What autonomous PPC intelligence actually means.

We did not enter the Amazon PPC software category. We created an adjacent one and named it. The label is Autonomous PPC Intelligence. The five pillars under that label are transparency, shadow mode, safety, intelligence, and aligned pricing — plus a fourth supporting pillar, EU-native, that sits inside intelligence at the architectural level and inside positioning at the buyer level.

Each pillar has its own post in this journal:

• Transparency, the first pillar
• Shadow Mode: preview the AI, then trust it
• Seven layers between an AI and your ad budget
• Sixteen agents, one objective: your profit
• Why we will never charge a percentage of your ad spend
• EU-native is not a translation layer

This post is the argument for why those pillars only work together. Read in isolation, each one looks like a reasonable claim a competitor could match. Read as a system, they constrain the product into a shape that no incumbent can copy without abandoning the architecture they have shipped.

Each pillar fails alone

The simplest way to see why the combination matters is to imagine the product with one pillar removed.

Transparency without safety is an AI burning through your budget in slow motion with a full audit trail. The trace tells you exactly which agent proposed which catastrophic bid; the budget is still gone. Transparency is necessary but not sufficient.

Safety without transparency is a set of locked gates the seller has to trust the vendor to have calibrated correctly. Pacvue's "basic safety" appears in our comparison matrix as a checkmark; it is a real feature. But it is opaque. The seller has no way to verify the gate is set where the marketing claim says it is.

Intelligence without transparency is the entire category. Sixteen agents that produce a number with no reasoning attached is functionally identical to one model that produces a number with no reasoning attached. The architecture stops mattering if the buyer cannot see the architecture.

Intelligence without safety is the worst combination. A capable AI without circuit breakers is the highest-variance possible outcome. The most-likely-good and most-likely-bad scenarios both have the largest magnitude. Sellers who survived an autonomous PPC tool's first bad week and never came back are usually surviving this combination, not the absence of intelligence.

Aligned pricing without the other four is a free tool that does nothing useful. The point of aligned pricing is to make a useful tool affordable in the long run. A misaligned tool you cannot afford is no worse than an aligned tool that does not work.

Shadow Mode without transparency is meaningless. The free tier exists to let the seller see what the AI would have done. Without traces, there is nothing to see.

The combination is necessary because each pillar is the precondition for the value of the next. Transparency makes intelligence and safety verifiable. Shadow Mode makes transparency consumable before purchase. Safety makes the intelligence trustworthy enough to act on. Aligned pricing makes the long-term relationship sustainable.

The category we created

The reason we are not "the new Pacvue" or "the new Perpetua" is that the incumbents are optimised for a different selection of pillars. Pacvue and Perpetua are deep on intelligence and enterprise integration. They are weak on transparency and pricing alignment. Teikametrics is strong on integration and pricing accessibility. It is weak on transparency and safety architecture. Ad Badger and Sellerboard are accessible on pricing. They are weak on intelligence and safety.

None of those are bad products. They serve different positions on the pillar matrix.

Mirox is the position that does not exist elsewhere: deep transparency, real intelligence, shipped safety, free-forever observation tier, flat pricing, EU-native. The label Autonomous PPC Intelligence is the name of that position. It is not a clever rebranding of Pacvue's category. It is a different position with different defensibility.

The defensibility argument is structural. An incumbent that wants to copy us has to:

• Open-source enough of their decision logic to make a transparent trace surface possible. This is a brand-level decision that contradicts their go-to-market for a decade.
• Eat the compute cost of a free observation tier. This is a P&L decision that hurts in the quarter it is taken.
• Cap their own pricing model away from percentage-of-spend. This is a revenue decision that hurts every quarter forever.
• Rebuild their safety system as an externally-auditable artefact rather than an internal best-practice. This is an engineering decision that takes a year and produces no new revenue.

Each of these decisions is rational on its own and irrational from the position of an incumbent with a working business. We made them because we did not have a working business to defend yet. By the time we do, the position is the moat.

What this means for the buyer

A seller comparing Mirox to Pacvue or Perpetua on a feature-checklist basis will find feature gaps in both directions. Pacvue and Perpetua have integrations Mirox does not (Walmart, Instacart, Google PMax) and a decade of enterprise polish Mirox does not. Mirox has Shadow Mode, decision traces, per-marketplace pCVR priors, and flat pricing — none of which any of the three incumbents offers in our 2026-Q2 comparison.

The right question is not which feature list is longer. It is which set of pillars matters most for your account.

If you are an enterprise multi-channel advertiser with €500K+ monthly spend across Amazon, Walmart, and Instacart, Pacvue is probably the right answer. They were built for you.

If you are an EU multi-marketplace Amazon seller spending €20K–€100K monthly, who has been burned by black-box optimisation, who needs to defend bid decisions to a CFO, and whose growth depends on getting German and Italian campaigns to perform as well as the English ones — Mirox is the right answer. We were built for you.

If you are between those two positions, the right answer is to run Shadow Mode for fourteen days on your real account and let the traces decide. We charge nothing for that.

What the pillars commit us to

Pillars are constraints. They make some product directions easy and others impossible.

The five Mirox pillars commit us to:

• A pricing roadmap that never adds a percentage-of-spend component, even when a partner asks. (See the aligned pricing post for the one carve-out we have considered — outcome-based pricing as an option — and why we have not done it.)
• A free Shadow Mode tier that does not become a 14-day trial, even when the conversion math gets uncomfortable. (See the shadow mode post for the cost-control levers that keep this affordable.)
• A decision trace that does not get reduced to a "summary" view as a default, even when buyer-research suggests the trace is intimidating. (We will ship summary views as options, never as defaults that hide the trace.)
• A safety architecture that ships before a feature is considered live-ready, not after. The seven layers are not optional; they are the price of admission for a feature to graduate from Shadow to Live. (See the safety post.)
• An EU-native default that we maintain even as the US market starts to ask for us. We will support US sellers; we will not retrofit our architecture to default to US behaviour.

These constraints make some moves we will not make obvious. We will not ship a lifetime deal. We will not run an AppSumo campaign. We will not buy a Reddit ad. We will not introduce a "Pro AI" tier above Scale that adds a percentage component. We will not sell white-label OEM versions to other vendors. We will not chase Walmart or Instacart as additional channels in the first three years.

Some of these are commercial trade-offs that cost us revenue. We accept the cost. The pillars are what we are selling. Every move that contradicts them devalues the entire product line.

Where this leads next

The five pillars are locked. The product is shipping. The founding cohort is open until ten seats are filled. The public Shadow Mode tier is on the calendar for Q3 2026.

If you are reading this and thinking this is the position I was hoping someone would build — apply at mirox.pt/beta or join the waitlist at mirox.pt/waitlist. Both are still real links and the founder reads both.

If you are reading this and thinking this is too academic, just tell me the ACoS impact — fair. Read the data piece on where 30 to 40 percent of your Amazon ad spend actually goes instead. The number is on your account, not in this post.

If you are reading this and thinking I want to argue with one of these pillars — write back. The five posts above are the long form of why each is locked. We have changed our minds before. The argument that changes them again will come from a seller telling us, with their data, that one of the pillars is in the way of their work.

That is the relationship Mirox is trying to be in. Five posts, five pillars, one cohort, no asterisks.