A house newsletter · daily · researched and written by the Kurzia AI
Tuesday 17 November 2026 · Issue 16 · 3 min read
Corma hits $1B three months in; five fresh rounds; Claude's watermark spreads; pricing for agent buyers
The main story
TechCrunch · today
Corma reaches $1B three months after launch
Takeaway: a $1 billion valuation
Corma, an AI cybersecurity startup, said on November 16 it raised a $125 million Series A at a $1 billion valuation, led by NEA — about three months after launching its product and shortly after leaving stealth in August.
Total funding is now $185 million from Sequoia, Khosla Ventures and Coatue, reported by TechCrunch. Demand jumped after October 21, when a critical PHP flaw, a Node.js flaw and an incident at LiteLLM broke on the same day — the kind of AI-enabled onslaught that creates a budgeted buyer overnight. For the next cheque, AI security is a category where a three-month-old company can command $1 billion on real demand. Tell any company you back that ships code or runs open-source dependencies: ask how it would have survived October 21, and whether an AI defense layer is now table stakes.
New rounds
Stoa — Helsinki AI that pulls drawings, schedules and safety reports into one view for construction teams; pre-seed; €3.5 million; led by Lifeline Ventures; reported by Tech.eu.
Valeria — Barcelona AI payroll and HR platform for high-turnover businesses; seed; €5.5 million; led by BlackFin; reported by Tech.eu.
Daisytuner — German compiler startup that tunes AI models to run faster on different chips; seed; €6 million; co-led by Elaia and LEA Ventures; reported by Tech.eu.
Dreamteam — AI-native CRM by Freshworks alumni challenging Salesforce and HubSpot; seed and Series A; $53 million; led by Together Fund and Accel; reported by TechCrunch.
Intergrid — Helsinki platform that autonomously runs industrial energy operations at 100+ sites; stage not stated; €23.5 million total funding, latest round co-led by Kvanted and Tesi; reported by Tech.eu.
What the AI labs did
Claude text watermark — Anthropic's November 16 release notes say text from Claude Opus 4.6, Opus 4.5, Sonnet 4.5 and Haiku 4.5 — four more models — carries its watermark on supported platforms; a non-default temperature or top_p can drop the mark. That undercuts startups selling probabilistic AI-text detection, since buyers of Claude output get provenance free, and hands compliance tools a hook to verify content origin. Ask any founder selling detection what they add beyond a lab watermark, and check your own pipeline's sampling settings don't strip the mark. Claude Platform release notes
For your portfolio
SaaStr · yesterday
Price for agents, or agents will route around you
Takeaway: cut those calls by about 90%
Jason Lemkin argued on SaaStr on November 16 that B2B companies rebuilding around AI should cut prices — possibly by half or more — because AI agents measure cost per call every day and quietly drop vendors that are too expensive.
His evidence is his own shop: SaaStr runs with 3 humans and 20+ agents, one vendor's bill rose 40% this year from about 35,000 agent calls a day, and his agents cut those calls by about 90%. Databricks, he notes, removed a $1,000 per-user fee and consumption went way up. Ask each B2B founder you back: do you publish a per-call price an agent can model, and when did you last test a 50-80% cut on one tier and watch what happened to usage?
“did you earn it? If you raised prices 8% last year, did you add 30% more value to your product?”
Lenny's Newsletter · yesterday
Free trials are the open faucet for token fraud
Takeaway: tens of thousands of dollars
Claire Vo, solo founder of the bootstrapped AI product-manager tool ChatPRD, wrote on November 16 that token fraud cost her tens of thousands of dollars over the past year: farmed free trials, resold promo codes, the product wrapped in relays, and a custom MCP installed as a 24/7 compute worker. She has ended free trials for good.
Any portfolio company selling AI access is selling compute, and an open free trial is a faucet someone will resell. The leak shows up in gross margin long before it shows up in churn. Ask founders with a free tier whether trial tokens are capped and whether anyone watches cost per signup; a team behind still runs open trials, a team ahead has already instrumented per-user cost.