Morning brief

Thursday, July 9

3 people · 0 companies · 5 topics

TL;DR Grok 4.5 went public today alongside GPT-5.6 in what multiple outlets called the first day all three major frontier labs have simultaneous public frontier models. Polymarket still prices Anthropic at 87% for “best AI model end of July” despite the launches, and OpenAI’s 2026 IPO market collapsed toward zero after a confirmed 2027 delay.

Models & research

  • model-releases elonmusk Grok 4.5 is live (SpaceXAI, V9 foundation, 1.5T params, 500K context, $2/$6 per M tokens). SWE-Bench Pro: 64.7% — behind Fable 5 (80.4%) and Opus 4.8 (69.2%), but 4.2x more token-efficient per task than Opus 4.8. Musk positioned it as “Opus-class” while acknowledging it competes with “last year’s Claude Opus.” Decrypt · ExplainX · Crypto.news

  • OpenAI shipped GPT-Live (GPT-Live-1 and GPT-Live-1 mini): full-duplex voice models that listen and speak simultaneously, delegating harder reasoning mid-conversation to GPT-5.5. simonw had preview access for weeks and called it “very impressive.” OpenAI · simonwillison.net · HN

People

  • steipete announced he joined OpenAI to work on “bringing agents to everyone”; OpenClaw (346k GitHub stars, the most-starred AI agent repo) is spinning out to a standalone foundation and staying open-source. steipete.me

Tools & dev

  • claude-code v2.1.205 (July 8): auto mode now blocks transcript-tampering rules; improved agent view adds classifier-written state headlines; /doctor becomes a full setup checkup with /checkup alias; auto-updates stream to disk, cutting ~400 MB peak RAM. GitHub releases

  • Microsoft Research released Flint, an intermediate visualization language for AI agents that lets models spec charts in a simple format that compiles to Vega-Lite, ECharts, or Chart.js without rewriting per target; 113 HN points, 299 comments. Microsoft Research · GitHub · HN

OSS

  • iOfficeAI/OfficeCLI — AI agent suite for reading, editing, and automating Word, Excel, and PowerPoint, 1,923 GitHub stars today. GitHub

  • VoltAgent/awesome-design-md — DESIGN.md files that give coding agents enough UI context to generate matching interfaces, 1,569 stars today. GitHub

  • Graphify-Labs/graphify (Python) — turns code, schemas, and docs into queryable knowledge graphs as a coding skill for LLM agents, 856 stars today. GitHub

Other

  • A Brown University professor ordered an in-person re-exam after suspecting AI cheating; class average on the take-home midterm was 96/100, in-person re-exam dropped to 48/100. Trending on HN overnight. Business Story · HN

  • Polymarket: Anthropic $1.25T valuation by Dec 31 sitting at 91% ($2.1M volume) — traders treat it as near-certain given Anthropic’s ~$1.11T private valuation in early July. (Polymarket: 91% yes, $2.1M volume) Polymarket

  • OpenAI confirmed its IPO is delayed to 2027, collapsing the Polymarket “OpenAI IPO by end of 2026” market toward zero. Companion market “Will Anthropic or OpenAI IPO first?” now prices Anthropic first at 73-87%. Polymarket · Yahoo Finance

  • UN Independent International Scientific Panel on AI published its preliminary report this month with “catastrophic harm” warnings and a push for binding global governance. UN report