inference chip developed in partnership with Broadcom, designed to power large language models like ChatGPT and Codex. The move marks OpenAI’s entry into custom silicon, mirroring strategies already adopted by Google, Amazon, and Meta to reduce dependence on NVIDIA and cut inference costs at scale.
Reading Time: 5 minutesByteDance unveiled Seedance 2.5 at its Volcano Engine conference — a video AI model that generates 30-second clips natively, benchmarked against Google Veo 3 at a fraction of the cost. Alongside a new language model, image model, and audio model launched in the same keynote, the announcement signals how rapidly the gap between Chinese and Western AI labs is closing.
Reading Time: 5 minutesAnthropic launched Claude Tag, a Slack-native agent that joins teams as a persistent, asynchronous member capable of monitoring metrics, summarising threads, and executing long-running tasks. It is currently in beta for Claude Enterprise and Team plan subscribers only, with meaningful open questions around memory partitioning, ambient mode accountability, and configuration complexity for non-technical teams.
Reading Time: 6 minutesClaude Code 2.1.186 introduces safer autonomous agent behavior by blocking destructive commands, adds CLI-based MCP authentication, and fixes over 60 bugs including startup slowdowns and file-write failures on network drives. The update makes multi-agent workflows more transparent and less risky, though the tool still requires technical setup and the agent teams redesign remains experimental.
Reading Time: 5 minutesSimon Willison used Claude Code to convert the Moebius 0.2B inpainting model from a Python-only tool into a fully browser-based app running on WebGPU — without writing a single line of code himself. The exercise reveals how Claude Code can handle model conversion, cloud deployment, and UI building autonomously, pointing toward a future where AI image tools run locally in a browser tab with no server required.
Reading Time: 5 minutesOpenAI’s o3 model helped Boston Children’s Hospital analyze 376 undiagnosed patient genomes and identify 18 previously unknown rare diseases. Reported by Mindstream, the breakthrough shows how advanced AI reasoning can accelerate diagnosis for conditions that have eluded specialists for years.
Reading Time: 6 minutesGoogle DeepMind has published an AI Control Roadmap that treats powerful internal agents as insider threats, borrowing cybersecurity frameworks to layer runtime monitoring, access controls, and AI supervisors on top of model alignment. The Neuron reports that DeepMind anchors the roadmap in concrete operational metrics — coverage, recall, and time-to-response — signalling that AI safety is evolving from a research discipline into an engineering operations practice.
Reading Time: 5 minutesAnthropic paused its most powerful AI models — Fable 5 and Mythos 5 — just four days after launch after US authorities raised national security concerns, forcing a blanket global access shutdown rather than a targeted geographic restriction. The episode exposes a critical gap between frontier AI deployment speed and the governance infrastructure needed to manage it responsibly.