The GTM Atlas disagrees with itself

Attio put fifteen operators in one place: chapter one says lead with the product, chapter five tried exactly that and converted nobody, and which one is right depends on who signs the cheque

Fifteen operators, one map

Attio published GTM Atlas over May and June: fifteen chapters, one per operator, laid out on a pannable canvas and organised along a customer journey from lead capture through to expansion. No form, no gate, no PDF. Elena Verna from Lovable on lead capture, Emily Kramer of MKT1 on qualification, Ben Thirlwall from Stripe on pricing, Cristina Cordova of Linear on hiring.

Worth saying up front what it is not. It is not written for you if you are one person shipping a small product. Kyle Norton's outbound chapter advises taking two BDR salaries and converting them into one GTM engineer. Bo Sun's expansion chapter splits a customer success org into three service tiers. Josh Epstein's retention chapter opens on who you hire on day one. That is four chapters of the fifteen solving a problem you will not have for years, if ever.

The rest is genuinely good, and the best thing in it is not a takeaway. It is an argument the Atlas has with itself and never resolves.

Chapter one and chapter five contradict each other

Verna's chapter is the loudest case for product-led anything published this year. Give the product away. Treat what freemium costs you as a marketing line item rather than a cost to squeeze down, because in the AI era it is replacing the ad spend you cannot afford. Her framing: "Your product is your lead generator. The lead gen gate is not a form: it's the experience of having that product already in people's hands." Lovable gives the product free to anyone running a hackathon, at a school, a company, a book club, and she calls it their best acquisition channel by a significant margin against any paid spend they do.

Four chapters later, Roniesha Copeland describes running that exact play at Vercel and watching it fail. Selling v0, the team started from a hypothesis she says "sounded airtight: the best way to experience the product is the product." So they custom-built a working application for the ideal customer at every target company and sent it to them.

"Open rates were very high. Conversions were zero."

Two operators, both credible, both describing the same tactic, one reporting their single best channel and the other reporting a clean nil. The Atlas prints them four chapters apart and lets it sit.

Altitude is the whole answer

Copeland's own diagnosis is the most useful paragraph in the collection, and it is not really about v0 at all:

"In a startup, the buyer and the decision-maker are the same person. The founder is using the product and owns the budget. As soon as you start targeting larger companies, those two roles get further and further apart. At enterprise scale, you're trying to reach a senior decision-maker who owns the budget, and they don't care about the product. They care that the product solves a business problem. Leading with product puts you at the wrong altitude."

So Verna is not wrong and Copeland is not wrong. Verna is describing a motion aimed at people who can decide alone. Copeland is describing what happens when you point the same motion at an organisation where the person who would love your product has to persuade someone who will never open it.

Which means the disagreement resolves entirely in your favour, if you sell to founders and small teams. There is no procurement, no security review, no committee, no business case written for someone who has not seen the thing. The user is the buyer. The hardest problem at Vercel's altitude does not exist at yours, and the advice that failed there is the advice you should be following.

It is also a ceiling worth marking on the map now. The day your customer is a fifty person company rather than a solo builder, the demo stops carrying the deal on its own, and that is a change in the job rather than a change in effort.

The second argument is about the prompt box

There is a quieter contradiction running between Verna and James Pastan at Framer, and this one bites harder if you have built anything AI-native.

Verna's case for giving everything away rests on friction collapsing: the onboarding steps that used to need a whole interface are now just a prompt box. Pastan agrees that this converts, and then names what it converts into. He calls it AI voyeurism: users acquired by curiosity about a new AI-native experience who "type something into a box and watch the model build it, but they don't actually have product intent, they're there to see what the hype is."

The sharp part is what he says next. That reaction is a real aha moment, it just is not yours: "that aha is on the model, not the product." Prompt boxes massively increase landing-to-signup conversion, and "frictionless experiences can be a trap, and there is such a thing as healthy friction."

If your signup is a text field and a spinner, this is the question to take from the Atlas: is your funnel collecting enough context from a user to make the payoff about your product rather than about the model you are calling? Pastan's activation metric at Framer is not sites published, it is publishing a site that is actually live and visited, because "zero edits before publishing isn't activation." Malhar Singh at Wispr Flow draws the same line more bluntly: the aha is the user thinking this could replace what I do today, and activation is the moment they actually do it.

Four more that survive being one person

Maja Voje's ECP before ICP is the most immediately actionable idea in the Atlas. Your early customer profile is not your ideal one. Win the segment that is obtainable now, with a burning problem, high risk tolerance and no compliance process, and earn the right to move up later. Her warning: "You might run out of runway while you're still talking to your ICP."

Emily Kramer's RAM, random acts of marketing, is the best name anyone has given the default failure. Teams "do what they've always done, copy what they see other teams do, and say yes to every founder request," without deciding which box is worth checking. Solo founders run RAM constantly, because every tactic looks equally plausible when you have tried none of them.

Copeland again, on qualification: "Enthusiasm is not the same as reason to buy." If the pain is not bad enough to fix now, you do not have a deal, you have a pleasant conversation with someone who likes talking about your product. Anyone whose Show HN went well and whose signups did not should read that twice.

Verna on frequency is the one nobody talks about. Building for a monthly-frequency use case early leaves you with "a product that constantly needs to reacquire customers who've forgotten you exist." That is a positioning decision disguised as a product decision, and it is close to irreversible.

The Atlas is an example of its own advice

Read Verna and Kramer together and the site you are reading them on starts to look like the point. Kramer: "Stop gating. It kills shareability and the data is available to you anyway." Verna: satellite apps, an actual interactive experience, are replacing the gated white paper, because nobody trades an email address for a PDF any more.

GTM Atlas is a gate-free interactive canvas that borrows the authority of fourteen operators at companies its buyers want to become, asks for nothing but an optional email at the end, and mentions Attio's product approximately never. It is the recommendation, executed.

The transferable half is not the canvas, which cost real engineering time. It is the sequencing: they decided what would be worth linking to before they decided what to build, which is the same instinct behind tactics that do not scale and work anyway.

What to do with it

Nicolas Sharp opens the Atlas by arguing that the one-size-fits-all era is over, that GTM is now a creative act, and that the winning teams will be the ones with a clear point of view rather than the right template. He is right, and it is also a lot to hand someone who has shipped a product and never run a distribution motion. "There is no universal playbook, build your own" is true and unhelpful in the same breath.

The narrowing is the work. Read Verna, Voje, Kramer, Copeland and Pastan, skip the four chapters about people you will not hire, and pick the two moves that fit your altitude. OctoLoops exists to do that narrowing: it reads your product, tells you which channels are worth your time, and gives you one distribution move a week instead of fifteen chapters and your own judgment about which ones apply.