In partnership with
Hey builders, happy Thursday.
Every story this week has the same question hiding in it: how much of your work are you ready to hand to AI? GitHub wants to pick your model, Shopify wants AI doing the buying, and you can build a bot today that does your Monday research while you’re away.
So, before the news, let us know 👇
What’s the first task you’d hand to an AI agent?
TL;DR
GitHub’s Copilot Fusion gets near-Opus coding results for about half the cost
Anthropic starts watermarking text from four more Claude models on Nov 3
Gemini 4 Carbon tops the intelligence index but opens to security teams first
Shopify’s Fall 2026 Edition brings 150-plus updates, AI purchasing, and shop.com
Build an AI analyst that emails you a research briefing every Monday
heyclicky draws on your screen to show you what to click next
The ice cream shop that makes money when it's cold
28 Wishes sells ice cream in Los Angeles. Below 70°F, sales fall about 20%. The weather is out of their hands. Rent isn't.
So the owners started putting about $20 a day into Kalshi weather markets, taking the cold side. The days that keep customers away now pay something back.
This is hedging. Big companies have done it for decades, buying protection against bad weather, fuel spikes and rising rates. It used to take a broker, a trading desk, and an order size no corner shop could meet.
Kalshi opens it up. Contracts on weather, fuel prices, inflation, tariffs and regulation, starting at a few dollars. Take a position on the outcome that would hurt you. If it hits, the payout softens it. If it doesn't, the contract expires and the good month was the point.
GITHUB’S NEW COPILOT NEARLY MATCHES OPUS AT HALF THE COST
If you’ve ever lost an hour picking the right model, GitHub just made that decision for you...
Two points. That’s how far GitHub says Copilot Fusion trails Claude Code running Opus 5.5 Max on the Artificial Analysis Coding Agent Index. Fusion Max scored 60.
Now for price... Fusion averaged about $5.59 per task against about $11.42, GitHub reports, and finished roughly 40% faster.
Fusion reads each task, works out whether it needs reasoning, code generation, debugging or tool use, then picks one model or lines up several to work together. You only choose the effort level: Standard, High or Max.
It’s live across Copilot plans in VS Code, JetBrains, the Copilot app and the CLI, and you’re billed for the underlying models.
GitHub claims agents opened more than 40% of pull requests in September, up from 17% in March. At that volume, nobody wants to hand-pick a model for every task. GitHub also opened every Copilot feature to Enterprise customers without a separate subscription, paid for with usage-based AI credits.
OUR TAKE
Picking one model for everything is starting to look like flying first class on every trip.
If a router gets you within two points of the best for half the price, the premium pick turns into a luxury.
Expect every coding tool to copy this.
Run it on your own repo before you trust any chart, GitHub’s included.
QUICK HITS
Anthropic adds invisible watermarks to four more Claude models: Claude is about to start signing its work, in ink you can’t see. On November 3, Anthropic starts watermarking text from Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6, a subtle pattern in word choice that doesn’t change quality, pricing or speed. Opus 4.6 and Sonnet 4.6 only carry it at default temperature, top_p and top_k, so if you rely on it for compliance, leave those settings alone.
Google’s Gemini 4 Carbon takes #1, but only defenders get in: Google built a model that is very good at finding holes in software, so the first keys went to the people who defend against them. Gemini 4 Carbon took the #1 spot on the Artificial Analysis Intelligence Index at 59 and scored 88.2% on SEC-bench Pro, a test of finding real vulnerabilities. Early access is limited to vetted security teams through Google’s Fairwind Program, with wider release later at an introductory $2/$10 per million input/output tokens.
Shopify’s Fall 2026 Edition swaps shop.app for shop.com: Shopify is trading shop.app for shop.com, the domain equivalent of upgrading from a studio to a house. The Fall 2026 Edition, its tenth, ships 150-plus updates built on 1 billion products and 400 million Shop Pay users, including AI that handles purchases and Shopify Protect for fraud. The pitch is a flywheel: every merchant and shopper added makes the network worth more to everyone else.
Shopify’s AI is now handling purchases. Here’s how to put one to work on your Monday mornings…
The 10 Best Stocks for Fall 2026
MarketBeat analysts just updated their 10 Best Stocks to Own for Fall 2026 report. The list includes AI growth names, a high-yield dividend payer, a steady portfolio diversifier and companies positioned for continued growth into year-end. Get all 10 stocks and the research behind them for free now.
THE AI ANALYST THAT EMAILS YOU A BRIEFING EVERY MONDAY
A cloud bot with a job, a saved skill and a schedule, built in one sitting.
⏱️ 20 to 30 minutes | 🔧 Needs: Grok Bot (desktop app) and a Gmail account
Why build this?
Most AI tools stop working the moment you close the chat. A bot with a job description, a saved process and a schedule keeps going without you. Grok Bot runs on xAI’s cloud, so per the video your laptop can be off and the Monday briefing still shows up. One setup turns a one-off prompt into a weekly habit you never have to remember.
Steps:
Install and sign in. Download Grok Bot from x.ai/bot for Windows, Mac or iOS, then answer the short onboarding questions.
Create a bot with a job. Hit the plus button, choose new bot, and ask for a “weekly intelligence analyst” that researches topics, uses verified information, writes concise summaries, emails them, and never invents facts or touches money or settings without permission.
Test its role. Ask it for its job title and how it handles unverified sources before you give it any tools.
Connect Gmail. Tell it to send summaries by Gmail, click the connect card, and approve the permissions. Plugins connect at account level, so every future bot can use it.
Save a research skill. Ask it to build a skill: research the past seven days from reputable sources and write a 300 to 400 word markdown report with a headline, three takeaways, a conclusion and source links.
Schedule a routine. Create “Weekly Monday research briefing” to run Mondays at 9:00 a.m. on your topic, using the skill, then email the result.
Test run it. Open the routines tab, click test run, and check your inbox. Optional: @mention a second bot to turn each report into X and LinkedIn posts.
Expected outcome:
You’ll have a research bot that wakes up every Monday at 9:00 a.m., digs through the past week of reputable sources on your topic, and emails you a short report with a headline, three takeaways, a conclusion and source links. You’ll also have a reusable skill and a second bot that turns each report into posts. Swap the topic and the same setup works for customers, competitors or your own niche.
Full tutorial:
That bot works while you’re away. This next tool works when you’re stuck, right on your screen.
TOOL OF THE DAY
An AI buddy that lives on your computer and shows you what to click next.
Most AI help ends with a wall of text you then have to translate into mouse movements. heyclicky skips that for you. It uses your screen as context, then draws on it to point you at the next move. Stuck inside FL Studio? Show it the screen and it walks you through the next step, right on top of the app. The same works in tools like Claude Code, where the hard part is knowing where to start. Think tutorial video, minus the pausing and rewinding.
