GPT-5.6 goes public, AI IPOs dwarf 25 years of tech exits
OpenAI gets government greenlight for GPT-5.6, three AI IPOs outvalue 25 years of VC exits, and Ollama hits 9M users with fresh $65M.
Subscribe free →OpenAI's GPT-5.6 Gets the Green Light — And Launches an Agent Too
After weeks of regulatory limbo, the Trump administration cleared GPT-5.6 for public rollout, and OpenAI wasted no time doubling down: the company simultaneously unveiled "ChatGPT Work," a new AI agent targeting enterprise users. CEO Sam Altman called GPT-5.6 "the best model we have ever produced," which he says every single time.
Why it matters: The fact that a model release required government approval before going public is a new and consequential norm for the AI industry.
Three AI IPOs Are Worth More Than Every U.S. VC Exit Since 2000
Anthropic, OpenAI, and SpaceX are on track to generate more value from their IPOs than the combined total of all U.S. venture-backed exits over the past 25 years. Let that sink in. The AI boom isn't just big — it's historically unprecedented in its concentration of value.
Why it matters: This reframes how investors, founders, and regulators should think about AI — this isn't a tech bubble, it's a value redistribution event unlike anything the industry has seen.
SpaceXAI Drops Grok 4.5 — Cheaper, Faster, "Opus-Class"
SpaceXAI shipped Grok 4.5 on Wednesday, with Musk positioning it as a cost-efficient rival to Anthropic's top-tier Claude Opus models. The pitch is simple: frontier-level performance without the frontier-level price tag.
Why it matters: If Grok 4.5 actually delivers Opus-class reasoning at lower cost, it puts real pricing pressure on Anthropic and OpenAI at exactly the moment both are trying to convert enterprise customers.
Ollama Raises $65M as Local AI Hits 9 Million Users
Ollama, the tool that lets developers run LLMs locally on their own machines, just closed a Benchmark-led $65M round. The open-source project has 176,000 GitHub stars and nearly 17,000 forks — numbers that reflect genuine grassroots adoption, not hype.
Why it matters: Ollama's growth proves there's a massive, underserved market of developers who want powerful AI without sending every query to a cloud API — and now it has the capital to build for them.
Meta's Instagram Photos Are Feeding Its AI — Unless You Stop It
Meta quietly rolled out an AI image generator trained on public Instagram photos, with opt-out buried in settings. Users who haven't actively disabled data sharing are already contributing their content to Meta's training pipeline.
Why it matters: This is the playbook: launch first, make opting out an afterthought, and collect training data from hundreds of millions of users who never explicitly agreed.
Google's Deepfake Detector Catches McConnell Hospital Hoax
A fabricated image showing Senator Mitch McConnell hooked up to tubes in a hospital bed circulated widely this week before Google's deepfake detection system flagged it as AI-generated. The image was a fake — and Google's tool caught it before it caused lasting political damage.
Why it matters: This is one of the first high-profile cases where AI detection infrastructure worked as intended on politically sensitive disinformation — a proof of concept that will matter a lot heading into election cycles.
Quick Hits
- →Meta's Muse Spark 1.1 model API is now open to developers, with the company claiming it's a "step-change" improvement in coding capabilities. The Verge
- →Anthropic's new Claude "Reflect" dashboard shows you how often you use Claude — which is also a very effective way to make you feel dependent on Claude. TechCrunch
- →Vibe coding darling Lovable is reportedly in talks to raise $300M at a $13.2B valuation — doubling its valuation in one round. TechCrunch
- →AWS GraphRAG cut pharmaceutical drug research cycles by 87% by unifying siloed proprietary databases into a single queryable knowledge graph. AI News
- →NHS hospitals are preparing to deploy an AI-powered blood test to screen for womb cancer before subjecting patients to invasive procedures. AI News
AI TLDR