The second week of July was a funding blitz — and the clearest sign yet that AI agents aren't just being built, they're doing the building.
An AI agent ran its own $100M fundraise
On July 9, enterprise agent startup Lyzr raised a $100M Series B (~$500M valuation) — and let its own agent, "SivaClaw," run the process: fielding questions from 130+ investors, drafting investment memos, and tracking which pitch slides backers lingered on. It's a vivid demo of where automation is heading — agents that don't just answer questions but own end-to-end workflows. See automation & agents.
The build-your-own-agent stack levels up
Two raises the same week fueled the infrastructure under those agents. Prime Intellect raised a $130M Series A (July 8) at a $1B valuation for a full-stack platform — compute, RL frameworks, evals — that lets enterprises train their own agents (customers include Ramp and Zapier). And Ollama, the tool for running open-weight models locally, raised $65M (July 9) and now serves nearly 9 million monthly developers. The picks-and-shovels layer for agents is consolidating fast. Compare automation platforms and coding tools.
Voice agents pull nine-figure money
Paris-based Gradium raised a $100M seed (July 9), backed by Nvidia, to build ultra-low-latency voice models for agent conversations. Real-time voice is becoming the interface for AI agents — worth watching if you're picking a voice AI tool.
Model watch: OpenAI's live voice + the GPT-5.6 family
OpenAI had a big two days. On July 8 it launched GPT-Live-1 and GPT-Live-1 mini, full-duplex voice models that listen and speak at once (natural interruptions, live translation) — the mini replaces Advanced Voice Mode by default for all ChatGPT users. Then on July 9 came the GPT-5.6 family in three tiers — Sol, Terra, and Luna — spanning flagship to budget ($5/$30 down to $1/$6 per million tokens), with Sam Altman claiming Sol is "54% more token efficient" on coding.
What it means if you're picking tools
Follow the money: it's flowing into agents that do work (Lyzr), the stack to build them (Prime Intellect, Ollama), and the voice layer to talk to them (Gradium) — even into fully agentic vertical services like AI law firm Norm ($120M, July 7). The practical upshot is that your automation, voice AI, and chat assistant options are getting deeper and cheaper by the week. Vote for what's actually working.
Sources: Lyzr $100M (TechCrunch) · Prime Intellect $130M (TechCrunch) · Ollama $65M (TechCrunch) · Gradium $100M seed (TechCrunch) · Norm $120M (TechCrunch) · OpenAI GPT-Live voice (TechCrunch) · GPT-5.6 family (TechCrunch)