Factory Raises $150M at $1.5B Valuation as OpenAI Turns Codex into Mac Super-App; DeepMind, Anthropic and Alibaba Ship Major Model Updates

Good morning — Aurora and Isabelle here with AI Tech News Today.
Date of this briefing: 2026-04-17T11:30:48.000Z

Quick hits to keep you current on the biggest moves in A‑I.

Factory raises one hundred fifty million at a one point five billion valuation

Headline: Factory closed one hundred fifty million dollars in new funding at a one point five billion dollar valuation.
Details: Investors include Keith Rabois, Shaun Maguire, Blackstone, and others. Founder Matan gets a shout-out in the coverage — people expect big things. Think of this as a bet that Factory can go from stealthy to unmissable.

OpenAI turns Codex into a true Mac super-app

Headline: OpenAI updated the Codex desktop app so it can see, click, and type inside any native app.
Key features:

• Iterate images without leaving your flow
• Run automations that pick up where they left off
• Tap into 90+ plugins for documents, project tools, and deployments

Goal: Make the desktop feel less like a tool and more like a helpful coworker that actually knows your shortcuts. The update starts rolling out today.

Google DeepMind launches Gemini three point one Flash TTS

Headline: Google DeepMind released Gemini three point one Flash TTS, a highly controllable text-to-speech model.
Highlights:

• Simple text tags to direct vocal style, delivery, and pace
• More natural speech in 70+ languages
• Adds Synth ID watermarking on every output

Availability: Developers via the Gemini API; ordinary users will see it inside Google Vids.

Anthropic drops Claude Opus four point seven and updates Claude Code

Headline: Anthropic shipped Claude Opus four point seven and redesigned Claude Code desktop.
Improvements in Claude Opus 4.7:

• Better long-task handling
• Stricter instruction following
• Self-verification of outputs
• Vision resolution triples to 2,576 pixels for better screenshot reading

Claude Code updates: Run multiple sessions side-by-side, edit files, use an integrated terminal, and preview HTML and PDF — a strong upgrade for developer workflows.

Alibaba open-sources Qwen three point six – thirty five B – A three B

Headline: Alibaba released Qwen 3.6-35B-A3B, a sparse mixture-of-experts model.
Model specifics:

• 35 billion total parameters, but only 3 billion active at a time
• Matches much larger models on agentic coding and multimodal reasoning

License & access: Available under the Apache 2.0 license on Hugging Face.

That’s your update. Stay curious, keep experimenting, and we’ll be back with more A‑I news.
Thanks for listening — I’m Aurora, with Isabelle at AI Tech News Today.