The AI agent market is repricing on a new clock. General Intuition, the New York startup building a foundation model that trains generalized agents to move through space and time, is in talks to raise fresh capital at a $6 billion pre-money valuation — nearly triple the $2.3 billion mark it set when it raised $320 million just weeks ago, according to TechCrunch.

A near-triple repricing in a matter of weeks is the sharpest signal yet of where venture conviction sits in this cycle. And it isn't an isolated data point. The same day, OpenAI made its flagship GPT-5.6 family available in the Kiro software development agent, and a LangChain case study revealed that Toyota now runs more than 50 agents in production. Agents are no longer a roadmap item. They're the product.

The Deal: Valor, Point72, and a Compressed Timeline

New investors in the proposed round include Valor Equity Partners, Point72 Ventures, and Seven Seven Six, with existing backers Khosla Ventures and General Catalyst also participating, TechCrunch reports. The round has not closed — a source close to the deal told TechCrunch it is oversubscribed as the startup continues to field investor interest — and terms could shift. If completed, the investment would mark Valor's first AI lab bet since it backed SpaceX.

CEO Pim de Witte spun General Intuition out of his video game clip-sharing platform Medal last October, converting an asset most founders would have left on the shelf: hundreds of millions of hours of gameplay annotated with "action labels" — records of which buttons a player pressed and when. Vinod Khosla told TechCrunch he believes those labels are a key part of the "emergence of intuition," the ability for a model to generalize across tasks it was never explicitly trained on.

The company plans to spend the new capital improving its general model with a focus on robotic embodiments, compute infrastructure through its partnership with CoreWeave, and talent, per TechCrunch. De Witte looked at what his first company had quietly accumulated and recognized it as the seed of something far larger. That is pattern recognition under pressure — the trait that separates founders who inherit assets from founders who understand them.

The Product Wave: Agents for Desktops and Codebases

The valuations are landing on top of a genuine shipping cadence. OpenAI's biggest bet is ChatGPT Work, released last month on its lowest subscription tier at $20 a month — a modified version of its Codex coding tool, rebuilt so non-engineers can field agents that complete multistep projects autonomously, TechCrunch reports.

"

"In this new factor, ChatGPT can actually do entire, very complicated tasks for you all autonomously in a way that is delightful and safe," Thibault Sottiaux, who leads OpenAI's core product work, told TechCrunch. "It's the very mission of OpenAI — to bring everyone along."

Getting there requires handing over the keys. Andrew Ambrosino, lead engineer for OpenAI's desktop app, has given the app access to and control over his inbox, Slack, phone, and apps like Notion and Figma — and concedes the privacy risk is real. "I'll do it for the job," he told TechCrunch. "I will take the personal hit here and there if I have to. And I haven't had to."

On the developer side, OpenAI's GPT-5.6 family — Sol, Terra, and Luna — is now available in Kiro, a software development agent built around spec-driven development, per the OpenAI Blog. OpenAI and AWS worked together to optimize the Kiro environment and OpenAI's models; in testing on Terminal-Bench 2.1, GPT-5.6 Terra completed successful tasks in Kiro at roughly an 82% cost reduction.

The Enterprise Proof Point: Toyota

If one number belongs in every enterprise AI pitch deck this quarter, it's this one: at Toyota Motor North America, shipping a new agent used to take 6 months and 6 engineers. Today it takes 4 days and 1 engineer, according to a LangChain case study.

Toyota's roughly 35-person enterprise AI team functions like an internal startup, with a mandate to clear a six-to-seven-figure annual ROI threshold on every project, per the case study. Its flagship platform, ToyotaGPT, now has more than 50 agents in production, grounded in internal data with permission-gated access — if an employee lacks SharePoint access to a dataset, the agent doesn't surface it.

"

"We went through the pain of building everything ourselves," Kordel France, Toyota's Director of AI Engineering, told LangChain. "It just costs a lot of labor, a lot of engineering time."

The team built on Deep Agents, LangGraph, and LangSmith, injecting reusable domain skills — manufacturing, supply chain, R&D, branding — into agents at runtime rather than hardcoding knowledge into each one. "With a single command, create deep agent, you get a powerful harness and an entire ecosystem for each new agent," said Ravi Chandu Ummadisetti, Director of Agentic AI and Product Research. Their GearPal application lets any technician walk up to a broken machine and ask, in plain language, why it failed and how to fix it — on a production line where every minute of unplanned downtime costs hundreds of thousands to millions of dollars.

What It Means for the Market

Three signals stand out. First, valuation velocity is now a feature, not a bug: General Intuition's near-triple repricing in weeks mirrors the infrastructure layer, where Hugging Face has been approached to sell at $13 billion or more — after turning down a $500 million Nvidia investment at a $7 billion valuation earlier this year — and Stripe has already paid $7 billion for OpenRouter, per TechCrunch.

Second, the economics reward autonomy. Agents that work longer burn more tokens, making them more lucrative per user — which is precisely why OpenAI needs ChatGPT Work to escape the engineering department. Vertical competitors like Harvey (law) and Clay (sales) are chasing the same customers with model-agnostic products. That sets up the strategic question of the cycle: as Christian Catalini warned on a16z's "It's time to build" blog, if the labs cannot rapidly secure the key complementary assets needed to scale AI in the market, value will accrue elsewhere.

Third, the agent wave is global. The European Commission's Apply AI Startup Award drew 24 nominated startups across 16 member states, with independent experts now selecting the top 10 finalists who will pitch on October 20, per the European Commission.

What to Watch

Watch whether General Intuition's round closes at $6 billion and who leads it — a completed round at this velocity will show how much farther the agent premium can stretch. Watch whether ChatGPT Work converts accountants and doctors the way Codex converted developers; the labs' training economics depend on it. And watch whether agents holding the keys to users' inboxes, Slack accounts, and phones survive contact with enterprise privacy and compliance — an agent with access to everything is a feature right up until it surfaces the wrong DM.

The founders in this space are teaching machines to act on our behalf while fighting, week by repricing week, to prove their companies deserve to exist. The fight to be seen as real — for time, for capital, for the right to keep building — is the oldest story in this market. The builders shipping real agents into real workflows — Toyota's technicians, Kiro's codebases, the accountants and doctors OpenAI is courting — are earning the premium. The rest are renting it. Track the deployment numbers, not the demos.