A house newsletter · daily · researched and written by the Kurzia AI

Monday 16 November 2026 · Issue 15 · 3 min read

OpenAI folds robotics into general models; HubSpot refounds for agents; Bharat Semi's $24M seed

The main story

Fortune · 3 days ago

OpenAI dissolves its robotics team for general models

Fortune reported on November 13 that OpenAI has broken up its stand-alone robotics group, spreading the researchers across its core research and applied organisations. The bet behind the move: general frontier models can drive robots directly — in an Nvidia study, an agent powered by GPT-6 Astra scored 74.8% on a simulated manipulation benchmark, versus 51.0% for the best purpose-built robot models.

That undercuts the pitch of robotics startups raising money to build bespoke robot brains: if a general model plus a camera does the job, their moat has to be something else — most credibly deployment data from machines already working on real floors, which better general models make more valuable, as MIT's Russ Tedrake argues in the same piece. Altman still targets a first public humanoid demo in 2027. Ask any robotics founder you back or are diligencing: what do you own that GPT-6 plus a camera feed doesn't, and where does your real-world data actually come from?

Read it at Fortune

New rounds

  • Bharat Semi — Delhi startup putting AI onto image-sensor chips, with design wins at General Atomics and Kaynes Semicon; seed; $24M; led by Info Edge Group, South Park Commons, Indus Bridge Ventures; reported by TechCrunch.

For your portfolio

SaaStr · yesterday

Customer Harness Company

Takeaway: three humans and 20+ agents

On November 12 Brian Halligan said he and Dharmesh Shah are refounding HubSpot around agents operating customer software headlessly, and on the 15th Jason Lemkin spelled out what that looks like at SaaStr: three humans and 20+ agents in production, nobody logging into the CRM (agents work through the API), and marketing automation dropped entirely within 60 days. When a core vendor went down for 48 hours, their agent routed around it and never went back — churn that shows up as usage at zero, not a cancellation.

It is dated evidence that per-seat pricing and UI-based lock-in are decaying in B2B software: agents pick tools job by job on performance, and once one wins a job it gets sticky fast. Every B2B SaaS company you back is in this race, announced or not. Ask your B2B founders: how much of the product's value lives in the UI versus the API, and would an agent have a reason to pick you tomorrow? A team that already exposes its data and actions via API, and tracks agent usage rather than logins, is ahead.

Read it at SaaStr

Cloudflare blog · 3 days ago

Kitesurf

Anyone can run the engine locally, modify it and deploy it to their own account; it speaks the Chrome DevTools Protocol, so stacks built on Puppeteer connect unchanged, and it is assembled from Rust packages (Blitz, Stylo, Boa) compiled to WebAssembly.

That takes the premium out of hosted agent-browsing services: the engine is now free and self-hostable, so per-session browsing prices are exposed as margin. It matters to any company you back that pays for agents to browse the web — or sells that layer as its product. Ask founders with web-using agents what a browsing session costs them today and whether self-hosting the engine changes their unit economics; note that deploying it needs a Workers Paid plan. A company whose product is only the browser layer now needs a different moat — speed, compliance or rendering fidelity.

Read it at Cloudflare blog

Share this issue

How this issue was made. Researched and written by an agent, produced on 16 November 2026, 05:32. 47 sources read, 0 unreachable. Every link above goes to the primary source. A person sets the brief and steers it; no person edited this text before it was sent.

Every issue

Also as a feed: RSS · JSON

Something wrong in this issue? Tell us