Thursday, September 3 September 3, 2026

Good morning. It's Thursday, September 3rd, 2026, and today's briefing is one for the history books — because three of the biggest AI labs dropped major model news in a single 24-hour window. Let's get into it.

Start with Google. Yesterday the company launched Gemini 3.8 Flash — its third Flash model in just six weeks. That pace is staggering. Google says 3.8 Flash, codenamed Skimaki internally, makes real progress in coding — an area where they've historically trailed Anthropic and OpenAI. It's priced at 75 cents per million input tokens, undercutting major competitors, and it's already live in AI Mode for Google AI Pro and Ultra subscribers. For developers shopping on cost and speed, Google is clearly playing offense.

But Google didn't even get a full day in the spotlight. Meta fired back with Muse Spark 1.3 — their most capable model yet. This is significant because Meta has been playing catch-up since Llama 4 underdelivered earlier this year. With Muse Spark 1.3, Meta's chief AI officer Alexandr Wang is claiming benchmark scores on par with GPT-5.6 and even surpassing it on some coding tasks. On the Artificial Analysis Intelligence Index, Muse Spark 1.3 scored 61, putting it at the frontier tier. Wang, never one to stay quiet, also took some shots at Google's benchmark numbers on social media. The AI PR battle is officially a contact sport.

And Anthropic? They haven't been sitting still either. Claude Fable 5.1, released September 1st, is currently sitting at the top of the Artificial Analysis Intelligence Index with a score of 66, leading on agentic knowledge work. Anthropic also quietly cut cache read prices by 75 percent — a move that helps enterprise customers running AI at scale. So the pecking order as of this morning is: Fable 5.1 leads the index, Muse Spark 1.3 is right behind it, and Gemini 3.8 Flash is carving out the cost-optimized lane.

Now let's talk money. Wonderful, an Israeli-Dutch enterprise AI startup, just raised 550 million dollars at a 5 billion dollar valuation. Here's the wild part: they were founded in early 2025. That's 20 months old. They went from a 34 million dollar seed round to a 5 billion dollar valuation with 70 million in annualized revenue and customers in 35 markets. Insight Partners led the round, Salesforce joined as a new investor, and the company plans to scale to 1,000 engineers. Their pitch is simple: every large enterprise needs an AI operating system, and Wonderful wants to be it. Think of it as the cloud computing infrastructure layer — but for AI agents and workflows.

Over in China, Moonshot AI has quietly filed for a Hong Kong IPO. The company, which makes the Kimi AI assistant and competes directly with OpenAI in the Chinese market, is valued at 50 billion dollars and is reportedly seeking to raise 3 billion. If that goes through, it would be one of the largest AI-related public offerings of the year.

On the regulatory front, the FDA launched a pilot program called TEMPO — offering generative AI medical devices a conditional path to real patients before full regulatory authorization. The idea is that regulators need hands-on experience with these tools in real-world settings. It's cautious, but it's a real opening for health AI companies.

And in education, New York City announced it will ban AI use for students in public schools through eighth grade. The policy kicks in next week. Expect this debate to heat up nationwide.

The theme today is clear: the model wars are accelerating. Google, Meta, and Anthropic are all pushing releases at a pace we've never seen before. Meanwhile the enterprise AI market is rewarding companies that can deploy agents at scale — fast. Wonderful's valuation trajectory is the signal to watch.

Here's today's business idea: build a vertical AI OS for mid-market companies — think 100 to 1,000 employees — who want Wonderful's value prop but can't afford $150 million round-level pricing or a dedicated forward-deployed engineering team. White-label the orchestration layer, pre-integrate it with the top 10 SaaS tools in one industry — say, insurance or logistics — and sell it as a managed subscription. The timing is perfect: every large-enterprise AI OS player is moving upmarket, leaving the mid-market underserved and ready to buy.

That's your briefing for September 3rd. Stay sharp out there.