<?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: Sahil Jaiswal</title>
    <description>The latest articles on DEV Community by Sahil Jaiswal (@sahil_jaiswal_6affec34584).</description>
    <link>https://dev.to/sahil_jaiswal_6affec34584</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%2F4158334%2F276a766f-7f96-4990-b841-4dccc8e5b85f.jpeg</url>
      <title>DEV Community: Sahil Jaiswal</title>
      <link>https://dev.to/sahil_jaiswal_6affec34584</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/sahil_jaiswal_6affec34584"/>
    <language>en</language>
    <item>
      <title>Bored 🥱 of scrolling your phone 📱? Try Outbound, and freshen your day ☘️</title>
      <dc:creator>Sahil Jaiswal</dc:creator>
      <pubDate>Fri, 09 Oct 2026 19:40:43 +0000</pubDate>
      <link>https://dev.to/sahil_jaiswal_6affec34584/bored-of-scrolling-your-phone-try-outbound-and-freshen-your-day-1jcm</link>
      <guid>https://dev.to/sahil_jaiswal_6affec34584/bored-of-scrolling-your-phone-try-outbound-and-freshen-your-day-1jcm</guid>
      <description>&lt;div class="ltag__link--embedded"&gt;
  &lt;div class="crayons-story "&gt;
  &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b" class="crayons-story__hidden-navigation-link"&gt;Outbound 🌱 An AI Quest App That Wants You to Close It&lt;/a&gt;
    &lt;div class="crayons-article__cover crayons-article__cover__image__feed"&gt;
      &lt;iframe src="https://www.youtube.com/embed/wlCcCtbvGxA" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" title="Outbound 🌱 An AI Quest App That Wants You to Close It"&gt;&lt;/iframe&gt;
    &lt;/div&gt;


  &lt;div class="crayons-story__body crayons-story__body-full_post"&gt;
      &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b" class="crayons-article__context-note crayons-article__context-note__feed"&gt;&lt;p&gt;Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Submission 🌿&lt;/p&gt;

&lt;/a&gt;
    &lt;div class="crayons-story__top"&gt;
      &lt;div class="crayons-story__meta"&gt;
        &lt;div class="crayons-story__author-pic"&gt;

          &lt;a href="/sahil_jaiswal_6affec34584" class="crayons-avatar  crayons-avatar--l  "&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%2Fuser%2Fprofile_image%2F4158334%2F276a766f-7f96-4990-b841-4dccc8e5b85f.jpeg" alt="sahil_jaiswal_6affec34584 profile" class="crayons-avatar__image"&gt;
          &lt;/a&gt;
        &lt;/div&gt;
        &lt;div&gt;
          &lt;div&gt;
            &lt;a href="/sahil_jaiswal_6affec34584" class="crayons-story__secondary fw-medium m:hidden"&gt;
              Sahil Jaiswal
            &lt;/a&gt;
            &lt;div class="profile-preview-card relative mb-4 s:mb-0 fw-medium hidden m:inline-block"&gt;
              
                Sahil Jaiswal
                
                
              
              &lt;div id="story-author-preview-content-4825056" class="profile-preview-card__content crayons-dropdown branded-7 p-4 pt-0"&gt;
                &lt;div class="gap-4 grid"&gt;
                  &lt;div class="-mt-4"&gt;
                    &lt;a href="/sahil_jaiswal_6affec34584" class="flex"&gt;
                      &lt;span class="crayons-avatar crayons-avatar--xl mr-2 shrink-0"&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%2Fuser%2Fprofile_image%2F4158334%2F276a766f-7f96-4990-b841-4dccc8e5b85f.jpeg" class="crayons-avatar__image" alt=""&gt;
                      &lt;/span&gt;
                      &lt;span class="crayons-link crayons-subtitle-2 mt-5"&gt;Sahil Jaiswal&lt;/span&gt;
                    &lt;/a&gt;
                  &lt;/div&gt;
                  &lt;div class="print-hidden"&gt;
                    
                      Follow
                    
                  &lt;/div&gt;
                  &lt;div class="author-preview-metadata-container"&gt;&lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
            &lt;/div&gt;

          &lt;/div&gt;
          &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b" class="crayons-story__tertiary fs-xs"&gt;&lt;time&gt;Oct 9&lt;/time&gt;&lt;span class="time-ago-indicator-initial-placeholder"&gt;&lt;/span&gt;&lt;/a&gt;
        &lt;/div&gt;
      &lt;/div&gt;

    &lt;/div&gt;

    &lt;div class="crayons-story__indention"&gt;
      &lt;h2 class="crayons-story__title crayons-story__title-full_post"&gt;
        &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b" id="article-link-4825056"&gt;
          Outbound 🌱 An AI Quest App That Wants You to Close It
        &lt;/a&gt;
      &lt;/h2&gt;
        &lt;div class="crayons-story__tags"&gt;
            &lt;a class="crayons-tag  crayons-tag--monochrome " href="/t/devchallenge"&gt;&lt;span class="crayons-tag__prefix"&gt;#&lt;/span&gt;devchallenge&lt;/a&gt;
            &lt;a class="crayons-tag  crayons-tag--monochrome " href="/t/hf26challenge"&gt;&lt;span class="crayons-tag__prefix"&gt;#&lt;/span&gt;hf26challenge&lt;/a&gt;
            &lt;a class="crayons-tag  crayons-tag--monochrome " href="/t/ai"&gt;&lt;span class="crayons-tag__prefix"&gt;#&lt;/span&gt;ai&lt;/a&gt;
            &lt;a class="crayons-tag  crayons-tag--monochrome " href="/t/hacktoberfest"&gt;&lt;span class="crayons-tag__prefix"&gt;#&lt;/span&gt;hacktoberfest&lt;/a&gt;
        &lt;/div&gt;
      &lt;div class="crayons-story__bottom"&gt;
        &lt;div class="crayons-story__details"&gt;
          &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b" class="crayons-btn crayons-btn--s crayons-btn--ghost crayons-btn--icon-left"&gt;
            &lt;div class="multiple_reactions_aggregate"&gt;
              &lt;span class="multiple_reactions_icons_container"&gt;
                  &lt;span class="crayons_icon_container"&gt;
                    &lt;img src="https://assets.dev.to/assets/raised-hands-74b2099fd66a39f2d7eed9305ee0f4553df0eb7b4f11b01b6b1b499973048fe5.svg" width="18" height="18"&gt;
                  &lt;/span&gt;
                  &lt;span class="crayons_icon_container"&gt;
                    &lt;img src="https://assets.dev.to/assets/sparkle-heart-5f9bee3767e18deb1bb725290cb151c25234768a0e9a2bd39370c382d02920cf.svg" width="18" height="18"&gt;
                  &lt;/span&gt;
              &lt;/span&gt;
              &lt;span class="aggregate_reactions_counter"&gt;2&lt;span class="hidden s:inline"&gt;&amp;nbsp;reactions&lt;/span&gt;&lt;/span&gt;
            &lt;/div&gt;
          &lt;/a&gt;
            &lt;a href="https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b#comments" class="crayons-btn crayons-btn--s crayons-btn--ghost crayons-btn--icon-left flex items-center"&gt;
              

              4&lt;span class="hidden s:inline"&gt;&amp;nbsp;comments&lt;/span&gt;
            &lt;/a&gt;
        &lt;/div&gt;
        &lt;div class="crayons-story__save"&gt;
          &lt;small class="crayons-story__tertiary fs-xs mr-2"&gt;
            7 min read
          &lt;/small&gt;
        &lt;/div&gt;
      &lt;/div&gt;
    &lt;/div&gt;
  &lt;/div&gt;
&lt;/div&gt;

&lt;/div&gt;


</description>
      <category>ai</category>
      <category>mobile</category>
      <category>showdev</category>
    </item>
    <item>
      <title>Outbound 🌱 An AI Quest App That Wants You to Close It</title>
      <dc:creator>Sahil Jaiswal</dc:creator>
      <pubDate>Fri, 09 Oct 2026 19:01:04 +0000</pubDate>
      <link>https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b</link>
      <guid>https://dev.to/sahil_jaiswal_6affec34584/outbound-an-ai-quest-app-that-wants-you-to-close-it-2e9b</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;🌱 &lt;strong&gt;What if an app's best outcome was that you stopped using it?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;You have fifteen minutes. You have been looking at a screen for too long. A workout sounds like too much, and "just go for a walk" does not quite give you a reason to start.&lt;/p&gt;

&lt;p&gt;That gap between wanting to do something and choosing something is the problem I wanted to work on. A feed offers the next thing to look at. I wanted to offer the next thing to do.&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;Outbound&lt;/strong&gt;, a gamified outdoor quest app using open-weight models that can run locally. It suggests three activities using your settings, interests, past feedback, and available local information. Choose one, put the phone away, and return with a short reflection.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The idea is not to make leaving the screen another demanding task. It is to make the next step small enough to take.&lt;/strong&gt;&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%2F9f4g4ex0buiqrwte83ah.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%2F9f4g4ex0buiqrwte83ah.png" alt="Outbound's homepage with outing settings and three quests" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Figure 1. Outbound's homepage: set your available time and how you feel, then choose one of three quests.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The experience is deliberately small:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Your profile:&lt;/strong&gt; editable interests, goals, dislikes, and comfort levels.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Today's context:&lt;/strong&gt; time, mood, energy, and an optional starting point.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Three choices, not a feed:&lt;/strong&gt; Easy Win, Useful Quest, and Stretch Quest, selected from &lt;strong&gt;120 activities&lt;/strong&gt;. Notice something new on a familiar walk, sketch outdoors, or check a detail for tomorrow's errand.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Useful feedback:&lt;/strong&gt; mark complete, partial, or skipped, optionally rate it, and leave a note. "Bring water next time" can become a reminder for a similar activity.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gentle rewards:&lt;/strong&gt; badges, streaks, and sounds acknowledge participation. &lt;strong&gt;Partial Counts&lt;/strong&gt; is a badge too; showing up need not go perfectly.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Optional voice:&lt;/strong&gt; ElevenLabs reads quests, transcribes notes, and speaks encouragement after your reflection.&lt;/li&gt;
&lt;/ul&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%2Fmuf5fao952s1z6ww0tuw.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%2Fmuf5fao952s1z6ww0tuw.png" alt="Mobile screens for choosing a quest, timing an outing, and saving feedback" width="799" height="644"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Figure 2. Choose a quest, head outside, and save a short reflection when you return.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;This is for someone between work, study, or chores who wants a little movement, curiosity, or confidence, but needs a reason to start.&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;📱 The design goal: spend less effort deciding on the screen, and more time doing something outside.&lt;/p&gt;
&lt;/blockquote&gt;

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

&lt;p&gt;🚀 &lt;strong&gt;&lt;a href="https://outbound-taqn.onrender.com/" rel="noopener noreferrer"&gt;Try Outbound&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Set a few interests and choose today's time and energy. Share a starting point only if you want nearby-place suggestions.&lt;/li&gt;
&lt;li&gt;Generate three quests, choose one, and start the quiet outing timer. Put the phone away; there is no feed to follow.&lt;/li&gt;
&lt;li&gt;Return and save a short reflection. Completed, partial, and skipped are all valid feedback.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;🔍 &lt;strong&gt;To inspect a decision:&lt;/strong&gt; Recommendation lab → Seed demo data → Run demo recommendation.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;See sources, rejected activities, predictions, and selection scores.&lt;/li&gt;
&lt;li&gt;Synthetic demo history stays separate from your personal progress.&lt;/li&gt;
&lt;li&gt;A new profile starts with preference-based ranking, not learned personalization.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  A Small Outing Still Has to Fit
&lt;/h3&gt;

&lt;p&gt;For an &lt;strong&gt;illustrative&lt;/strong&gt; fifteen-minute outing when you feel tired:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Too far:&lt;/strong&gt; a garden with a twenty-eight-minute walking round trip is ruled out before prediction.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Too demanding:&lt;/strong&gt; low-energy settings exclude strenuous activities.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Still possible:&lt;/strong&gt; a short, familiar walk might remain.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A confident description cannot make the garden closer.&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%2Fb065oyzboq9j28vukam2.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%2Fb065oyzboq9j28vukam2.png" alt="An example quest with a destination and time breakdown" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Figure 3. An example quest with time for walking, the activity, and a return buffer.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;That distinction matters: &lt;strong&gt;Outbound should give you a usable reason to leave, not a convincing reason that falls apart at the door.&lt;/strong&gt;&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;💻 &lt;strong&gt;&lt;a href="https://github.com/Sahil-Jaiswal-189/Outbound" rel="noopener noreferrer"&gt;Source code&lt;/a&gt;&lt;/strong&gt;
&lt;/li&gt;
&lt;li&gt;🛠️ &lt;strong&gt;&lt;a href="https://github.com/Sahil-Jaiswal-189/Outbound#prerequisites" rel="noopener noreferrer"&gt;Setup and prerequisites&lt;/a&gt;&lt;/strong&gt;
&lt;/li&gt;
&lt;li&gt;🧩 &lt;strong&gt;&lt;a href="https://github.com/Sahil-Jaiswal-189/Outbound/blob/main/ARCHITECTURE.md" rel="noopener noreferrer"&gt;Detailed architecture&lt;/a&gt;&lt;/strong&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Stack:&lt;/strong&gt; JavaScript, Lucide, and Leaflet in the browser; Node and SQLite in the backend; Python/TabPFN and Ollama/Qwen for inference. Run locally or together in one Docker service.&lt;/p&gt;

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

&lt;p&gt;🧩 &lt;strong&gt;The most useful improvement was giving each part one clear job.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;An earlier version added mood and place bonuses on top of predictions. That made the decision harder to explain. I simplified it: &lt;strong&gt;check what fits, predict outcomes, choose three, then add friendly wording.&lt;/strong&gt;&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%2F5fg93vzwms77mmsaumpx.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%2F5fg93vzwms77mmsaumpx.png" alt="A diagram of how Outbound chooses quests and learns from feedback" width="800" height="267"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Figure 4. How Outbound checks what fits, predicts outcomes, and chooses three quests. AI-generated diagram.&lt;/em&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Facts Decide What Fits
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Open-Meteo:&lt;/strong&gt; weather and modeled air quality.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;OpenStreetMap / Overpass:&lt;/strong&gt; mapped nearby places.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;openrouteservice:&lt;/strong&gt; walking-time estimates.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Our code combines these facts with your settings to remove unsuitable activities &lt;strong&gt;before prediction&lt;/strong&gt;. A named destination must fit travel, activity, and return time. Failed lookups are reported; suitable general activities can remain, without inventing a place.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. TabPFN Asks Whether It Might Work for You
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Remember:&lt;/strong&gt; SQLite stores your profile, outings, and feedback.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Predict:&lt;/strong&gt; TabPFN v2 uses past outcomes and selected details, including mood, weather, duration, and effort, to estimate completion and liking separately.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Check:&lt;/strong&gt; each outcome must beat a simpler estimate on history not used for fitting. Until enough usable history exists, goals and hobbies guide ranking.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Update:&lt;/strong&gt; feedback adds examples, not a new round of model-weight training.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Finishing something and enjoying it are different signals. The model has to earn its place.&lt;/strong&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Selection Makes a Choice, Not an Endless Ranking
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Model-based score = average of completion and liking estimates − repetition penalty.&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Partial model use:&lt;/strong&gt; an outcome that fails the check keeps its simpler estimate.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No validated model:&lt;/strong&gt; use goal/hobby fit minus repetition instead.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Variety:&lt;/strong&gt; avoid duplicate activities and destinations, seek different categories, and occasionally explore a less-tried category.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;These are selection rules, not another AI model.&lt;/strong&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  4. Qwen Makes the Wording Friendly
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Wording:&lt;/strong&gt; Qwen2.5:3b runs through Ollama after selection, writing titles, reflection prompts, and later encouragement.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Personal context:&lt;/strong&gt; profile details and relevant notes, not per-user fine-tuning.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Facts:&lt;/strong&gt; destinations, timings, instructions, and evidence-based explanations stay backend-controlled.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Fallback:&lt;/strong&gt; failed or invalid model output leaves the template quests usable.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Qwen's job is warmth, not deciding where you should go.&lt;/strong&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  What I Have Verified
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;🔬 &lt;strong&gt;75 automated tests passed&lt;/strong&gt; in build verification, covering filtering, separate demo history, repetition, and model-failure fallbacks.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Real local TabPFN inference:&lt;/strong&gt; 120 synthetic outings; both outcomes beat simpler estimates on examples held out of fitting.&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Prediction&lt;/th&gt;
&lt;th&gt;Baseline error&lt;/th&gt;
&lt;th&gt;TabPFN error&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Full completion&lt;/td&gt;
&lt;td&gt;0.0975&lt;/td&gt;
&lt;td&gt;0.0588&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Liking&lt;/td&gt;
&lt;td&gt;0.1203&lt;/td&gt;
&lt;td&gt;0.0689&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Reading the table:&lt;/strong&gt; Brier scores measure prediction error; lower is better. This small synthetic test demonstrates the working system, &lt;strong&gt;not real-user performance or better habits&lt;/strong&gt;.&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%2Flwed0owrrk6q517hmoi3.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%2Flwed0owrrk6q517hmoi3.png" alt="The recommendation lab showing TabPFN predictions and quest scores" width="800" height="667"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Figure 5. TabPFN's predictions in the recommendation lab, using simulated outings.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/Sahil-Jaiswal-189/Outbound/blob/main/docs/assets/lab-model.evidence.json" rel="noopener noreferrer"&gt;View the results&lt;/a&gt; | &lt;a href="https://github.com/Sahil-Jaiswal-189/Outbound/blob/main/scripts/capture-model-lab.mjs" rel="noopener noreferrer"&gt;Reproduce this screenshot&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  What I Learned Outside
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;I used Outbound during development, trying a few quests that took me to a local shop and a nearby café.&lt;/strong&gt; The early recommendations felt too generic, so I improved the real-world context and expanded the activity catalog. I also added a timer, feedback, badges, and optional voice to make the experience more engaging. These were informal personal trials, not a study of long-term habit change.&lt;/p&gt;

&lt;h3&gt;
  
  
  What Still Needs Work
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Speed:&lt;/strong&gt; cold models and public map services can still delay recommendations despite caching. Shortening that wait is the next priority.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Place details:&lt;/strong&gt; access and opening hours remain unverified.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Completion:&lt;/strong&gt; self-reported, not location-tracked.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Those few outings were a starting point. The next step is to test with more people: can they choose comfortably, leave the app alone, and return with useful feedback?&lt;/p&gt;

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

&lt;p&gt;🔓 &lt;strong&gt;A personal recommendation needs personal feedback. I wanted to control where that feedback goes and how it shapes the next choice.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Qwen's open weights, TabPFN's model and code, and the open-source Ollama runtime give me three practical benefits:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Control over personal history.&lt;/strong&gt; Laptop mode keeps storage and inference on the laptop. The Render demo runs them on its server, not the visitor's phone.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;An inspectable decision.&lt;/strong&gt; I can compare TabPFN with a simpler estimate, keep the fallback when it wins, and see which path was used.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Local computation.&lt;/strong&gt; Downloaded models can compute offline. Fresh weather, places, routes, and tiles need internet; coordinates go to providers. Optional speech text and recordings go to ElevenLabs.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For this project, that worked better than relying entirely on a hosted chatbot: I can run and test both models without a required hosted-inference fee, and control the selection rules. Closed models can also sit behind careful rules. &lt;strong&gt;Our advantage is ownership of the setup, not universal model superiority.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The boundaries are important too:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Costs:&lt;/strong&gt; hosting, hardware, and optional voice are not free.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Model terms:&lt;/strong&gt; &lt;a href="https://huggingface.co/Qwen/Qwen2.5-3B-Instruct" rel="noopener noreferrer"&gt;Qwen uses the Qwen Research license&lt;/a&gt;; &lt;a href="https://huggingface.co/Prior-Labs/TabPFN-v2-clf" rel="noopener noreferrer"&gt;TabPFN v2 uses the Prior Labs License with attribution requirements&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Optional proprietary service:&lt;/strong&gt; ElevenLabs is separate from the local inference setup.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The freedom I care about is not making the app bigger. &lt;strong&gt;It is being able to make the app ask less of you.&lt;/strong&gt;&lt;/p&gt;

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

&lt;h3&gt;
  
  
  📊 Best Use of TabPFN
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Use:&lt;/strong&gt; predict completion and liking from outing feedback; check each outcome against a simpler alternative before model-based ranking.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Purpose:&lt;/strong&gt; recommend what someone might do and enjoy, not what keeps them clicking.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Evidence:&lt;/strong&gt; the lab, real-model screenshot, and saved results make the process inspectable.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🎙️ Best Use of ElevenLabs
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Use:&lt;/strong&gt; narrate the selected quest, transcribe a recorded reflection, and speak Qwen's encouragement after the note is saved.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Purpose:&lt;/strong&gt; less reading and typing, without a conversation running throughout the outing.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Alternatives:&lt;/strong&gt; typed notes and browser speech remain available.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🚀 Best Use of Render
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Use:&lt;/strong&gt; host the public demo with a single-service Docker setup for the app, Qwen, and TabPFN.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Purpose:&lt;/strong&gt; let people try Outbound without installing models. An attached persistent disk preserves history and model downloads.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Deployment:&lt;/strong&gt; only the web app is exposed publicly; model services stay internal. Configuration and setup instructions are in the repository.&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;🌿 &lt;strong&gt;Outbound does not need you to become an outdoors person overnight.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It needs to make one small outing easier to choose today. Whether it succeeds will be decided outside, not by how long someone stays in the app.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;This project and write-up were developed with AI assistance. Screenshots use controlled demo state; the prediction capture uses synthetic history. The app's park photograph is by &lt;a href="https://unsplash.com/photos/a-path-winds-through-a-sunny-green-park-lh_MesNhkbI" rel="noopener noreferrer"&gt;Tina Devidze on Unsplash&lt;/a&gt;, under the &lt;a href="https://unsplash.com/license" rel="noopener noreferrer"&gt;Unsplash License&lt;/a&gt;. Model and source-provider credits appear above and in the repository.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>ai</category>
      <category>hacktoberfest</category>
    </item>
    <item>
      <title>GardenBuddy: A Weather-Aware Plant Assistant That Remembers</title>
      <dc:creator>Sahil Jaiswal</dc:creator>
      <pubDate>Sun, 04 Oct 2026 17:18:08 +0000</pubDate>
      <link>https://dev.to/sahil_jaiswal_6affec34584/gardenbuddy-a-weather-aware-plant-assistant-that-remembers-2m0g</link>
      <guid>https://dev.to/sahil_jaiswal_6affec34584/gardenbuddy-a-weather-aware-plant-assistant-that-remembers-2m0g</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-weekend-2026-10-01"&gt;Hacktoberfest Weekend Challenge: Build for a Friend&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;GardenBuddy&lt;/strong&gt;, a small plant care dashboard for a friend who loves plants but does not want plant care to become another thing to constantly remember.&lt;/p&gt;

&lt;p&gt;The idea is simple: when you add a plant, you also select where it lives. That can be a city and region, plus whether the plant is indoors, outdoors, or on a balcony. GardenBuddy then checks the local weather for that area and turns it into practical care advice.&lt;/p&gt;

&lt;p&gt;So instead of only saying “water every 3 days,” it can notice the situation around the plant. If rain is coming, it may tell you to skip watering. If the day is too hot, it may suggest filtered light or shade. If the night is too cold, it may suggest moving the plant inside. The goal is to make the daily decision easier: what should I do for this plant today?&lt;/p&gt;

&lt;p&gt;Each plant gets its own card with the plant image, weather strip, care status, main action, reason, placement advice, and growth advice. The basic care rules are always available, so the app still works even without AI.&lt;/p&gt;

&lt;p&gt;On top of that, GardenBuddy has a local AI assistant powered by Qwen through Ollama. There are two AI experiences. The first is an AI care plan, where Qwen looks at the plant profile, weather, rule signal, watering history, calendar tasks, and saved memory to create a more detailed plan. The second is plant chat, where you can ask questions about one specific plant.&lt;/p&gt;

&lt;p&gt;This is where GardenBuddy is different from a normal chatbot. A normal AI bot can answer “why are tomato leaves yellow?” but it does not know your tomato plant. GardenBuddy can use the plant’s actual history: when it was watered, what the weather looked like, what advice was given earlier, and whether a similar issue happened before.&lt;/p&gt;

&lt;p&gt;For example, if I ask about yellow tomato leaves today, GardenBuddy can save that incident with the weather, care signal, and answer. Later, if I ask again about the same tomato plant, it can bring that memory back before Qwen answers. That makes the assistant feel more like a plant journal with memory, not just a generic Q&amp;amp;A box.&lt;/p&gt;

&lt;p&gt;There is also a care calendar. It shows upcoming watering checks, weather warnings, and tasks grouped by plant. I can add my own task, remove something that does not apply, or view all plants together. This makes the app useful beyond one-time advice because it helps plan what needs attention over the next few days.&lt;/p&gt;

&lt;p&gt;GardenBuddy is meant to be calm and practical. It does not try to replace gardening knowledge. It helps someone who likes plants make better daily decisions with less guessing.&lt;/p&gt;

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

&lt;p&gt;&lt;iframe src="https://player.vimeo.com/video/1232801749" width="710" height="399"&gt;
&lt;/iframe&gt;
&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/Sahil-Jaiswal-189" rel="noopener noreferrer"&gt;
        Sahil-Jaiswal-189
      &lt;/a&gt; / &lt;a href="https://github.com/Sahil-Jaiswal-189/GardenBuddy" rel="noopener noreferrer"&gt;
        GardenBuddy
      &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;GardenBuddy&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;GardenBuddy is a weather-aware plant care dashboard for people who want simple, practical help with their plants.&lt;/p&gt;
&lt;p&gt;It tracks each plant separately, checks the local forecast for that plant's city, creates a care calendar, and can use a local open-weight model through Ollama for plant-specific planning and chat. The app is designed to stay useful even when the AI model is not running.&lt;/p&gt;
&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;What It Does&lt;/h2&gt;
&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;Add plants with name, type, city, region, stage, placement, sunlight, watering rhythm, and notes.&lt;/li&gt;
&lt;li&gt;Fetch a 7-day forecast for each plant location using Open-Meteo.&lt;/li&gt;
&lt;li&gt;Show clear care signals such as &lt;code&gt;Needs water&lt;/code&gt;, &lt;code&gt;Skip watering&lt;/code&gt;, &lt;code&gt;Heat watch&lt;/code&gt;, &lt;code&gt;Protect from rain&lt;/code&gt;, and &lt;code&gt;Too cold outside&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Build a horizontal care calendar with watering checks, weather warnings, and user-added tasks.&lt;/li&gt;
&lt;li&gt;Let you log care actions such as watering.&lt;/li&gt;
&lt;li&gt;Keep plant-specific memory, so later AI answers can refer to past issues for that…&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/Sahil-Jaiswal-189/GardenBuddy" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


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

&lt;p&gt;GardenBuddy is a lightweight full-stack app with a plain HTML, CSS, and JavaScript frontend and a small Node.js backend.&lt;/p&gt;

&lt;p&gt;When a user adds a plant, the app saves the plant profile and location. The backend uses Open-Meteo to fetch weather for that city and region. The frontend then combines that weather with the plant’s watering rhythm, placement, sunlight preference, and last watered date to show the basic rule-based care signal.&lt;/p&gt;

&lt;p&gt;I wanted the rules to exist separately from AI because they make the app dependable. Even if the model is offline, GardenBuddy can still show useful care actions like watering, skipping watering because of rain, protecting from heat, or moving a plant away from cold weather.&lt;/p&gt;

&lt;p&gt;The AI layer runs locally through Ollama using &lt;code&gt;qwen2.5:3b&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;Qwen is used after the basic signal is calculated. It does not blindly decide everything from scratch. It receives structured context from the app: the selected plant, today’s weather, the care signal, care logs, calendar tasks, previous AI plans, and relevant plant memories. That context is what makes the answer feel connected to the actual plant.&lt;/p&gt;

&lt;p&gt;MongoDB Atlas is used as the persistent data layer. It stores the garden state, plant profiles, calendar tasks, care logs, generated AI plans, and plant memories.&lt;/p&gt;

&lt;p&gt;The most important collection is &lt;code&gt;plantMemories&lt;/code&gt;. When the user chats about a plant, GardenBuddy can save the question, answer, current weather, care signal, and tags. The next time the user asks something about that same plant, the backend searches those memories and passes the relevant ones to Qwen.&lt;/p&gt;

&lt;p&gt;The calendar is also part of the main workflow. GardenBuddy automatically creates useful upcoming tasks from watering rhythm and weather forecast, and the user can add custom tasks too. The calendar is grouped by plant so advice does not become one messy list.&lt;/p&gt;

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

&lt;p&gt;Open innovation matters here because GardenBuddy works with personal routines and home context.&lt;/p&gt;

&lt;p&gt;A plant care app can know someone’s city, balcony setup, indoor placement, watering habits, travel rhythm, and repeated plant problems. Using a local open-weight model through Ollama means the AI part can run on the user’s own machine instead of depending on a paid closed API for every answer.&lt;/p&gt;

&lt;p&gt;It also makes the system more flexible. The model can be swapped, prompts can be changed, and the app can still work when the model is offline because the rule layer remains available.&lt;/p&gt;

&lt;p&gt;For me, the useful part is the combination: rules for reliable daily decisions, weather for real-world context, MongoDB for memory, and Qwen for personalized guidance. The AI is not just pasted on top. It is grounded in the plant’s saved history and the app’s actual care logic.&lt;/p&gt;

&lt;p&gt;That is what makes GardenBuddy different from asking a normal chatbot. It does not only know general plant advice. It can remember what happened to your plant.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Best Use of MongoDB Atlas&lt;/strong&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;GardenBuddy uses MongoDB Atlas as the data layer for plant profiles, garden state, calendar tasks, care logs, generated AI plans, and plant-specific memories.&lt;/p&gt;

&lt;p&gt;MongoDB is especially important for the memory feature. Before Qwen answers a plant question, GardenBuddy retrieves relevant memories for that same plant and includes them in the context. This lets the assistant respond with awareness of past symptoms, watering decisions, weather stress, and previous care advice.&lt;/p&gt;

&lt;p&gt;MongoDB Atlas is not just where the data sits. It is what lets GardenBuddy become more useful the longer someone uses it.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>weekendchallenge</category>
      <category>hf26challenge</category>
    </item>
  </channel>
</rss>
