Morning Briefing
Signals, investigations, and field notes on the AI economy.
On July 10, 2026, South Korea’s SK Hynix started trading on Nasdaq (when-issued SKHYV, then SKHY), after pricing about $26.5 billion of American Depositary Receipts at $149—the largest ADR offering on record and a landmark U.S. listing for the HBM memory supplier that underpins modern AI GPUs.
On July 9, 2026—the same day GPT-5.6 went public—OpenAI evolved its GPT-5.5 Bio Bug Bounty into an ongoing private OpenAI Bio Bounty Program focused on universal jailbreaks against a predefined biosafety challenge, raised the top award from $25,000 to $50,000, and put GPT-5.6 in scope alongside GPT-5.5.
On July 9, 2026, Anthropic launched a beta “Reflect” experience in Claude settings—usage summaries, topic patterns, quiet hours, break nudges, and AI Fluency coaching—for Free, Pro, and Max users with Memory on, framed as intentional AI use rather than pure engagement growth.
On July 9, 2026, Anthropic appointed former Federal Reserve Chair and Nobel laureate Ben Bernanke to its Long-Term Benefit Trust—the independent body that can appoint Anthropic board members and is charged with holding the Public Benefit Corporation to its long-term public mission as AI’s economic effects scale.
On July 9, 2026, Meta Superintelligence Labs shipped Muse Spark 1.1—its strongest agentic and coding model yet—plus a public preview of the Meta Model API for developers. The multimodal upgrade targets tool use, computer use, coding, and long-horizon agents, available in Thinking mode on Meta AI and meta.ai.
Alongside the July 9 GPT-5.6 public unlock, OpenAI shipped ChatGPT Work—an agent for multi-hour knowledge-work jobs with plugins across Drive, Slack, Teams, and more—and folded the standalone Codex desktop app into a unified ChatGPT desktop for macOS and Windows that combines Chat, Work, and Codex.
On July 9, 2026, OpenAI moved GPT-5.6 Sol, Terra, and Luna from a government-coordinated trusted-partner preview (started June 26) to broader public availability after U.S. officials cleared a wider rollout. The three-tier family—Sol flagship, Terra balanced (~2× cheaper vs GPT-5.5-class positioning), and Luna low-cost—had been gated at the administration’s request.
On July 8, 2026, OpenAI published an audit finding roughly 30% of SWE-Bench Pro tasks broken—hidden requirements, contradictory instructions, overly strict tests, or incomplete grading—and retracted its prior recommendation that the research community treat SWE-Bench Pro as a leading coding evaluation.
On July 8, 2026, SpaceXAI (formerly xAI under the SpaceX/AI stack branding) launched Grok 4.5—its strongest model yet for coding, agentic work, and knowledge tasks—trained alongside Cursor. It ships at $2/$6 per million tokens, ~80 tokens/sec, as default in Grok Build and on all Cursor plans, with API access via the SpaceXAI console; EU availability deferred to mid-July.
On July 8, 2026, OpenAI introduced GPT-Live—a new generation of full-duplex voice models powering ChatGPT Voice. GPT-Live-1 (default for Go/Plus/Pro) and GPT-Live-1 mini (default for Free) can listen and speak simultaneously, with global consumer rollout on web and mobile; Business/Enterprise/Edu excluded at launch.
On July 7, 2026, Anthropic announced Claude Cowork on web and mobile—ending the desktop-only phase of its agentic knowledge-work product. Beta access starts with Max plan users, with broader paid plans to follow; sessions can continue in the cloud when a laptop is closed and prompt users on phone when a decision is needed.
On July 7, 2026, Meta introduced Muse Image—the first image-generation model from Meta Superintelligence Labs—plus a Muse Video preview. Muse Image uses agentic tool use, multi-reference composition, and Instagram social context; it is rolling out in Meta AI, Instagram Stories, and WhatsApp in limited countries.