Good morning. Here's what's worth checking out today
BREAKING STORY
ChatGPT now lets businesses claim a profile and launch their own agent
OpenAI launched Businesses in ChatGPT in beta on 10 November. In its Business Hub, a verified company can update its details, add product feeds with prices and availability, and see aggregated insights: how often ChatGPT mentions it, how many people click through to its website, and the topics of conversations that cite it. Eligible US ecommerce and online-service businesses can also launch a Business Agent, a chat that answers questions from their own product data. Reuters reports that Shopify merchants can connect from their Shopify admin. Agents must say they are AI, and purchases need the user's explicit approval. No price is given.
My take: Claim your ChatGPT profile now, the way you once claimed your Google Business Profile. OpenAI says 1.2 billion people use ChatGPT each week, and until you claim your listing you can't correct what it shows or see how often you come up. This week, claim your business and check that your name, hours, prices and stock match your website. Then read the discovery topics, which are the questions people ask about you, and answer them on your own site. If you sell online in the US, test a Business Agent on ten real customer questions before you switch it on.
MARKET TRENDS
Forward-deployed engineer openings are up 4.6x in a year
Lenny Rachitsky's late-2026 job-market analysis, built on TrueUp data from more than 9,000 tech companies, found that openings for forward-deployed engineers rose 4.6x in a year, to about 1,500. A forward-deployed engineer (FDE) works inside a customer's business to get the product, now usually AI, actually working. Product design openings fell 28% to about 1,900, the lowest since tracking began, while open PM roles are near a four-year high.
How to make money from it: Sell the small-business version of the FDE job as a fixed-scope offer. For example: two weeks to connect an AI assistant to a client's inbox or CRM for a flat fee, then a monthly retainer to keep it running. Designers can bundle AI-workflow setup with their design work. After the first client, write up what changed in hours saved; that case study sells the next one.
GROWTH HACK / DIY
Turn every reviewer into your next user
The idea: Medium launched TK on 9 November, a writing app where, TechCrunch reports, writers can share early drafts with other people for feedback. That points to a loop any small product can build. Each request for feedback brings a new person into your product with a real task, and some of those reviewers later have drafts of their own and ask for feedback the same way.
Do it today:
Add a "Get feedback" button beside Share, with comment-only access as the default.
Let reviewers comment before they sign up. Ask only for a name, and keep their comments if they later create an account.
Tell reviewers when their feedback lands: "Priya accepted 3 of your suggestions."
Prompt once, after the draft is final: "Writing something yourself? Get feedback the same way." Suggest the person who invited them as their first reviewer.
What you'll need: a share link with a comment permission, inline comments, email notifications and a sign-up path that keeps guest comments; about one to two weeks for a small team.
Illustrative maths (invented numbers, not Medium data): 2 drafts a month with 3 reviewers each is 6 invitations. If 70% comment and 10% of those start a draft, each writer brings in 0.42 new writers a month. That won't grow on its own, but it turns every 100 writers you win elsewhere into about 172. Lift reviewer-to-writer conversion to 15% and it is about 270.
AGENT LAB
Make your agent check what it made
LangChain's sales agent, described on 10 November, fills a deck template, turns each page into an image and reads the images back to check the layout, fixing and re-rendering when something is wrong. It stops as soon as one review finds no real defect. On a one-pager, LangChain says that rule cut a run from 18 minutes to 7. An agent that only reads its own code can't see a heading wrapping onto three lines; a render step gives it eyes.
How it works: fixed template → render to images (a headless browser such as Playwright for HTML, LibreOffice for slides) → a vision model checks five things: cut-off text, bad wraps, missing images, content off the page, unreadable text → pass and stop, or fix and repeat, with a cap of three loops before a person takes over.
Try it: pick a document you make every week. Plant five defects by hand and keep the reviewer only if it catches at least four. Tell it to report real defects only, or it will never pass. Visual review won't catch a wrong figure, so check facts separately.
TRENDING AI RESEARCH
AI agents don't know what to ask, and your users won't tell them
What it says: In Ask2Decide, a benchmark posted to arXiv on 9 November, four AI models tackled 43 planning tasks with key facts hidden. Given every fact, their first plans were unworkable on 5–28% of tasks; made to ask for the facts, on 58–98%. Their questions found the facts that mattered barely more often than look-alikes that changed nothing. And real people rejecting a plan left out the correcting fact in 43% of replies, against 7% for a standard simulated user.
Why it matters: A confident first plan may rest on facts the agent never asked about, and polite test users hide the problem. It is an early preprint with 19 human participants, so treat the ranges as an early signal.
How you can use it: Before asking an assistant for a plan, write two or three lines on constraints and deal-breakers. If you build agents, make them collect the five to eight facts that most often change the plan before drafting, and add: "List the facts you don't have, say which could change the plan, and ask me only about those." When you reject a plan, say why.
TOOL REVIEW
Cardboard: edit videos on your Mac by describing the cut you want
What it does: Cardboard is an agentic video editor: you write the cut you want ("remove pauses, add captions, make it vertical") and it makes the edits. Its new Mac app finished tenth on Product Hunt on 9 November. Cardboard says it keeps your footage on your computer, though frames and the transcript are still sent to the AI. It needs Apple Silicon and can use your paid Claude or ChatGPT plan as the editor.
Price (as of 12 Nov 2026): Lite is free in the Mac app, with three watermarked 1080p exports. Starter is $9 for the first month, then $29; Creator $30, then $59; Pro $89, then $175.
Who it's for: solo founders, coaches and creators making talking-head clips, captions and short promos.
Verdict: Good for solo creators' quick edits. It is Mac-only, and the first-month promo prices roughly double after month one, so budget on the regular price.
WHAT GOT FUNDED YESTERDAY
Five rounds announced on 11 November.
Tano (tano.ai) · creator marketing · UK. Founded by Will Caplan (CEO) and Sagar Shah (CTO), the London company uses AI agents to find creators for consumer brands, handle outreach, send products, pay creators and track the sales they drive. It says this lets brands buy creator marketing against a sales target, the way they buy ads. It raised a $4M seed round led by Cherry Ventures.
Anorak (anorak.tech) · AI games · UK. Led by CEO Shaun Rutland, co-founder of mobile games studio Hutch, with co-founder Tak Fung and Chief Creative Officer Jonathan Alpine, it builds interactive generative video: AI video that responds to your choices in real time, like a game. It says its mission is "making any pursuit playable". It raised a $5M pre-seed led by Play Ventures and Backed VC, with Supercell and MTG.
Palomma (palomma.com) · proptech and fintech · Colombia. Founded by Cipriano Echavarría (CEO), Felipe Monsalve (CTO) and Nicolás Gómez del Campo, the Medellín company sells property managers AI leasing and collections software, rent processing, rent advances and guarantees. It says Latin American firms need all four to scale. It raised a $10M Series A led by Cometa, with Volpe Capital, Clocktower Technology Ventures, Tekton Ventures, Simma Capital and Yango Ventures.
Helix (helix.id) · identity verification · US. Co-founder and CEO Alex Oberg runs the Los Angeles company, which confirms a user is human by having them say four numbers aloud, with the checks run on their own device. It says it stores no biometric or personal data on a server. It raised a $6.2M pre-seed led by Chicago Ventures, with BAM Ventures, Corazon Capital, Tusk Ventures and others.
Rox (rox.com) · AI sales software · US. Founded by Ishan Mukherjee (CEO), Shriram Sridharan (CTO) and Diogo Ribeiro, the San Francisco company's AI agent finds leads, books meetings and manages deals for large enterprises. It says its recurring revenue grew over 10x in year one. It raised a $100M Series B at a $1.3B valuation, co-led by General Catalyst and March Capital, with Sequoia Capital, Factory, Databricks Ventures and Prosperity7 Ventures.
QUICK HITS
ElevenLabs says Scribe v3 makes 27% fewer word errors than v2 across 102 languages, and its live version now labels who said what in real time.
Anthropic added Claude Sheets, Whiteboards and interactive connectors from Adobe and Atlassian; its marketplace now lists more than 9,000 connectors and plugins.
OpenAI has bought Odessia, the AI travel agent built by Sonder founder Francis Davidson, in what TechCrunch says looks like an acqui-hire; terms were not disclosed.
That's today. Hit reply and tell me what you're building. I read every email.
Dev
