I am writing this on a Tuesday in the first full week of January, seventeen weeks and roughly two hundred and twenty commits into building DuranteOS. Studio — the credit-metered gateway I have been describing for two months — launches in three days. The launch is small, internal, and load-bearing for everything else I am building. By the end of this week, both DOS and Altyaa will route their AI inference through the same Bearer-token surface, the same credit ledger, the same SpendCap policy.

I am writing the positioning essay for Altyaa now, before that ships, because the moment Studio is live I am going to stop being able to think about the two products as cleanly as I can right now. Once the substrate is shared, the temptation will be to frame both products as features of the substrate. They are not. They are two distinct bets, with two distinct customer bases, that happen to ride on the same infrastructure. The shared infrastructure is what makes running both viable for a single founder. It is not what makes either of them a business.

There is a particular kind of Brazilian business that every English-language social media tool tries to serve and almost always fails.

A neighborhood bakery in Vila Madalena. A floricultura on a side street in Botafogo. An artisan coffee shop in Lapa. A small dental practice in a residential building in Higienópolis. These businesses are real, profitable, and they all post on Instagram and Facebook several times a week. The owner is the marketer, the cashier, the buyer, and the closer. The owner is also tired.

These businesses are roughly ten million strong in Brazil. They speak Portuguese. They serve customers within a 2-kilometer radius. They do not have the cash for an agency, the time for a course, or the patience for a tool whose UX assumes English-speaking power users with quarterly OKR cycles.

Altyaa is the product I am building for them. Not for the influencer with 200K followers. Not for the agency managing forty client accounts. For the bakery owner who needs to post about today's pão de queijo before noon and would prefer the AI handle the photo, the caption, and the schedule without needing to learn anything new.

This is the positioning post. It is the second product I run alongside DuranteOS. It is also the only one with a path to revenue inside the next quarter, which is part of why it gets its own essay instead of being absorbed into the DOS narrative.

Two products, one founder — why both

The reasonable question on first hearing about my two products is "why two?" The reasonable answer is that they serve different functions in my life and they share the substrate that makes both possible.

The two are independent products with independent positioning, independent go-to-market, and independent customer bases. They share one thing: the DOS Studio gateway — the credit-metered API that will route Altyaa's AI calls through the same infrastructure DOS uses, starting Friday. That sharing is what makes running both viable for a single founder. Without it I would be paying provider markups twice and reinventing rate-limiting twice. With it I get one consolidated bill across providers and one consolidated audit trail across both products.

This is the only piece of infrastructure I do not regret over-investing in.

What "wedge" actually means

In go-to-market vocabulary, a wedge is the specific narrow problem you solve so well that customers pull you in over alternatives that solve a broader problem less precisely. The wedge has to be sharper than the alternatives. The wedge has to address a buyer who is currently in pain, not one you are trying to convince.

Altyaa's wedge is not "AI for social media." That is a category, not a wedge. The category is crowded with English-language tools (Buffer, Hootsuite, Later, Sprout) that work fine for a US-based marketer who has time to learn them.

Altyaa's wedge is something more specific:

"Post what's in your shop today, in good Portuguese, with a photo that looks like your shop, on three channels, in under two minutes, without learning a new app."

Read that sentence three times. Each clause is a constraint that the broader category does not satisfy.

The right column is not exotic. It is what a Brazilian SMB owner already does mentally, expressed in software. The left column is what they get from current tools, and it is not what they want.

Why this wedge is the bet I'm making in 2026

The argument for the wedge is sharper this quarter than it was when I started thinking about Altyaa. The dispatches I have been writing for the last seven weeks have one consistent finding: the substrate is plural, the protocols are public goods, the labs are paying for retention rather than capability, and the workflow layer above the substrate is where the next billion-dollar companies get capitalized.

Resolve AI raised at a billion-dollar valuation last week on an incident-remediation-agent thesis. The framing was not "we built a better model." It was "we built the workflow that wraps the model for a specific buyer in a specific context."

That is also the Altyaa thesis. Not "AI social media." AI social media for the Brazilian neighborhood SMB owner, in Portuguese, with the platform constraints and payment rails and support channels that buyer actually uses.

The labs are not going to build this. They will not build the carioca-vs-paulistano caption distinction, the Pix-payment CTA, the WhatsApp support thread, the Instagram-square-vs-LinkedIn-horizontal image pipeline, or the four-question onboarding that assumes the owner has a phone and not a marketing degree. They will keep shipping general-purpose models that get better every quarter. The work above those models is the bet.

The four kinds of customer Altyaa is NOT for

Naming what you are not for is more useful than naming what you are. The four anti-personas:

The discipline of saying no to four large groups is what makes saying yes to the bakery owner credible.

The three things that have to be excellent

If Altyaa gets these three things wrong, the wedge collapses. If it gets them right, the wedge defends itself.

What this implies if you are building for a non-English market

Three suggestions, each costly enough that I considered cutting them and was wrong each time.

One: tune the model for the language, not for English. Translation-quality output is identifiable to native speakers within one sentence. The cheapest path is to use the LLM at face value with a translation post-step. The right path is to invest in language-specific prompts, regional variants, and a reranker against a corpus of native examples. The cost is real (weeks of work). The benefit is the difference between a tool the customer uses and a tool the customer cancels.

Two: respect the customer's existing vocabulary. Brazilian SMB owners do not say "schedule this content asset to the social calendar." They say "posta isso quarta-feira de manhã" — "post this Wednesday morning." Match the vocabulary. Translate every UX label through a fluent native speaker, not a translator app. The same advice applies to every non-English market — the vocabulary mismatch is the primary churn driver, and it is invisible to teams that only test in English.

Three: meet customers where they already pay. Brazilian SMBs pay via Pix and boleto, not credit-card subscriptions. They use WhatsApp for support, not email. They do not have a CFO to approve a USD-denominated SaaS. Build for the local payment, support, and pricing infrastructure or lose to whoever does. The substrate (Stripe, generic credit-card flows) does not solve this for you. The buyer notices.

The wedge is not glamorous. It is not a category-defining moonshot. It is a tool the bakery owner uses for two minutes a day and barely thinks about. That is exactly the bar.

Altyaa exists because I needed to demonstrate to myself that I could build something a customer would pay for while I bet on the longer DOS thesis. By the end of this quarter I want it to be doing both. The two products together are more interesting than either would be alone, and the synergy is mostly invisible — exactly how shared infrastructure should be.

Studio launches Friday. The first thing it routes will not be a DOS prompt. It will be an Altyaa caption for a paulistano coffee shop's afternoon post. That is the right shape of priority. The substrate ships when the product needs it, not before.

The bakery owner does not know they are also using DOS Studio's credit gateway. They do not need to know. They just need their post to go up before noon.

By the end of this week, it will.