When writing code is free
More of your code is AI-written now — and you still ship it under your name. But the agents keep churning out fixes, the diff looks pretty good, and “pretty good” quietly becomes “ship it” — without anyone actually using it. Macro makes using it instant.
From the team behind Sunsama.
The empty states feel a bit sad — give them some life.
Added a friendly illustration and a one-line CTA to each empty state.
The new bottleneck
AI flipped the economics. Producing code is heading toward free; the bottleneck is being sure it feels right. And there’s a law underneath it: if verification is hard, people skip it. Hard to write tests, they don’t get written. Hard to use the change, it gets shipped on a glance at the diff.
Your product quality isn’t capped by the verification available to you. It’s capped by the verification you’ll actually do — and as output climbs, the verification you skip is the quality you lose.
What each check can reach
Verification has two dimensions: the code vs. the product, and correctness vs. taste. AI is taking three of the four corners. It can’t take the last one.
Does it do what we said it would?
AI: kindaDoes it feel right? The last 10%.
AI: noThe human’s corner — Macro lives here
Does it run? No bugs, no regressions.
AI: yesSimple? Well-abstracted? Or just more code?
AI: kindaFor a high-taste product, the last 10% is the unknown-unknowns — the menu that’s now too crowded, the flow that feels off — and they only show up when you use the thing. AI is taking correctness. Taste is what’s left, and for a product like yours it’s the whole game.
A note from the founder
At Sunsama, nothing shipped until I’d used it. Every button, every interaction, every edge I could think to push — and not because I didn’t trust the tests. Tests are great. But a passing suite and a product that feels right are two different things, and only one of them is the product.
That instinct is the whole reason Macro exists. AI writes more of the code now, but the question I care about hasn’t changed: how does it actually feel to use this? You can’t get that from a diff — you get it by using what you built. I can feel myself getting tempted to ship small changes an agent probably got right without really trying them. So I wanted testing what my agents build to be as easy as jumping into a Slack chat — and helping them land that last 10% that makes a product wow customers.
I’ve spent the last decade building tools that help modern workers stay sane and focused. I want to agentmaxx AND stand behind what I ship.
How it works
Install the GitHub app and Macro reads your repo, detects the stack, boots a full isolated dev environment for every agent, and wires the tooling they need. The setup that normally takes days is finished before the first agent runs — and you and the agent share the same runtime, so there’s nothing to reproduce by hand.
What it is
Macro takes care of all the annoying infra work for you. Landing an agent’s final details and using the app feels as easy as sending a couple Slack messages.
Install the GitHub app; Macro reads the repo, detects your stack, boots the environment, and wires the tooling. The setup that took days is already done.
Own RAM, own dev server, own branch, own live preview URL. A hundred can run at once and none touches your laptop, your ports, or your place in the code.
Stable public URLs for every environment, so logins, OAuth callbacks, and webhooks just work — no ngrok, no tunnels to babysit.
Teleport into any agent’s environment and take the keyboard. It’s a real dev box — pick up where it left off, hand it back when you’re done.
Linear, GitHub, Intercom, your own tools — wired on boot via MCP, so agents work against your real systems, not a mock.
Set the stack, the secrets, the agent types, and who can do what — once, for the whole team.
Preview environments
Each change runs on its own environment with a real, shareable preview URL. Click through it like a customer, send it to a teammate, catch the thing the diff couldn’t show you. It’s the running app — not a screenshot, not a recording.
one per change — open any of them and use it
Who this is for
For a commodity surface, a diff and a passing test is plenty. For a product where the last 10% is the brand, shipping on “the diff looked fine” is how the polish quietly erodes. Only one rung of verification reaches taste, and it’s the one Macro makes free.
You’re putting your name on code you didn’t write. Standing behind it used to mean you wrote it, or read every line. Now you didn’t.
Macro is how you earn the right to stand behind it anyway — because you used it.
Macro is the easiest way to playtest, iterate, and land your agents’ changes in a way you can stand behind. In the time you’d usually spend unblocking colleagues on Linear, Slack, or GitHub, you can now help your agents get across the finish line. It doesn’t replace your tests or your judgment — you’re still the one who looked, and the one who shipped it.
Macro is in private alpha. Tell us what you’re building and we’ll get you in.