presented by moonshot
concept render
moonshot is a personal ai you wear behind your ear. preorders are $99, expected early december 2026, and fully refundable before it ships.
if i still ran my dropshipping store, gemini 4 argon would get my orders export before it got anywhere near my ads.
profit first. then revenue. then conversions.
every prompt is below so you can paste it tonight.
what google shipped, and the part i'd skip
google announced gemini 4 argon on sept 30. it's built for long, detailed work (google's examples include finance research and legal drafting), and it can read charts and long videos. so you can screenshot your ads dashboard and ask what changed week over week.
the output limit went to 1 million tokens, up from 64,000.
ngl i don't think that part matters much for a store. a profit report on 90 days of orders fits on a few pages. the chart and video reading is what i'd actually use.
google is rolling it out in stages. security partners first, then paid api customers and google ai ultra subscribers. if you can't get it yet, the text prompts here work in claude or chatgpt too. step 4 needs a model that can watch video.
profit first
step 1. export your last 90 days of orders
from shopify or whatever your store runs on, export orders with product, price, shipping, payment fees and refunds. add ad spend by product from your ad account, and cost of goods from your supplier invoices if your store doesn't track it. give argon the files with this prompt:
here are my last 90 days of orders and costs.
for each product, work out revenue, cost of goods, shipping, payment fees, refunds and ad spend, and give me profit per order.
list the 5 products losing the most money per order. show the math for each one.
for each of those 5, tell me which single cost line is doing the most damage.
flag any row where the data looks wrong instead of guessing.
line 4 is there because exports often have a few broken rows, and you want argon to point at them instead of quietly averaging them in.
and if your best seller shows up on that list, the one you've been scaling ads on since summer, you'll scroll back up and check the math, then check you exported the right dates, and it'll still say you lose money every time someone buys it.
fix the cost line it points to on those 5 before you touch anything else.
if you're not subscribed yet, subscribe here. it sends setups like this with every prompt written out.
then revenue
step 2. test one price per product
price feels like it's gonna be step 1. i'd run it second, because a price test on a product that loses money on every order won't tell you much until the cost line from step 1 is fixed.
give argon the same export plus the dates you changed prices:
for my 10 best-selling products, look at price, margin and how sales moved when i changed prices. suggest one price test per product, explain why, and tell me the result that would make me keep it or undo it.
run the tests one at a time so you know which change moved the number. i'd give each one 2 weeks and keep ad spend flat while it runs, so the price is the only thing that changed.
step 3. bring back one-time buyers
most stores have a pile of people who bought once and never came back:
from these orders, find customers who bought once and never ordered again. group them by what they bought, and draft one short follow-up email per group, written like a person, with no discount unless i ask for one.
the revenue no export will show argon
if you've ever sold wholesale, you've probably had a thread like this:
buyer: still good for the fall order at the price we said at the show
you: yes
you: 100%
you: remind me what we said
we're building moonshot (yc s26) for that part. while listening is on, it hears what gets promised out loud and brings it back before you ask, so the price you said at the show is already written down when he texts. anything it would send for you waits for your yes.
jalapeño, the earpiece we're building, is up for preorder, expected early december.
then conversions
step 4. record your checkout
on your phone, logged out and on cellular, go from a product page to the payment step the way a first-time buyer would, and screen record the whole thing. give argon the video:
watch this as a first-time buyer on a phone. list every moment where you'd hesitate or leave, with the timestamp, and the smallest fix for each. don't suggest a redesign.
google says argon is strong on long videos, and this is the cheapest way i can think of to use that. a 2-minute recording can turn up something you stopped noticing months ago, like a shipping cost that only shows up at the last step.
step 5. rewrite your best product page from your reviews
paste your top product page and your last 50 reviews:
using only what customers say in these reviews, rewrite this product page. lead with the reason people actually bought it, answer the 2 most common worries in the first screen, and keep every claim to something a review supports.
imo pages written from reviews sell better, because they answer the worries buyers actually had.
start with step 1 this weekend
tbh i don't know yet if argon beats the model you already pay for at this, so test it. run step 1 in both, check 3 orders by hand against what each one says, and keep the one that got them right.
if you're reading this on the web or someone forwarded it to you, subscribe so the next issue comes to you.
sources
google, "gemini 4 argon: our next era of frontier intelligence" (sept 30, 2026): https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/
9to5google, gemini 4 argon announcement, rollout and pricing (sept 30, 2026): https://9to5google.com/2026/09/30/gemini-4-argon-announcement/
engadget, "google's first gemini 4 model is 'argon'" (oct 1, 2026): https://www.engadget.com/2274263/google-gemini-4-model-argon/
