<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Bmartin2000</title>
    <description>The latest articles on DEV Community by Bmartin2000 (@bmartin2000).</description>
    <link>https://dev.to/bmartin2000</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4175744%2F0c1ef84c-8699-4f4b-ac1b-3d389902760b.png</url>
      <title>DEV Community: Bmartin2000</title>
      <link>https://dev.to/bmartin2000</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/bmartin2000"/>
    <language>en</language>
    <item>
      <title>Code Your Reality: film how your business works, get your own app (built on Gemma 4)</title>
      <dc:creator>Bmartin2000</dc:creator>
      <pubDate>Sun, 18 Oct 2026 18:12:23 +0000</pubDate>
      <link>https://dev.to/bmartin2000/code-your-reality-film-how-your-business-works-get-your-own-app-built-on-gemma-4-2m9p</link>
      <guid>https://dev.to/bmartin2000/code-your-reality-film-how-your-business-works-get-your-own-app-built-on-gemma-4-2m9p</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;Here's something I see all the time: a company is a mirror of its founder's head. If that head is a mess (sticky notes, half-remembered promises, "I'll just call Peter"), the company slowly turns into the same mess. Orders get lost, two people book the same van, nobody knows which spreadsheet is the real one. Nobody notices it happening, but it quietly strangles the business.&lt;/p&gt;

&lt;p&gt;The textbook answer is a CRM or an ERP. And yet most small and mid-sized businesses never get one, for good reasons:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Fitting it to your company takes forever.&lt;/strong&gt; Off-the-shelf systems make you work their way, not yours.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Building your own takes even longer&lt;/strong&gt;, and it's expensive.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;And every change starts the whole thing again:&lt;/strong&gt; another quote, another wait, more time, energy and frustration.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;So they stay with the chaos, because the cure looks worse than the disease.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;But what if…&lt;/strong&gt; you just took out your phone, opened an app, and filmed how your company actually works, talking through it like you'd explain it to a new colleague? And out came a complete system, built around &lt;em&gt;your&lt;/em&gt; process, that you can start using right away?&lt;/p&gt;

&lt;p&gt;That's &lt;strong&gt;Code Your Reality&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;And when it isn't quite right, or your business changes, you don't start over. Record a new video, describe the change in one sentence, or edit it by hand. The &lt;strong&gt;Plus&lt;/strong&gt; plan gives you room to grow, and real people behind the project will help you along the way.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;🎬 &lt;strong&gt;Demo video:&lt;/strong&gt; &lt;a href="https://www.youtube.com/watch?v=i4IlB6sVqO8" rel="noopener noreferrer"&gt;https://www.youtube.com/watch?v=i4IlB6sVqO8&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;💻 &lt;strong&gt;Code:&lt;/strong&gt; &lt;a href="https://github.com/Bmartin2000/code-your-reality" rel="noopener noreferrer"&gt;github.com/Bmartin2000/code-your-reality&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  How it works, in a bit more detail
&lt;/h3&gt;

&lt;p&gt;You film the job once (a clerk renting out a camera, a workshop lending a drill, a repair shop taking in a bike) and say out loud what you check and what must never happen. Google's open-weight &lt;strong&gt;Gemma 4&lt;/strong&gt; model looks at frames from the video and reads the transcript. It then proposes what the business keeps track of, how work moves along, and which rules it follows. You keep or drop each finding in plain language, press &lt;strong&gt;Build my app&lt;/strong&gt;, and get a real application: forms, lists, a step-by-step lifecycle, and business rules that the &lt;strong&gt;server enforces&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;Try to rent the same camera twice and the app refuses, quoting the rule that came from the video. The app works like a small CRM: a pipeline board, filters, notes on records, and spreadsheet import that goes through the same rules. You can rename sections, fields and steps in your team's own words.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Who pays for it, and why.&lt;/strong&gt; I work at Integrint, a Hungarian digital agency. Our smallest clients are rental shops, workshops and event-equipment companies, and they are exactly the businesses described above. They would love custom software, but they can't write a specification, and they can't afford weeks of meetings to produce one. What they &lt;em&gt;can&lt;/em&gt; do is show us how they work. So the pricing is simple:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Free demo&lt;/strong&gt;: build one app from your own process, no card.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Plus, €19/month&lt;/strong&gt;: 30 analyses a month instead of 3, so you can keep re-recording as you improve. Deeper analysis (12 frames instead of 6), reports, colleagues with roles (owner, manager, staff, viewer), a calendar feed and webhooks, and up to 5 apps.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Integrint Care&lt;/strong&gt;: we record and maintain it with you.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Subscribers see their period, usage and invoices. Invoices are issued through Számlázz.hu, Hungary's most common invoicing service, once the operator connects it. Until then they stay clearly marked drafts, and this preview takes no payments. On Plus, the generated app can also invoice the business's own customers, straight from a record.&lt;/p&gt;

&lt;p&gt;That's less than an hour of developer time per month, for software that matches how the business actually works, and keeps matching it as the business changes.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbumjgmfvnj8gsoiwkib8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbumjgmfvnj8gsoiwkib8.png" alt="Plain-language review: what the app tracks, how things move along, and rules quoting the moment they were said" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;The launch film tells the story of a company tangled in its own workarounds. Halfway through, the film itself turns out to be the recording: the app's real recorder appears over the owner's shot, the video is sent, and from there every app scene is a real screen recording of Code Your Reality analysing that clip with Gemma 4 and building the app.&lt;/p&gt;

&lt;p&gt;You can run it yourself in a few minutes on Windows, macOS or Linux. The guided example works &lt;strong&gt;without any AI model&lt;/strong&gt;, so you can see the whole journey right away: review, build, rent, get refused, return. With Ollama and Gemma installed, record your own process.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8afie9cg1isssdrmbmqz.png" alt="Record with the phone camera" width="780" height="1688"&gt;&lt;/td&gt;
&lt;td&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fk71cvexpjcwkccbaztdj.png" alt="The app is ready: open, download, share, record again, ask for a change" width="780" height="1688"&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpep6vdgearncaza0mcir.png" alt="The generated rental app" width="800" height="500"&gt;&lt;/td&gt;
&lt;td&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg9fc3kzywhuisaz6t3kz.png" alt="The server refuses renting the same item twice" width="800" height="500"&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/Bmartin2000" rel="noopener noreferrer"&gt;
        Bmartin2000
      &lt;/a&gt; / &lt;a href="https://github.com/Bmartin2000/code-your-reality" rel="noopener noreferrer"&gt;
        code-your-reality
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Reality Compiler — turn an observed business process into a working application. Local open-weight AI, human-verified rules, schema-driven runtime.
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;Code Your Reality&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Film how you work. Get your own app.&lt;/strong&gt;
&lt;em&gt;By Integrint. Don't describe your workflow. Show it.&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;Code Your Reality turns a recorded business process into a working application. A small business owner films how the work is done with a phone. &lt;strong&gt;Google's open-weight Gemma 4 model, running locally through Ollama&lt;/strong&gt;, proposes the process structure: what is tracked, how things move along, and the business rules. Each finding is traced back to the second of the recording it came from. The owner keeps or drops each finding in plain language. The approved workflow then compiles into a schema-driven application that actually runs: records persist, and the server enforces the business rules on every operation. When the business changes, the owner records again or asks for a small change in one sentence.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Customer journey:&lt;/strong&gt; see how it works → pick a plan (free pilot) → record with the…&lt;/p&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/Bmartin2000/code-your-reality" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;&lt;strong&gt;Stack:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;FastAPI and SQLite on the backend, React and TypeScript on the frontend, plus FFmpeg, faster-whisper and Ollama.&lt;/li&gt;
&lt;li&gt;One Docker image for servers (HTTP or HTTPS on port 7413), plain &lt;code&gt;pip&lt;/code&gt;/&lt;code&gt;npm&lt;/code&gt; for local use.&lt;/li&gt;
&lt;li&gt;The frontend can also be hosted separately, on any origin or in a subfolder.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Sign-in protects the Studio, so nobody can use the speech-to-text or the model anonymously on a public server. The first account becomes admin and adds the rest.&lt;/p&gt;

&lt;p&gt;The repo has 50 backend tests plus frontend unit tests and Playwright end-to-end tests. The end-to-end tests include a full "review → build → rent → refuse → return → reload" journey.&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;The AI pipeline&lt;/strong&gt; (all open source or open weights):&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;FFmpeg&lt;/strong&gt; extracts a 16 kHz audio track and samples 6 still frames evenly across the video (12 on Plus).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;faster-whisper&lt;/strong&gt; (CPU, int8) transcribes the narration into timestamped segments.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gemma 4 through Ollama&lt;/strong&gt; reads the frames and the numbered transcript segments in &lt;strong&gt;two constrained passes&lt;/strong&gt;:

&lt;ul&gt;
&lt;li&gt;
&lt;em&gt;Pass A&lt;/em&gt; proposes the data structure: entities, typed fields, references, and the lifecycle of the main transaction.&lt;/li&gt;
&lt;li&gt;
&lt;em&gt;Pass B&lt;/em&gt; proposes rules and actions. Its JSON schema is &lt;strong&gt;generated from Pass A's answer&lt;/strong&gt;. Every rule kind (precondition, exclusivity, effect on a linked record, minimum value, date order) is its own &lt;code&gt;anyOf&lt;/code&gt; branch, whose parameters are enums of the entities, fields, states and events that actually exist. The model &lt;em&gt;can't&lt;/em&gt; reference a field it didn't define.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;Every finding cites the &lt;strong&gt;segment numbers&lt;/strong&gt; it came from, and the server turns those into timestamps and quotes. A finding that claims to come from the video but cites no segment is downgraded to "AI guess, check this".&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;strong&gt;The part I'm proudest of: no generated code ever runs.&lt;/strong&gt; The model only produces JSON constrained by a schema. That JSON is:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;normalized (identifiers re-derived, references checked, unmappable rules kept as documentation);&lt;/li&gt;
&lt;li&gt;reviewed by a person;&lt;/li&gt;
&lt;li&gt;validated and compiled into a manifest.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A small runtime interprets a &lt;em&gt;closed&lt;/em&gt; set of rule kinds, so the app is safe by construction. The AI decides &lt;em&gt;what&lt;/em&gt; the rules are, never &lt;em&gt;how&lt;/em&gt; they execute.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Small change requests&lt;/strong&gt; use the same idea. One sentence becomes at most five operations from a fixed set (add a field, make it required, add an option, add or remove a rule…), each constrained by a schema built from the current app. "Build me a full accounting system" is refused, not guessed.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Honest numbers.&lt;/strong&gt; I tested on a 48-second narrated clip about lending tools in a community workshop, deliberately not the rental domain of the built-in example.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Gemma 4 E4B, locally on my laptop&lt;/strong&gt; (Intel Core Ultra 7 155U, 15 GB RAM, no GPU): about 8.4 minutes for the two passes.

&lt;ul&gt;
&lt;li&gt;It found Tool, Member and Loan with correct references, put the lifecycle on the loan, and cited a transcript timestamp for all 9 rules.&lt;/li&gt;
&lt;li&gt;Two rules mapped straight to enforceable rules: the deposit can't be negative, and a new loan records today's date.&lt;/li&gt;
&lt;li&gt;One proposed effect was wrong, which is exactly the kind of thing the review step catches. The rest were kept as notes for the owner to confirm.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A change request&lt;/strong&gt;, "members should also give an email address, and it is required": 50 seconds, and exactly one new required field.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gemma 4 31B on Ollama Cloud&lt;/strong&gt;, same clip (transcript and 6 frames): 7.7 seconds. All 6 rules came back as enforceable kinds, each citing its transcript segment, from "a tool can only be lent to one member at a time" to "returning it turns the card back to available".&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Ollama Cloud doesn't support structured outputs, so there the schema goes into the prompt instead. The server validates the answer against it. If the answer doesn't fit, the server asks for at most two repairs that name exactly what was wrong. This run fit on the first try.&lt;/p&gt;

&lt;p&gt;Small local models are good at structure and less consistent at rules. That's exactly why the human review step and the evidence quotes exist.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Because the customer's video is their business.&lt;/strong&gt; A recording of how a shop works shows its customers, prices, deposits and habits. With an open-weight model running in Ollama, the analysis happens on the machine that runs Code Your Reality, so the video never has to leave it. For a privacy-minded small business, and for GDPR in the EU, that's the difference between "maybe" and "yes". When a server has no GPU, the same code switches to Ollama Cloud with two environment variables (the URL and an API key, kept on the server). The app then tells the user honestly that the transcript and a few still frames are sent for analysis.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Because the business model only works if inference is cheap and predictable.&lt;/strong&gt; A €19/month product can't pay per-token prices for long video prompts and two analysis passes every time someone re-records. Running Gemma locally makes the marginal cost of an analysis close to zero, so the price covers support, not tokens.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Because the open stack made the safety design possible.&lt;/strong&gt; Ollama's structured outputs let me compile a &lt;em&gt;grammar&lt;/em&gt; from the first pass's answer and force the second pass to stay inside it. I could also inspect and swap the model (I tested Qwen 3.5 4B and Gemma 4 E4B side by side on the same clip), set its context and temperature, and turn its "thinking" off for extraction. A closed API wouldn't have given me that control.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;And because no one gets locked in.&lt;/strong&gt; The model is replaceable and the runtime is mine. Every customer can download their app definition and all their data as spreadsheets at any time.&lt;/p&gt;

&lt;h2&gt;
  
  
  My Agent Session
&lt;/h2&gt;

&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Gemma&lt;/strong&gt;: Gemma 4 is the default model. It reads the video frames and the transcript in two schema-constrained passes, and it handles the one-sentence change requests: locally as Gemma 4 E4B, or as Gemma 4 31B on Ollama Cloud.&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;em&gt;AI disclosure: I built this project with the help of Claude Code, an AI coding assistant, and drafted this post with AI assistance. I reviewed, tested and edited everything, and the numbers above come from my own runs. The launch film's cinematic shots and narration were generated with AI video and voice tools. Its app scenes are real screen recordings. The product's own AI is Google's open-weight Gemma 4 model running through Ollama.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>opensource</category>
    </item>
    <item>
      <title>AFK.exe: Grow a Living Office by Stepping Outside with Local AI</title>
      <dc:creator>Bmartin2000</dc:creator>
      <pubDate>Sun, 11 Oct 2026 16:33:51 +0000</pubDate>
      <link>https://dev.to/bmartin2000/afkexe-grow-a-living-office-by-stepping-outside-with-local-ai-1eod</link>
      <guid>https://dev.to/bmartin2000/afkexe-grow-a-living-office-by-stepping-outside-with-local-ai-1eod</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week1-2026-10-05"&gt;Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;AFK.exe&lt;/strong&gt;, a cozy developer game where your workspace grows when you step away.&lt;/p&gt;

&lt;p&gt;You begin in a small 3D office. Sit at the desk, open the monitor, and work through short coding contracts. Your tiny plant-headed companion, &lt;strong&gt;Inez&lt;/strong&gt;, keeps you company. Fix a bug, run its tests, and earn money, experience, and skill practice.&lt;/p&gt;

&lt;p&gt;But completing contracts uses energy. Going outside restores it and earns fresh-air tokens, the currency for plants. A plain office gradually becomes a green workspace with vines, warm lighting, better equipment, and a second monitor.&lt;/p&gt;

&lt;p&gt;The loop is simple:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Code → take a walk → notice something in nature → bring life back to your office.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The phone view asks you to start a walk, pocket the phone, and briefly return when a nature-photo mission appears. A normal qualifying walk needs &lt;strong&gt;10 minutes and at least 150 metres from the starting point&lt;/strong&gt;. Walking itself earns rewards; the optional photo mission adds a reason to notice your surroundings.&lt;/p&gt;

&lt;p&gt;AFK.exe is for developers and learners who enjoy progression games and could use a more inviting reason to take a break. Five skills—Debugging, TypeScript, Testing, Performance, and Async—lead through ten career titles and continuing mastery. Career milestones unlock developer workflows for your own code, so the rewards extend beyond decorating a room.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/Y_kr_XDbQHI" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

&lt;p&gt;&lt;a href="https://bmartin2000.github.io/Gooutdoors/" rel="noopener noreferrer"&gt;Play the browser demo&lt;/a&gt; · &lt;a href="https://youtu.be/Y_kr_XDbQHI" rel="noopener noreferrer"&gt;Watch on YouTube&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The video shows a coding fix passing its tests, payment and skill rewards, an outdoor mission submission, office upgrades, and progression to Distinguished Engineer.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Demo transparency:&lt;/strong&gt; the career tour uses labelled admin acceleration. The outdoor sequence uses a sample image and admin approval; it is not a demonstration of a real GPS-qualified outing or a successful local-model photo check. The developer-tool screens show the available workflows, not AI results generated during this recording.&lt;/p&gt;

&lt;p&gt;The hosted browser build lets you explore the game and phone experience. &lt;strong&gt;Local AI requires the repository version running on your computer with Ollama.&lt;/strong&gt; GitHub Pages does not host the model.&lt;/p&gt;

&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/Bmartin2000" rel="noopener noreferrer"&gt;
        Bmartin2000
      &lt;/a&gt; / &lt;a href="https://github.com/Bmartin2000/Gooutdoors" rel="noopener noreferrer"&gt;
        Gooutdoors
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;AFK.exe&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Code. Explore. Upgrade.&lt;/strong&gt; · &lt;em&gt;Your workspace grows when you step away.&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;You sit at a desk in a small 3D office, and the whole game runs on the computer in front of you. Playable contracts refill immediately; extra offers arrive every two minutes by default. You take the ones you're qualified for, ship the fix and get paid. Every job uses up energy, and the only way to fill it back up is to &lt;strong&gt;go outside&lt;/strong&gt;. Your phone measures the walk with GPS, offline, and buzzes when a photo mission appears. When you get back, an &lt;strong&gt;open-weight AI running on your own computer&lt;/strong&gt; (Qwen3.5 4B through Ollama) checks the photo. Fresh air buys the plants that make your office come alive.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;Built for the &lt;strong&gt;Hacktoberfest 2026 Open-Source AI Challenge: Week 1, "Touch Grass."&lt;/strong&gt;
No account. No cloud. No API key. Photos, code and location never leave your devices.&lt;/p&gt;
&lt;/blockquote&gt;…&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/Bmartin2000/Gooutdoors" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The project is &lt;strong&gt;MIT-licensed&lt;/strong&gt;. The repository includes the game, phone PWA, local Ollama adapter, tests, setup instructions, and &lt;a href="https://github.com/bmartin2000/Gooutdoors/blob/main/ATTRIBUTIONS.md" rel="noopener noreferrer"&gt;third-party attributions&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;The interface uses &lt;strong&gt;React, TypeScript, and Vite&lt;/strong&gt;. The office is built with procedural &lt;strong&gt;Three.js geometry through React Three Fiber&lt;/strong&gt;. Game progress, walk records, and photo evidence are stored in &lt;strong&gt;IndexedDB&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;The open AI component is &lt;strong&gt;Qwen3.5 4B&lt;/strong&gt;, an open-weight multimodal model, served locally through &lt;strong&gt;Ollama&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;It has two roles:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Nature-photo checking.&lt;/strong&gt; Back at the desk, the model examines a downscaled photo against the selected mission's visual criteria. It returns structured JSON containing a match decision, estimated confidence, a reason, and a review flag. Zod validates the response. Uncertain results hold the bonus reward instead of silently granting it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Developer tools.&lt;/strong&gt; Level 2 unlocks an AI Debugger, level 4 a Code Optimizer, and level 6 a Test Generator. These are different workflows over the same local model. Generated suggestions are displayed as text and are never executed automatically.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;I kept other checks deterministic. Coding contracts run authored test fixtures. Walk qualification uses GPS calculations on the phone. Neither relies on a language model to decide whether a test passed or a walk met its thresholds.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Phone: GPS + mission photo + local storage
                     |
           LAN transfer or outing file
                     |
Desk: React game + IndexedDB
                     |
        local, loopback-only adapter
                     |
          Ollama → Qwen3.5 4B
                     |
       validated result → bonus reward
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The phone-to-desk transfer carries a walk summary and re-encoded photos, rather than the GPS coordinate history. Reward grants use a transaction so importing the same outing again does not repeatedly pay out.&lt;/p&gt;

&lt;p&gt;To run the full version:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/bmartin2000/Gooutdoors.git
&lt;span class="nb"&gt;cd &lt;/span&gt;Gooutdoors
npm &lt;span class="nb"&gt;install
&lt;/span&gt;ollama pull qwen3.5:4b
npm run dev
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Ollama must be running; use &lt;code&gt;ollama serve&lt;/code&gt; if needed. The README also covers the installed phone PWA, local HTTPS, LAN pairing, and file transfer.&lt;/p&gt;

&lt;p&gt;The practical trade-off is device capability. Model downloads and initial setup need internet access, inference uses local compute, and browsers can pause GPS with the screen off. Pocket mode keeps a minimal screen awake where supported. The app also makes clear that photo matching is a plausibility check, not proof of where or when a photo was taken.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;For AFK.exe, local open-weight AI makes the outdoor loop possible without turning a walk into a cloud-upload workflow.&lt;/p&gt;

&lt;p&gt;After setup, a player can track a walk and save a mission photo without mobile data. The photo can be checked later on the player's own computer. Their submitted code and mission images do not need to go to an external inference service.&lt;/p&gt;

&lt;p&gt;Open code also makes the reward rules inspectable. Someone can read the prompts, review the uncertainty threshold, add a nature mission, or adapt the local-model integration. There is no inference API key or per-request provider bill, although running the model still requires suitable hardware and electricity.&lt;/p&gt;

&lt;p&gt;A cloud API could perform image analysis, but it would add a network dependency and third-party data transfer to this particular design. Keeping inference local lets a player bring the experience home on their own terms.&lt;/p&gt;

&lt;p&gt;The idea I want to preserve is simple: &lt;strong&gt;the most valuable thing the app asks you to do happens outside the app.&lt;/strong&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  My Agent Session
&lt;/h2&gt;

&lt;p&gt;I used &lt;strong&gt;Codex&lt;/strong&gt; to help implement, debug, and refine the project. The iteration included faster contract replenishment, clearer skill-training requirements and rewards, mobile navigation back from a walk, an office-first opening, and a camera transition that completes consistently on slower devices.&lt;/p&gt;

&lt;p&gt;The implementation is visible in the &lt;a href="https://github.com/bmartin2000/Gooutdoors/commits/main/" rel="noopener noreferrer"&gt;commit history&lt;/a&gt;. I have not attached a saved agent-session transcript to this draft.&lt;/p&gt;

&lt;p&gt;The demo was produced with Higgsfield, English female narration through Seed Audio, and an original electronic instrumental. This post was drafted with AI assistance and is being reviewed before publication.&lt;/p&gt;

&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Best Use of GitHub Copilot — GitHub Actions automation.&lt;/strong&gt; The &lt;a href="https://github.com/bmartin2000/Gooutdoors/blob/main/.github/workflows/pages.yml" rel="noopener noreferrer"&gt;deployment workflow&lt;/a&gt; installs dependencies, runs the tests, builds the static app, and deploys it to GitHub Pages when &lt;code&gt;main&lt;/code&gt; changes. The challenge lists GitHub Actions project automation as a qualifying use for this category; this entry does not claim that Copilot authored the game.&lt;/p&gt;

&lt;h2&gt;
  
  
  Related Community Reading
&lt;/h2&gt;

&lt;p&gt;While preparing this write-up, I read &lt;a href="https://dev.to/sanskarcodes/building-local-first-ai-apps-what-changes-when-the-data-stays-on-the-device-15mi"&gt;Sanskar's &lt;em&gt;Building Local-First AI Apps: What Changes When the Data Stays on the Device&lt;/em&gt;&lt;/a&gt; and &lt;a href="https://dev.to/arun2232/lookout-the-offline-ai-field-companion-that-tells-you-to-put-your-phone-away-26na"&gt;Arun Mahajan's &lt;em&gt;Lookout: The Offline AI Field Companion That Tells You to Put Your Phone Away&lt;/em&gt;&lt;/a&gt;. They offer related perspectives on keeping useful work local and letting an outdoor app have a stopping point. They are contextual reading, rather than code dependencies.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Code. Explore. Upgrade.&lt;/strong&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>opensource</category>
    </item>
  </channel>
</rss>
