From Farm to Mac App Store: How Heberti Built Pinery for Authors
with Heberti Almeida, Pinery
From Farm to Mac App Store: How Heberti Built Pinery for Authors
Show Notes
Heberti Almeida spent eight years in Canada - most of them as a software engineer at Square, keeping payment systems and receipt printers alive on-call. He built smart indoor greenhouses in a Toronto apartment using HomeKit and Siri. He raised three kids in cottage country outside Huntsville. And then, after years of prayer and quiet questioning about food security and sustainable living, he and his wife made a decision that most developers would find incomprehensible: they left Canada, moved to a farm in the mountains of Minas Gerais, Brazil, and started planting olives.
He also, finally, shipped Pinery. He had been building it as a side project for seven years - two hours here, five hours there - every time his full-time job allowed. Pinery is a Mac App Store self-publishing tool for authors: write in Markdown, design your book, export to EPUB, PDF, or static website. No cloud. No AI data uploads. No distractions. Under 30 megabytes, almost zero external dependencies, and a privacy architecture that means Heberti does not know who his users are or what they are writing. Three weeks after launch: 600 downloads and 40 paid subscriptions. His wife, it turns out, was right - he should have shipped sooner.
What Pinery Solves
Heberti spent five years at a major Brazilian publisher before moving to Canada, watching the traditional book production pipeline up close. A manuscript arrives in Word, gets passed to an editor, back and forth in files named "final_final_2" - then on to a designer who rebuilds it in Adobe InDesign, which forces the editor to re-read the entire manuscript again because layout changes cause text to reflow. When the InDesign file is finally exported to EPUB, it comes loaded with garbage styles and structural problems that require a separate cleanup pass. The whole cycle takes months and is entirely manual.
Pinery replaces that pipeline with Markdown at the center. Markdown carries structure (headers, bold, italic, lists) in plain text that does not corrupt when moved between apps. You write once, see a real-time preview, and customize your book's design - fonts, colors, typography from any Mac font - in the same place. When you export, Pinery produces clean EPUB 3, a digital PDF, or a static website. No InDesign. No cleanup. No file named "final_final_final." For self-published authors who want professional output without a publisher, it eliminates every bottleneck in the traditional pipeline.
5 Frameworks from This Episode
1. The Privacy-First Product Architecture
- Heberti's experience at Square handling sensitive financial PII shaped a design principle for Pinery: he does not want to know who his users are or what they are writing
- The app is offline-first and file-based - your writing stays on your device; iCloud or Dropbox sync is optional and user-controlled
- When AI features were added, the architecture did not change: Pinery uses Apple Foundation Models (on-device, macOS 26) so no text is ever uploaded to external servers
- The privacy constraint is not a limitation - it is a feature that differentiates Pinery from cloud-first writing tools in a market where authors increasingly worry about training data usage
2. The Patient Side Project
- Pinery took seven years from concept to launch - not because Heberti worked slowly, but because it was built in two- to five-hour windows alongside full-time employment
- The patient approach produced a product with near-zero external dependencies, a codebase Heberti understands entirely, and an app bundle under 30MB - qualities that come from building deliberately rather than fast
- The launch was triggered not by perfection but by his wife's intervention: "you have been working for so long and you are waiting for it to be perfect - it's never going to be perfect"
- Three weeks post-launch: 600 downloads, 40 paid subscriptions. The seven years of refinement were not lost time - they were product quality that converted on first exposure
3. The Markdown-Native Publishing Pipeline
- The traditional publishing pipeline (Word → editor → InDesign → EPUB cleanup) is sequential and lossy at every handoff - each step introduces formatting corruption and requires rework
- Markdown solves this by keeping structure in plain text: a hash symbol defines a header, asterisks define bold - the styles travel with the content without becoming invisible formatting landmines
- By building around Markdown, Pinery makes every stage of the process - writing, editing, designing, exporting - happen in one file in one app
- The open source byproduct: Heberti extracted Pinery's Markdown syntax parsing into a standalone library he published for anyone to use - the same open source ethos as his earlier Folio Reader Kit for EPUB rendering on iOS
4. On-Device AI as the Right AI
- When ChatGPT arrived, Heberti tested whether LLMs could replace what Pinery was building - hallucination rates and cloud data requirements ruled out direct integration
- Apple's Foundation Models (shipping in macOS 26) changed the equation: on-device LLMs that can proofread, rewrite for tone, and surface grammar suggestions without sending a single character off the device
- The current challenge is determinism: LLMs produce variable output, so Heberti is fine-tuning prompts to produce consistent, high-quality proofreading results within the structured output Pinery needs
- The principle: use AI that fits the product's values, not AI that is merely available - on-device AI is slower to ship but consistent with the privacy promise the product was built on
5. The Disconnected Work Rhythm
- Heberti codes in the early morning, then climbs on his tractor and works the farm - physical, disconnected labor between coding sessions
- His observation: he solves programming problems faster now than when he was doing it full-time, because the farm gives him genuine mental reset between sessions
- The Square on-call experience was the counterexample: payments not processing, printers not printing - high-stress availability with no mental recovery window
- The farm is not a distraction from the product; it is the rhythm that makes the product better - a practical case for the kind of deliberate recovery that most founders never build into their schedule
Founder Experiment: Build Something You Would Use Every Day
Step 1 - Find the pipeline in your domain that is broken at every handoff. Heberti's insight was that publishing had lossy steps between Word, editor, InDesign, and EPUB. Every domain has these: the thing that gets renamed "final_v3_actual_final" is a symptom of a broken pipeline. Map yours and identify where the most rework happens.
Step 2 - Design for your own privacy constraint, not the market average. Before adding any analytics, telemetry, or cloud sync, ask: what is the minimum data I actually need to run this product? Heberti does not know who his users are. That is not naivety - it is a deliberate design choice that protects users and simplifies the product simultaneously.
Step 3 - Minimize external dependencies before launch, not after. Pinery ships under 30MB with almost zero third-party libraries. This was not an accident - it was years of building the libraries himself. Identify the one or two open source libraries that are genuinely irreplaceable and build everything else yourself, even if it is slower.
Step 4 - Choose AI that matches your product's values. Heberti waited for on-device AI rather than shipping cloud-dependent AI features that violated Pinery's privacy architecture. Map the AI features your product needs against the privacy guarantees your product makes. If they conflict, either change the feature or change the AI - do not compromise the architecture for a feature.
Step 5 - Ship before it is perfect. Seven years of side-project work produced something stable and beautiful. His wife said: ship. He shipped. 40 paying customers in three weeks. The version that ships is always more valuable than the version still being polished. Establish a clear "stable enough" threshold and honor it.
Glossary
Tools & Resources Mentioned
Q&A
What is Pinery and who is it for?
Pinery is a Mac App Store self-publishing tool for authors who want to go from manuscript to professional ebook without a publisher or Adobe InDesign. You write in Markdown inside a focused writing environment, customize your book's design (fonts from any Mac font, colors, typography), see a real-time preview, and export to EPUB 3, digital PDF, or static website. It is built for the self-published author - novelists, technical writers, bloggers compiling a book - who needs professional output without the traditional publishing pipeline.
Why did Heberti choose Markdown as Pinery's core format?
The traditional publishing pipeline breaks at every handoff because Word documents carry invisible formatting that corrupts when moved. Markdown solves this by encoding structure in plain text characters that travel cleanly across apps and collaborators - a hash is a header, asterisks are bold, that is the entire system. A Markdown file is readable by any text editor, safe to back up anywhere, and produces clean output when converted to EPUB or HTML. Heberti's years at a Brazilian publisher watching InDesign receive mangled Word exports made Markdown feel like the obvious answer.
How does Pinery handle AI features while staying privacy-first?
Heberti's initial concern with adding AI to Pinery was that every LLM integration he evaluated required uploading user text to cloud servers - which violates the product's core design principle of not knowing what users are writing. Apple's Foundation Models, shipping in macOS 26, resolved this: on-device LLMs that can proofread, suggest improvements, and rewrite for tone without any data leaving the device. The current technical challenge is consistency - LLMs produce variable output, so Heberti is fine-tuning his prompts to get reliable structured results for the proofreading tool he is building.
Why did Pinery take seven years to ship?
Because it was built in two- to five-hour side-project windows alongside full-time engineering jobs at companies including Square. Heberti did not have continuous time to work on it - every session was carved out of evenings and spare hours over seven years. The upside: the product was refined over a long time rather than rushed, which is why it shipped with near-zero external dependencies, a 30MB app bundle, and a codebase Heberti fully understands. The launch trigger was not a milestone - it was his wife telling him to stop waiting for perfection.
What is Pinery's business model and early traction?
Free download on the Mac App Store with a subscription: $4.99/month or $49.99/year. Heberti participates in Apple's Small Business Program, which reduces Apple's commission from 30% to 15% for developers earning under $1M annually. Three weeks after launch: approximately 600 downloads and 40 paid subscriptions. Distribution has been entirely organic - a Product Hunt launch, design and writing newsletters that picked it up, and cautious Reddit posting (which Heberti notes requires "some guts" because of the community's hostility toward subscription pricing).
What is on Pinery's product roadmap?
Two major additions. First, a full on-device proofreading tool powered by Apple Foundation Models - grammar error detection, syntax suggestions, and improvement recommendations that run entirely on the user's Mac. Second, audiobook export: Pinery will be able to convert a manuscript to a full audiobook file using an on-device text-to-speech model that Heberti plans to bundle with the app. Both features extend Pinery's output formats while preserving the privacy-first, offline-first architecture.
How did moving to a farm in Brazil change how Heberti works?
He codes in the early morning, then works the farm - tractor, planting, building. The physical disconnect between sessions is not a productivity tax; it is the source of his productivity. He reports solving programming problems faster now than when he was doing it full-time at Square, because the farm provides genuine mental reset. The comparison he draws is to Square on-call: constant availability with no recovery, where a printer not printing meant someone did not get their food order and his phone would not stop. The farm gave him back his ability to leave a problem and return to it with fresh eyes.
What is Heberti building on the farm and why did he leave Canada?
He and his wife had been questioning food security - realizing that financial safety means little if you cannot produce your own food. In Canada, year-round production requires a heated greenhouse running on natural gas, which creates its own dependency. In the mountains of Minas Gerais, Brazil, the climate allows year-round growing. They are building a 240 square meter greenhouse, planting grapes, apples, and olive trees (five years from first harvest), and renovating the farmhouse - Heberti built the furniture himself. The decision was partly practical and partly spiritual: years of prayer about simplicity, sustainability, and what it means to actually be secure.
What is Heberti's take on college and formal education?
He would not enroll today. He has a business degree that he says served him for very little, and has built a career as a software engineer without a computer science degree. His point is not that education has no value - it is that the internet has made domain-specific knowledge accessible to anyone willing to learn it, and the economic case for a credential is weaker than it has ever been. If he were starting over with his current mindset, he would go directly into farming: high-quality food production is genuinely undersupplied, and the skills required are learnable outside any institution.