IN PARTNERSHIP WITH
Every app makes small decisions, like routing a request or flagging an outage. Jev is a System One model that takes the information you give it and answers your predefined questions in three ways (choice, score or yes/no). Each answer has a probability, so your app knows when to act or ask a human.
Welcome back. On Thursday OpenAI gave ChatGPT agents a wallet, so they can now pay for things on your behalf. And an hour ago it shipped GPT-6.2 Astra and Sol, two new models built for professional work. Hot take: the agent that can spend your money is still a bigger change for your business than the model behind it.
Today: ChatGPT Wallet, GPT-6.2 Astra and Sol, Anthropic's IPO, and a 15-minute setup for letting an agent spend safely.
TODAY IN AI
1. ⚡ ChatGPT agents can now pay for you: ChatGPT Wallet lets your agent complete purchases after you approve each one with Face ID or a fingerprint. Mastercard, Visa and American Express issue a payment token, so the agent never sees your real card, and you keep your card's rewards and dispute protection. It's rolling out to ChatGPT Pro users in the US (18+, US-issued cards), and OpenAI takes no cut at launch. Some sites still say no: United plans to block agents, and Amazon doesn't allow them.
Why it matters: if you sell online, agents are about to become a sales channel. Check that your checkout works without pop-ups and captchas, and that your product pages state price, stock and delivery in plain text an agent can read.
2. ✨ GPT-6.2 Astra and Sol landed an hour ago: Astra is OpenAI's new frontier model for hard projects, and OpenAI says its slides, spreadsheets and documents now follow your templates and read less like AI. Sol is the efficient model for everyday work. On Agents' Last Exam, OpenAI reports Astra 5.5 points ahead of Claude Fable 5.2 at 5% of the cost. In the API, Astra costs $8 per million input tokens and $40 per million output (20% less than GPT-6 Astra), and Sol stays at $2 and $10. Both are live today for Plus, Pro, Enterprise and Edu in ChatGPT Work and Codex.
Why it matters: if you build decks, sheets or docs with AI, give Astra your own template today and see how many edits you skip. If you pay per token, Sol handles most everyday work at a quarter of Astra's price.
3. 🚀 Anthropic goes public this week: the maker of Claude set a range of $70 to $78 a share for 825 million shares on Nasdaq, under the ticker ANT. At the top of the range that values it at about $1 trillion. Pricing is expected tomorrow, November 18, with trading the day after. The final price can still change.
Why it matters: once Claude's maker answers to public shareholders every quarter, expect more pressure on revenue: price changes, new paid tiers and a harder push into business tools. If your team runs on Claude, write down what you pay and use today, so you can spot the changes when they come.
What this means for you: agents just got a card and a budget. The people who set limits and approvals now will be the ones who can safely let them work.
Why Gopuff replaced its insurance broker with us.
Gopuff's legal team cut premiums by 20% while improving their coverage. WithCoverage is the insurance broker and platform for companies paying six figures a year for business insurance. Start with a free, confidential risk analysis.
FROM THE FRONTIER
Give an AI agent $200 and it spends less than half.
The good news. Induction Labs' FrontierML test gave agents 21 long machine-learning jobs, a 12-hour sandbox and a wallet of $10, $50 or $200 for GPUs and tokens. More money did buy better work: Claude Opus 5.5 scored 45.4 with $10 and 58.8 with $200, against 66.1 for a human researcher.
The bad news. Agents spent only about 45% of their wallets on average and never more than 80%. They used 5.3 of their 12 hours, and over 67% of their data-generation calls went to models more than a year old. They didn't stop to check what better options existed.
My call: tell agents what to spend, not just what not to
A budget cap alone makes agents timid. Give them a target too: "use up to $50, try at least two current models, report what each cost." Then compare results, not just the bill. I'd change my mind if the next round shows agents using most of their budget without being told.
IN THE KNOW
𝕏 ChatGPT @ChatGPT
Now your dot can securely complete purchases on your behalf—with your approval. With ChatGPT Wallet, agents pay without access to your actual card details…
♥ 5.2K · ↻ 413 · View on X
𝕏 Harrison Chase @hwchase17
Today we announced that LangChain raised $300M at a $3B valuation. We think agents will transform the economy…
♥ 1.3K · ↻ 101 · View on X
See the GTM Stack Behind Pebl’s Handoff
In global employment, the sale is only the start. Customers stay or leave based on what happens after signature.
At The GTM Garage, Pebl’s Chief Business Officer, Dutta Satadip, walks through the tools behind prospecting, pipeline, and handoff to onboarding, and how Pebl uses AI across the customer lifecycle.
15 min conversation. 15 min open Q&A. For Seed to Series A teams.
TOOLBOX
🤖 Relevance AI: build AI agents for sales and support without code, and set which steps need your approval before the agent acts.
⚙️ Make: visual automations. Send every agent order over your limit to Slack or email for a yes/no before it goes through.
📅 Reclaim AI: AI calendar that protects focus time and books a weekly 15-minute slot to review what your agents did and spent.
TUTORIAL
Let an agent spend money without losing sleep (15 minutes)
Pick one buying job: something you repeat, like reordering supplies, booking a tool trial or topping up ad credits.
Set the numbers: the most it may spend per purchase and per month, and the shops it may use.
Run the prompt below to turn that into written rules your agent follows.
Use a separate card with a low limit for agent purchases, so one mistake can't reach your main account.
Give it a hard stop: one line that says "if anything here is unclear, ask me before paying."
Review weekly: compare what the agent bought with what you would have bought, and tighten the rules where it drifted.
I want an AI agent to handle this buying job for me:
[job, e.g. reorder office supplies every 2 weeks]
My limits: [max per purchase], [max per month], allowed shops: [list]
Write the rules the agent must follow:
1. What it may buy without asking, and what always needs my approval.
2. What to check before paying: price vs last time, delivery date, return policy.
3. When to stop and ask me (price jump over 15%, unknown shop, item out of stock).
4. What to report after each purchase, in 3 lines.
Then list 5 ways this could go wrong and the rule that prevents each one.PROMPT STATION
Agent-Ready Product Page Audit
When you sell online and want AI shopping agents to understand, trust and buy from your pages.
## Role
You are an e-commerce consultant who checks if AI shopping agents can buy from a store.
## Task
Audit my product page and tell me what stops an agent from buying.
## Context
Page text: [paste the full product page text]
Checkout steps: [list them, e.g. cart, account, address, payment]
## Output
1. Missing facts an agent needs: price, stock, sizes, delivery time, returns.
2. Anything only shown in images or pop-ups that an agent can't read.
3. Checkout steps that would block an agent, with a fix for each.
4. A rewritten product summary of 5 lines with every key fact in plain text.
Rank fixes by how likely they are to lose a sale.👉 Browse more prompts like this. Built for ChatGPT, Claude, Gemini, Grok.
Want one written for your business instead? Describe the task in a sentence and the prompt generator writes it, now UNLIMITED with your context profiles inside NjectAI.
IMAGE PROMPT
Phone Checkout Close-Up, Clean Studio Light
When you need a crisp "buy in one tap" visual for a store ad or a launch post.
A hand holding a smartphone showing a simple checkout screen with one large "Pay" button.
Background: [soft beige / light grey / pastel] seamless studio backdrop.
Next to the phone: [your product], small, sharp, slightly in front.
Camera: close-up, slight top-down angle, 85mm look, phone screen in focus.
Light: soft even studio light, gentle shadow under the hand, no glare on the screen.
Frame: vertical 4:5, empty space at the top for a headline.
Style: clean commercial photo, true colours. No brand logos, no extra text, no watermark.👉 More image prompts in the library. Works in Grok Imagine, ChatGPT and similar image models.
Want more? the free prompt library · the free guides
Hit reply and tell me: would you let an agent buy for you, and what's your limit? I read every reply.
🔑 God of Prompt
P.S. If you only do one thing today, run the TUTORIAL prompt before you hand any agent a card. The prompt library has more like it.



