Zero to Launch · Session 02 · Wed Aug 19, 2026
▄▄███▄▄·██████╗ ██╗ ██╗ ██╔════╝╚════██╗██║ ██╔╝ ███████╗ █████╔╝█████╔╝ ╚════██║ ╚═══██╗██╔═██╗ ███████║██████╔╝██║ ██╗ ╚═▀▀▀══╝╚═════╝ ╚═╝ ╚═╝
Zero to $3,000
B2C, with the dashboard open
Ship AIPhoenix, ArizonaPress → to begin
What Ship AI is
Builders showing each other the work.
- Builders & founders
- Phoenix
- Free, always
The best AI education isn't behind a paywall or on a stage — it's in the open, for free.
Your host
Santos Hernandez
Founder & Host
Founder and Lead Product Engineer building agentic AI systems. Founding product hire at ZBD — the money layer for games — $0 to $12M ARR, the EU's first MiCAR approval, and money transmitter licenses in 26 states and D.C. He started Ship AI to give Phoenix builders a room where you show the work, not talk about it.
Tonight
- 01The brief — seven stories, then a five-minute break
- 02Zero to $3k — the case study, with the dashboard open
- 03The knowledge share — the same structure, pointed at your product
- 04Q&A — anything with a number in it
- 05Next event
Where we are
- Aug 5GTM Engineering
- Aug 19Zero to $3,000
- Sep 2Outbound Agents
- Sep 16Unfair Advantage
- Oct 7Ship the Surface
- Oct 14Agentic Growth
- WeekendHackathon
Six Wednesdays, then the weekend. Every session is free and standalone. Join late and you can still compete in October.
Community partners
WorkuityPlatinum sponsor · Venue
CEI GatewayCommunity partner
Venture Café PhoenixCommunityThe rooms, the nights, the coffee. Ship AI is free because these three show up for builders.
Current sponsors
desicGold · AI deployment
AutomationInterns.comGold · Everyday AISponsors
This room is free because somebody funded it
Ship AI is seeking sponsors. Tiers run $1,000 to $10,000. Cash, credits and hours all count the same.
shipai.club · Free, always
Segment one
The brief
Seven stories from the last week, and the argument under each one. Then five minutes, then we open the dashboard.
The brief · 01 / 07
Grok 4.6 ties Opus 5 Max on agentic work
@changis_k · Artificial Analysis
- Both score 59 on the Agentic Index. It measures tool use, planning and autonomy, not single answers.
- Grok finishes a task in about 53 turns. Opus 5 Max takes about 103.
- $0.84 per completed task.
You pay per finished task, not per benchmark point. Which of your agents survives that maths?
The brief · 02 / 07
Claude Code can design now
- A /design skill in research preview, in the CLI and the desktop app.
- Editable artboards for your interface. Pick one, change it, then have Claude build it.
- Design moves into the same window as the code.
Who still opens a design tool before they open an editor?
The brief · 03 / 07
Whop goes native inside Grok
- Whop is now a native connector in Grok.
- The pitch is turning tokens into economic value.
- Distribution is moving inside the assistant.
If your product were a connector in somebody's assistant, what would they ask it for?
The brief · 04 / 07
Cursor runs Git like a database
- They built their own Git storage, called Origin, and run it like a database.
- The post traces twenty years of Git infrastructure to explain why.
- Agents write far more commits than people do. The old assumptions break.
What in your stack was sized for human traffic and now takes agent traffic?
The brief · 05 / 07
Grok Bot ships quality-of-life fixes
- Mobile notifications now group by bot and carry their own icon.
- Small, unglamorous, shipped.
- This is most of the work.
What is the smallest fix on your list that users would actually notice?
The brief · 06 / 07
Two labs took 43% of all venture dollars
- AI captured more than 70% of global venture funding in Q2.
- OpenAI and Anthropic together: $217B, about 43% of every venture dollar in the quarter.
- Infrastructure keeps clearing too. Fireworks AI raised $1.5B, Together AI $800M.
None of that money is available to you. So what is the advantage of being small?
The brief · 07 / 07
Phoenix: MiiHealth raises $2.8M seed
- An AI patient-intake assistant, built here, funded this month.
- Not a frontier lab. One workflow, one regulated industry, sold to people who have that problem.
- That is the shape of company this room can build.
What boring workflow do you already understand better than almost anyone?
Intermission
05:00
Five minutes. Get a refill, meet somebody you didn't come with. When the bar empties we open the dashboard.
Segment two
The case study
One product, one dashboard, no redactions. What it made, what it lost, and what it took to climb back.
Case studies are usually told by someone with an outcome to sell.
Told with the analytics open, including the money wasted on what didn't work.
First 30 days
$3,318
Feb 7 to Mar 8, 2026, straight off the dashboard. No ad spend. The first paid dollar went out on July 14.
The first thirty days
$3,318
day 30 · cumulative
Cumulative from the day it shipped. $159 by day seven, $3,318 by day thirty. Look at day seven before you panic in your own first week.
The trough, seen from August
$15,504
August MTD · banked, Aug 1–17
$30,401
projected EOM · not banked
Monthly revenue, complete days only. Solid is banked. Dashed is a forecast: the trailing 7-day average of $1,064/day held to the 31st. At this scale February is a bump and May's $303 is flat.
It isn't your graph. It's the graph.
Paul Graham named this shape twenty years ago. Cal.com posted theirs at $10m ARR this week, h/t @peer_rich. The trough is where most people quit.
Where it is now
$1,064
revenue per day, 7-day average
$61
ad spend per day
68%
of this month's installs organic
Aug 11–17: $7,449 of revenue on $426 of spend, blended. The product had already done $4,949 before the first paid dollar on July 14.
The organic half
Thirty days, no budget
What the channels actually were, in what order, and how much of it was repeatable versus lucky.
The arc, in order
- Manualby hand until it works
- One channelsequence beats coverage
- Repeatablemethod vs. lucky post
- Then paidonly once the maths held
Find and onboard the first users one at a time. Automating a guess only makes the guess arrive faster.
One channel at a time, fed until it repeats or clearly doesn't. Run two at once and you can't tell which one moved the number.
Then the honest audit. Which part was the method, and which part was one lucky post? You scale the method.
Paid came last, once the funnel converted organically and payback was a number I could say out loud.
Thirty days, four moves. None needed a budget. They needed doing in this order.
The shipping sequence, in order
- Store prepEverything that has to be true before you submit. Review turnaround is a schedule input, not a surprise.
- ASOThe listing is a ranked surface, not a formality. Title, subtitle, keyword field, screenshots.
- Social pagesSet up where the audience already is. Two pages done properly beat six abandoned.
- Marketing siteThe one asset every channel points at. Oct 7 builds it properly.
- Articles, SEO, pSEOWritten once, ranking for months. Programmatic pages where the data supports them: one template, many queries.
- Automated contentDrafts off competitor coverage, scheduled and distributed. A pipeline, not a habit.
- Friends and familyThe ask nearly everyone skips. Your first reviews and installs come from people who already know you.
The same sequence, with the real dates on it
- Nov 2025Marketing site. A landing page existed months before there was an app to land on.
- Jan 4Twenty-two articles, written for search rather than for us.
- Jan 21Social automation: generated images, scheduled to five platforms.
- Feb 3Programmatic pages. Five template families, four locales.
- Feb 7The app goes live in both stores.
- Jul 14First paid dollar. Five months and one week after shipping.
Read the gap. Every distribution machine was running before the product existed, and paid came five months after that. The order is the lesson, not the dates.
The SEO half, mostly not typed by hand
/seo-audit → /programmatic-seo
35 articles on disk5 template families × 4 localesrank proof: none yet
Four skills off skills.sh: seo-audit, programmatic-seo, ai-seo and content-strategy, from coreyhaines31/marketingskills. The pages exist. The rankings aren't proven.
Automated content, end to end
Drafting
- Competitor coverage as the input
- AI SDK does the drafting
- Every claim marked verified or not
- A human reads it before it goes out
Distribution
- Postiz, until Jul 22
- Then direct, in our own service
- Now moving to Zernio
- Three publishers in seven months
How the old machine actually worked
- TemplateA post template with an image prompt and variables: tournament banner, prize pool, activity.
- Brand contextColours and style injected into every prompt automatically, so it all came out looking related.
- Nano Banana Progoogle/gemini-3-pro-image through the Vercel AI Gateway. Flux 2 Pro as fallback, Gemini Flash as the cheap seat.
- PostizImage buffer uploaded, post created, scheduled. No human between the model and the calendar.
- Five platformsFacebook, Instagram, X, YouTube, TikTok — on a schedule, indefinitely.
Five steps, no approval gate anywhere. That's the whole design, and it's why it produced 783 posts.
Volume does work. For people with something to say.
| Followers | Brand searches / mo | |
|---|---|---|
| 5 posts a week | 5,194 | 302 |
| 10 a week | 8,042 | 634 |
| 15 a week | 11,587 | 591 |
| 25 a week | 19,531 | 2,833 |
| 26+ a week | 53,002 | 6,037 |
Neil Patel's numbers, published this week. Cadence really does compound. We posted 783 times and got none of this, because the posts were about nothing in particular.
Two machines, one idea
SEO
- One template, many queries
- 5 families × 4 locales
- 35 articles written once
- Ranks while you sleep
Social
- One brief, many posts
- Images generated, not commissioned
- Scheduled to five platforms
- Posts while you sleep
Leverage is work you do once that keeps paying. Both machines are cheap to build now and neither one cares how big you are. That is the whole advantage you have over a company with a marketing department.
The compounding half
Community and retention
Paid stops the day you stop paying. This doesn't. Neither does a leak.
Community as a growth loop
- 01A user gets real value
- 02They're given somewhere to say so
- 03Superfans become ambassadors
- 04Their reach brings new users
It's a loop with a conversion rate, not a Discord you open and hope. Ours: 51 referrals started, 5 completed, no ambassador programme yet.
The top of the funnel is four buttons
- Google and Apple are one tap. No password, no inbox.
- Email and phone both cost you a code and a context switch.
- Every provider you add is a reason not to bounce here.
This screen is the first thing every ad you buy pays for. Four options rather than a form is the cheapest conversion work available, and it is done once.
An event per stage, verified firing
- Installedsession_started
- Registeredsng_complete_registration62%
- Activatedgg_first_reward_earned50%
- Repeatinggg_reward_earned58%
- Cashed outshop_purchase39%
Our actual events, not a website funnel. The middle three are what Google bids on. Take a baseline today; in eight weeks it's the before half of your October story.
Fix the leak before you turn on the tap
- Jul 20913
- Jul 27785
- Aug 31,316
- Aug 101,430
Four weekly cohorts, day one to day seven. Aug 3 is the last closed window, so read that one. Aug 10 is dashed and will still move.
The paid half
Then paid
It came last, on purpose. What had to be true first, what it actually bought, and where the ceiling really sits.
Spending before the unit economics were clear would have killed it.
Paid doesn't find product-market fit. It buys more of whatever you already have, leak included.
The switch to paid
- Preconditionsthree things, all true
- Angles, not variationsfive reasons, not five wordings
- The daily loopraise · hold · cut
- The ceiling$61/day, and not by choice
A funnel converting organically, a payback period I could state, and a number I'd stop at. All three written down before the first dollar.
Five wordings of the same thing is one test. Five reasons to care is five. The angles come out of the value props, not the ad account.
One look a day. Three decisions, judged against a threshold you set in advance.
August: $1,098 of spend against $15,479 of revenue. Android is capped at $61/day. Google limits the ad groups over the video claims, and Meta has sat on a billing hold since Aug 13.
Paid didn't create the business. It bought more of a funnel that already worked.
The other half
What didn't work
The same length as what did. This is the half nobody publishes, and the half that saves you a month.
Every paid channel, lifetime
| Spend | Installs | CPI | |
|---|---|---|---|
| Google Android | $1,847 | 3,132 | $0.59 |
| Apple Search Ads | $21 | 44 | $0.49 |
| $144 | 87 | $1.65 | |
| TikTok | $60 | 15 | $4.01 |
Google does the volume. Apple Search Ads is the cheapest install we buy, on almost no money. TikTok cost seven times Google per install and was killed on day six.
The rest of what didn't work
- The onboarding rewriteEight steps at ~81% completion became five steps at ~57%. Step-one abandonment went 6.6% to 33.8%. Shorter is not easier.
- Reverting didn't fix itCompletion went 68% to 58% anyway, with region blocks and network failures underneath. The rewrite wasn't the only thing broken.
- The reward-progress card94% of the people who tapped it could not afford the reward. The button's main job was disappointing people.
- The sub-$5 cashout testMeasured nothing. A provider sync wiped the treatment group mid-test, and it ended on 2 redemptions and $4. We shipped the $5 floor on judgement.
- The Google video flag13 of 14 videos limited for exaggerated claims. Recutting the copy changed nothing. The claim Google read was inside our own app UI, in a screenshot.
- Play prominent disclosureThe ad SDK could start before consent. Correct finding, our bug, fixed in a later build. The appeal is still open.
Segment three
Your turn
Same structure, your product. One channel, one test, an event per stage. Receipts are thirty percent of the October score.
Your first channel
/channel-plan
reading 01-roadmap/README.mddrafting test, budget, go/no-go…wrote 02-b2c/README.md
One channel, the test, the budget, and the go/no-go number. Decide them before you spend.
Now point the same structure at your own product.
First channel, the test, the budget, the go/no-go number. Four lines. Write them tonight.
Next program
Growth Loops
Tonight is acquisition. Growth Loops is the other half: the value event, the path to first value, cohorts, and one proven behaviour turned into a loop. Six sessions, dates soon.
shipai.club · Free, always
Q&A
Ask me anything with a number in it.
- What did it cost?
- What broke?
- What would you do first?
The dashboard is open and nothing is off the table. Spend, retention, the appeal, the parts I got wrong.
Thank you
Dan & Workuity
Platinum sponsor · The venue
Workuity Biltmore hosts and sponsors every meetup in this series and the October weekend. Rooms like tonight don't happen without Dan and this community — thank him on your way out.
Thank you
AutomationInterns.com
Gold sponsor
Affordable AI for everyday business — and the newest name on the Ship AI sponsor wall. Welcome Jon S and the interns.
Sponsors
Your logo belongs on this wall
If tonight was useful, this is how it stays free. Every tier is published and every arrangement is negotiable — cash, credits or hours, at every price point. Talk to Santos before you leave, or scan for the menu.
shipai.club · Free, always
Say hi
- X@5antoshernandezBuild in public. Arguments welcome.
- Discorddiscord.gg/kZSJMNveYMWhere tonight's links, slides, and skills land.
- Work with medesic.xyzWant systems like these built and run for you? That's my team.
Everything is open source: the deck, the skills, the template repo. Take it all.
Just ship it.
Next
Outbound Agents — Wed Sep 2, Workuity Biltmore
The B2B counterpart. An ICP tight enough to write the email to, twenty named accounts, and a sequence built live from a blank page.
shipai.club · Free, always