Aaron (@aaronshipsit) profile picture

I run a tiny AI factory — we ship a new product every month, priced like we're not greedy. Real numbers monthly, receipts attached. Now shipping: FeatureJet.

Michigan
Joined June 2011
Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)Photo by Aaron (@aaronshipsit)
I run a tiny AI factory: I orchestrate agents that build, test, and ship real SaaS. One new product a month, real numbers monthly, receipts attached. First up: FeatureJet — feedback boards at $15 flat, because we're not greedy. The streak starts now.
1
77
The Apple HomeView reviewed in one photo. This is easily the buggiest product Apple has put out in years. What were they doing the last two years after they finished the hardware? (note: they *are* on the same WiFi network).
64
China will not pause the frontier irrespective of what the US does
8
Gmail spam detection has gotten ridiculously bad. How does this get past the filters?
1
14
The advantage of building alone is not total control. It is short decision latency. Pick a direction and ship fast. Fall flat on your face if you have to, but learn from everything you do.
6
Prompts are suggestions. Hooks are policy. Use hooks for anything that must happen every time. Use skills for judgment you want to reuse. Use MCP servers only when the work genuinely needs another system. If a rule can be enforced by code, do not leave it in a paragraph and hope the model remembers. It won't end well.
7
A one-time price is a useful constraint. You cannot fund a mediocre product with six months of billing. You have to make the value clear, keep the scope honest, and ship something people would rather own than rent. That is a business model, not a pricing gimmick.
7
The honest price is the one a customer can predict before clicking buy. If the bill depends on seats, usage, exports, support, and three settings nobody can find, that is not flexibility. It is pricing designed to extract as much money as possible.
8
small teams do not need a feedback tool with 40 features. they need a list, a vote button, and updates people can actually follow. everything else is noise that makes the roadmap feel busier than it is.
8
Most wedding software is built for vendors, planners, and enterprise budgets. The couple is left juggling spreadsheets, group chats, and twelve tabs. That gap is interesting. Not because weddings are glamorous. Because a focused product can remove a very specific kind of chaos.
5
chat is where agents die slow. real work lives in your codebase, your queue, your version control. agents that only talk are agents that never shipped anything.
1
13
It’d be nice if a frontier AI model would hack an extra TW of capacity to our grid
9
Flat pricing tells customers what the product costs. Per-seat pricing makes the bill a function of your org chart, every new hire quietly raises it. One is a price. The other is an invoice that grows every time you succeed.
4
Most wedding software is built to help vendors sell, schedule, and upsell. The couple needs less admin, clearer decisions, and fewer tabs. Build for the people getting married, not the businesses billing them.
10
killing an idea early is the cheapest thing you'll ever ship. the dangerous ideas are the ones you keep past week one, that's when you start lying to yourself about them.
8
agents don't wait for meetings. they branch, they ship, they queue up the next thing. you wake up to a diff instead of a standup. the interesting question isn't whether agents can ship, it's who reviews what they shipped. the review gate is the product now.
11
magic links are the dumbest idea the internet has ever had
10
A feedback board is useful when it changes what gets built. Otherwise it is just a suggestion box with better typography. Close the loop, record the decision, and say no plainly. Customers can handle no, or at least they'll tell you why it's still important to them. Silence is what makes them leave.
1
1
15
Shipping one product a month is less about speed than refusing to hide in preparation. Thirty days gives you enough time to make something useful and not enough time to build a small monument to your own assumptions.
6
One month, one bet. The goal is not to predict the perfect product. It is to put one thing in front of real users every 30 days, then let usage, refunds, and silence do the judging. The numbers go up either way. At least this way, they are attached to something shipped.
8
One person with AI can ship what used to take a team. The trap is letting the agent decide what matters. I keep the judgment, the AI just types faster. One month, one bet. The numbers go up either way. aaronshipsit.com
11