<?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: Tanzil Ahmed</title>
    <description>The latest articles on DEV Community by Tanzil Ahmed (@tanzil_ahmed_e6a826eee8c4).</description>
    <link>https://dev.to/tanzil_ahmed_e6a826eee8c4</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%2F4184875%2F39e2737e-b474-409d-863d-801780da12fc.png</url>
      <title>DEV Community: Tanzil Ahmed</title>
      <link>https://dev.to/tanzil_ahmed_e6a826eee8c4</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/tanzil_ahmed_e6a826eee8c4"/>
    <language>en</language>
    <item>
      <title>Choreseed: Make the chore leave behind a routine</title>
      <dc:creator>Tanzil Ahmed</dc:creator>
      <pubDate>Thu, 22 Oct 2026 06:58:13 +0000</pubDate>
      <link>https://dev.to/tanzil_ahmed_e6a826eee8c4/choreseed-make-the-chore-leave-behind-a-routine-343c</link>
      <guid>https://dev.to/tanzil_ahmed_e6a826eee8c4/choreseed-make-the-chore-leave-behind-a-routine-343c</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19"&gt;Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;Before I start something, I tend to collect advice: websites, videos, AI conversations, one more explanation. Then I spend my energy deciding which pieces apply and where to begin.&lt;/p&gt;

&lt;p&gt;For this challenge, I picked one small chore: cleaning a contact CSV. I wanted the effort of figuring it out to leave behind something I could use again.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Choreseed turns ordinary instructions and selected guidance into a routine you can inspect, rehearse, save, and replay.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It is for individuals, freelancers, and small teams who need to tidy contact or customer lists without writing code. The first release handles selected-field trimming, explicit case changes, column renaming, completely empty rows, and deduplication with an explicit key and keep rule.&lt;/p&gt;

&lt;p&gt;The useful part happens on the next file: load the saved routine, choose a compatible CSV, and run it. No new prompt or inference.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;&lt;a href="https://choreseed.onrender.com/demo.webm" rel="noopener noreferrer"&gt;Watch the 85-second real-integration recording&lt;/a&gt;&lt;/strong&gt; · &lt;a href="https://github.com/tanzilamd/choreseed/blob/main/docs/assets/choreseed-demo.webm" rel="noopener noreferrer"&gt;GitHub video mirror&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://choreseed.onrender.com" rel="noopener noreferrer"&gt;Try the free preview&lt;/a&gt;&lt;/strong&gt; — choose &lt;strong&gt;Try a fictional sample&lt;/strong&gt;, edit its steps, download the output, and save the routine. A video and text transcript are on the home page.&lt;/p&gt;

&lt;p&gt;The recording uses the development app with &lt;strong&gt;real Backboard inference and real Browserbase reference reading&lt;/strong&gt;. Its contacts are fictional, and playback condenses a roughly two-minute scripted run. My research fatigue is real; the demo is not a claim about personal CSV usage, customers, or human time saved.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Current hosting limit:&lt;/strong&gt; live AI creation and public-source reading are paused in the free preview until usage accounting has durable storage. The sample, imported routines, and local replay work. The recording demonstrates the full creation workflow; it is separate from the deployed preview. Free hosting can sleep between visits.&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%2F5ex3gajaoh2be7x34glh.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%2F5ex3gajaoh2be7x34glh.png" alt="Editable steps, before-and-after contacts, and an audit of actual changes. All contacts are fictional." width="800" height="1259"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The recording starts with this instruction:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Trim Name and Email. Lowercase Email. Remove completely empty rows and duplicate contacts.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;“Duplicate contacts” leaves a decision unresolved. The model asks which key and which occurrence to keep. I answer: &lt;strong&gt;Email, keep first; keep blank Email keys separate.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;After review, six fictional input rows become four output rows. Leading-zero IDs stay strings. Two rows with missing emails survive separately. The original file stays untouched.&lt;/p&gt;

&lt;p&gt;Then a second file arrives with the same headers in a different order. The saved routine handles it with &lt;strong&gt;zero additional AI calls&lt;/strong&gt;. A changed header set is blocked, rather than silently remapped.&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/tanzilamd" rel="noopener noreferrer"&gt;
        tanzilamd
      &lt;/a&gt; / &lt;a href="https://github.com/tanzilamd/choreseed" rel="noopener noreferrer"&gt;
        choreseed
      &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;Choreseed&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;A little less by hand.&lt;/strong&gt; Turn contact CSV instructions and selected guidance into a small routine you can inspect, edit, export and run again without another AI call.&lt;/p&gt;
&lt;p&gt;Built for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19" rel="nofollow"&gt;Hacktoberfest Open-Source AI Challenge, Week 3: Slow and Steady&lt;/a&gt;. This is an early release, not a general spreadsheet or arbitrary-code agent. All demo contacts are fictional.&lt;/p&gt;
&lt;p&gt;&lt;a href="https://choreseed.onrender.com" rel="nofollow noopener noreferrer"&gt;Free public preview&lt;/a&gt; · &lt;a href="https://choreseed.onrender.com/demo.webm" rel="nofollow noopener noreferrer"&gt;Watch the 85-second real-integration recording&lt;/a&gt; (&lt;a href="https://github.com/tanzilamd/choreseed/docs/assets/choreseed-demo.webm" rel="noopener noreferrer"&gt;mirror&lt;/a&gt;) · &lt;a href="https://github.com/tanzilamd/choreseed/docs/DELIVERY.md" rel="noopener noreferrer"&gt;Delivery status&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The free preview is usable for the fictional sample, editing, CSV export, recipe import/save and compatible-file replay, including offline replay after caching.&lt;/strong&gt; Public browser checks verified exact downloads and zero repeat inference. Live AI remains paused because durable usage storage is unavailable: paid hosting requires payment information, and the free database provisioning API is blocked. Runtime keys are present. See the &lt;a href="https://github.com/tanzilamd/choreseed/docs/DELIVERY.md" rel="noopener noreferrer"&gt;delivery record&lt;/a&gt; for the distinction between this usable free preview and full live-AI production acceptance.&lt;/p&gt;
&lt;div class="markdown-heading"&gt;…&lt;/div&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/tanzilamd/choreseed" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;Original code is &lt;strong&gt;MIT licensed&lt;/strong&gt;. The repository includes &lt;a href="https://github.com/tanzilamd/choreseed/blob/main/docs/SETUP.md" rel="noopener noreferrer"&gt;setup&lt;/a&gt;, &lt;a href="https://github.com/tanzilamd/choreseed/blob/main/docs/PRIVACY.md" rel="noopener noreferrer"&gt;privacy and limitations&lt;/a&gt;, &lt;a href="https://github.com/tanzilamd/choreseed/blob/main/THIRD_PARTY_NOTICES.md" rel="noopener noreferrer"&gt;model/dependency attribution&lt;/a&gt;, and &lt;a href="https://github.com/tanzilamd/choreseed/tree/main/docs/evidence" rel="noopener noreferrer"&gt;actual evaluation evidence&lt;/a&gt;.&lt;/p&gt;

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

&lt;p&gt;The model interprets the task. Deterministic code changes the data.&lt;/p&gt;

&lt;p&gt;A React/TypeScript workbench sends instructions, column names, clarification, and selected reference excerpts to Fastify. Backboard calls the open-weight &lt;strong&gt;&lt;a href="https://huggingface.co/Qwen/Qwen3.5-35B-A3B" rel="noopener noreferrer"&gt;Qwen3.5-35B-A3B&lt;/a&gt;&lt;/strong&gt;, pinned through OpenRouter to &lt;code&gt;deepinfra/fp8&lt;/code&gt;. Memory, web search, provider fallback, and automatic model routing are off.&lt;/p&gt;

&lt;p&gt;The result is a typed recipe, a necessary clarification, or an actionable unsupported-operation response. Zod and semantic validation reject unknown columns, rename collisions, invented reference IDs, unsupported steps, and executable code. “Ready” means ready to preview; the user still reviews it.&lt;/p&gt;

&lt;p&gt;The raw CSV stays in a browser Web Worker. PapaParse preserves strings. The interpreter applies ordered steps, computes changed cells and excluded rows, and exports a new CSV. Formula-like cells receive spreadsheet-safe escaping by default, visible in the preview. Explicit saves store recipes in IndexedDB, not raw files.&lt;/p&gt;

&lt;p&gt;Browserbase reads supported public references in bounded, read-only sessions. Users review the excerpts before compilation: at most three sources and 4,000 characters per source, with visible truncation and a paste fallback. Source text is untrusted data. A citation ID existing does not establish that its source supports the proposed step.&lt;/p&gt;

&lt;h3&gt;
  
  
  Two failures that improved the result
&lt;/h3&gt;

&lt;p&gt;Qwen3-32B produced malformed output in an early probe and weak Bangla results. I documented the failure and tested an explicitly named replacement, rather than silently changing models.&lt;/p&gt;

&lt;p&gt;The replacement then produced correct Bangla operations but put quoted instructions where reference IDs belonged. Validation rejected it. I made the allowed IDs explicit and reran the affected real case successfully.&lt;/p&gt;

&lt;p&gt;A final public offline test also caught a build-marker bug: the service worker had an empty asset list. I fixed the injection, made a missing marker fail the build, redeployed, and repeated the public journey. Cached-app replay then passed. Both failures remain in the evidence.&lt;/p&gt;

&lt;p&gt;The backend implements conservative paid-request reservations and session/IP quotas in SQLite. Restart-persistence tests pass with a persistent local ledger. The free host has an ephemeral filesystem, so public live AI stays off; a free-database provisioning API block remains unresolved. Public live-AI quota persistence has &lt;strong&gt;not&lt;/strong&gt; been verified.&lt;/p&gt;

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

&lt;p&gt;The Apache-2.0 open-weight model does the core semantic work: turning an English or Bangla instruction into a constrained plan. Its identity, weights, and license are available for inspection. Evaluating and explicitly replacing the first model was a real part of building the application.&lt;/p&gt;

&lt;p&gt;The recipe schema and MIT interpreter are also inspectable and modifiable. A user can export the routine as JSON, keep it, and run it again without buying another model response. That makes the result portable beyond the original AI conversation.&lt;/p&gt;

&lt;p&gt;The tested offline capability is precise: &lt;strong&gt;cached app + saved recipe + compatible local CSV&lt;/strong&gt;. Creation uses hosted inference. Local inference, self-hosted models, and provider zero retention are not claimed.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Lazy Payoff
&lt;/h2&gt;

&lt;p&gt;The measured result is &lt;strong&gt;zero new inference calls on the second compatible file&lt;/strong&gt;. Both recorded exports matched independent expected rows. Public downloads also matched independently written CSV bytes; public offline replay passed without caching API responses.&lt;/p&gt;

&lt;p&gt;Validation includes &lt;strong&gt;22 deterministic/API/compiler/accounting tests and five browser journeys&lt;/strong&gt;, plus real English/Bangla model cases and real reference reading. Fictional samples, mocked failures, and actual provider results are identified separately.&lt;/p&gt;

&lt;p&gt;I have not measured a matched manual baseline or human review time. First setup still takes care: source reading, clarification, provider waits, and review all count. The hypothesis is that this care can leave behind a reusable routine, instead of being spent again on the next file.&lt;/p&gt;

&lt;p&gt;That is my version of “Slow and Steady”: do the deciding once, keep the small routine, and bring it the next file.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Backboard:&lt;/strong&gt; actual open-weight model comparison and semantic compilation through Backboard.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Browserbase:&lt;/strong&gt; actual public reference reading supplies reviewed context to the compiler; browser journeys also verify the application.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;AI-assisted development disclosure: I directed the product and used OpenAI Codex to implement, test, and document it. All recorded contacts and evaluation fixtures are synthetic. No private agent-session export or DevRelay upload was made.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>opensource</category>
      <category>ai</category>
    </item>
    <item>
      <title>ScopeProof: An evidence-first answer to “one more thing”</title>
      <dc:creator>Tanzil Ahmed</dc:creator>
      <pubDate>Thu, 15 Oct 2026 19:36:37 +0000</pubDate>
      <link>https://dev.to/tanzil_ahmed_e6a826eee8c4/scopeproof-an-evidence-first-answer-to-one-more-thing-3odi</link>
      <guid>https://dev.to/tanzil_ahmed_e6a826eee8c4/scopeproof-an-evidence-first-answer-to-one-more-thing-3odi</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;“Could we add customer login? It should be a quick change.”&lt;/p&gt;

&lt;p&gt;A contact form and a login system can both sound like ordinary website work. But if the agreed price covers one and excludes the other, someone has to explain that difference before the work starts.&lt;/p&gt;

&lt;p&gt;That is the problem I chose for &lt;strong&gt;ScopeProof&lt;/strong&gt;: help a freelancer check a new request against the agreement and prepare a clear response.&lt;/p&gt;

&lt;p&gt;The app reviews up to eight requests and suggests &lt;strong&gt;Included&lt;/strong&gt;, &lt;strong&gt;Extra work&lt;/strong&gt;, or &lt;strong&gt;Needs clarification&lt;/strong&gt;. Each suggestion comes with a short reason and a scope quote when one is available. You can change any decision, estimate the extra work, edit a client reply and export the review. Nothing is sent automatically.&lt;/p&gt;

&lt;p&gt;Freelance developers and small web studios might pay for that convenience. I have not tested the business assumption with paying users yet.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;&lt;a href="https://scopeproof.onrender.com" rel="noopener noreferrer"&gt;Try ScopeProof&lt;/a&gt;&lt;/strong&gt; · &lt;strong&gt;&lt;a href="https://github.com/tanzilamd/scopeproof" rel="noopener noreferrer"&gt;View the repository&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The free Render service may need a moment to wake up. Start with &lt;strong&gt;Load illustrative sample&lt;/strong&gt;: a fictional agreement for five pages, a contact form and two revision rounds, excluding accounts and online payments.&lt;/p&gt;

&lt;p&gt;The contact form belongs in the original work. Login and payments need a separate discussion. A third revision goes beyond the stated limit. “Make the website more engaging” needs a clearer definition before anyone can sensibly estimate it.&lt;/p&gt;

&lt;p&gt;Enter two extra hours at $50/hour, for example, and the app calculates a proposed $100 quote for the editable reply. The client still needs to approve it.&lt;/p&gt;

&lt;p&gt;The sample is labelled prewritten output and sends nothing. For real inference, acknowledge the hosted-AI disclosure and click &lt;strong&gt;Analyze with AI&lt;/strong&gt;. Copy the reply, download the text report, or print to PDF.&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%2Fx4z8hfpelx8f65n78yd8.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%2Fx4z8hfpelx8f65n78yd8.png" alt="ScopeProof's actual interface before loading any results" width="799" height="562"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;There is also a &lt;a href="https://github.com/tanzilamd/scopeproof/blob/main/docs/screenshots/live-review-local.png" rel="noopener noreferrer"&gt;live-analysis screenshot from the local app&lt;/a&gt;, captured with real Backboard inference.&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/tanzilamd" rel="noopener noreferrer"&gt;
        tanzilamd
      &lt;/a&gt; / &lt;a href="https://github.com/tanzilamd/scopeproof" rel="noopener noreferrer"&gt;
        scopeproof
      &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;ScopeProof&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Keep the work. Protect the scope.&lt;/strong&gt; A small, evidence-first freelance scope-change assistant for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12" rel="nofollow"&gt;Hacktoberfest Open-Source AI Challenge: Week 2 — Side Hustle&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;ScopeProof compares an original agreement with up to eight client requests. It suggests &lt;strong&gt;Included&lt;/strong&gt;, &lt;strong&gt;Extra work&lt;/strong&gt;, or &lt;strong&gt;Needs clarification&lt;/strong&gt;, checks supporting quotes verbatim, and leaves every final decision to the freelancer. User-entered hours and rates produce a proposed change quote; an editable reply and text/print report make the next conversation easier. Nothing is sent to a client automatically.&lt;/p&gt;
&lt;p&gt;Live demo: &lt;strong&gt;&lt;a href="https://scopeproof.onrender.com" rel="nofollow noopener noreferrer"&gt;https://scopeproof.onrender.com&lt;/a&gt;&lt;/strong&gt; · &lt;a href="https://dev.to/tanzil_ahmed_e6a826eee8c4/scopeproof-an-evidence-first-answer-to-one-more-thing-3odi" rel="nofollow"&gt;Published DEV submission&lt;/a&gt;. See &lt;a href="https://github.com/tanzilamd/scopeproof/docs/evidence/deployment.json" rel="noopener noreferrer"&gt;deployment evidence&lt;/a&gt; and &lt;a href="https://github.com/tanzilamd/scopeproof/docs/evidence/manual-public-verification.json" rel="noopener noreferrer"&gt;manual public verification&lt;/a&gt; for how the workflow was checked.&lt;/p&gt;
&lt;p&gt;&lt;a rel="noopener noreferrer" href="https://github.com/tanzilamd/scopeproof/docs/screenshots/overview.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Ftanzilamd%2Fscopeproof%2FHEAD%2Fdocs%2Fscreenshots%2Foverview.png" alt="Actual ScopeProof interface, before loading sample or live results"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Quick start&lt;/h2&gt;
&lt;/div&gt;
&lt;p&gt;Requires &lt;strong&gt;Node.js 24&lt;/strong&gt; (tested on 24.19.0). No production dependencies, database, login, build step or external fonts.&lt;/p&gt;
&lt;div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"&gt;
&lt;pre&gt;npm ci --ignore-scripts
npm &lt;span class="pl-c1"&gt;test&lt;/span&gt;
npm start&lt;/pre&gt;

&lt;/div&gt;
&lt;p&gt;Open the address printed by the server on port 3000. The labelled illustrative sample works without credentials and…&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/tanzilamd/scopeproof" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The application is MIT licensed. The repository includes &lt;a href="https://github.com/tanzilamd/scopeproof#quick-start" rel="noopener noreferrer"&gt;setup instructions&lt;/a&gt;, &lt;a href="https://github.com/tanzilamd/scopeproof/blob/main/docs/architecture.md" rel="noopener noreferrer"&gt;architecture notes&lt;/a&gt; and &lt;a href="https://github.com/tanzilamd/scopeproof/tree/main/docs/evidence" rel="noopener noreferrer"&gt;test evidence&lt;/a&gt;.&lt;/p&gt;

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

&lt;p&gt;I built ScopeProof with OpenAI Codex assistance for implementation and testing.&lt;/p&gt;

&lt;p&gt;I kept it to one Node.js 24 app on Render with plain HTML, CSS and JavaScript. There are no production package dependencies, login or database. The Backboard key stays on the server.&lt;/p&gt;

&lt;p&gt;The live model is &lt;strong&gt;&lt;a href="https://huggingface.co/Qwen/Qwen3-32B" rel="noopener noreferrer"&gt;Qwen3-32B&lt;/a&gt;&lt;/strong&gt;, selected explicitly as &lt;code&gt;qwen/qwen3-32b&lt;/code&gt; through Backboard's OpenRouter integration. Qwen publishes its weights under &lt;strong&gt;Apache 2.0&lt;/strong&gt;. Each analysis gets a separate conversation, with memory and web search off, followed by an attempt to delete the temporary thread.&lt;/p&gt;

&lt;p&gt;My main rule for the app was simple: a suggestion should point to something the person can check.&lt;/p&gt;

&lt;p&gt;The server checks every quotation against the agreement and requires a result for every request. Invented quotes are removed; unsupported definitive suggestions become &lt;strong&gt;Needs clarification&lt;/strong&gt;. A real quote can still be misinterpreted, so the user keeps the final decision.&lt;/p&gt;

&lt;p&gt;Prices come from user-entered hours, rate and currency. The calculation runs in code, rounding each line before adding the total. Blank hours show &lt;strong&gt;Estimate incomplete&lt;/strong&gt;, with a count of the estimates still needed. An explicitly entered zero is valid. Editing the agreement clears the old review.&lt;/p&gt;

&lt;h3&gt;
  
  
  Checking that it works
&lt;/h3&gt;

&lt;p&gt;There are &lt;strong&gt;51 passing automated tests&lt;/strong&gt; and &lt;strong&gt;19 passing Chromium checks&lt;/strong&gt; for validation, quotes, arithmetic, overrides, errors, exports and desktop/mobile use. Provider calls in these suites are mocked. Desktop/mobile axe scans found zero WCAG 2.1 A/AA violations; a full accessibility audit remains outside that coverage.&lt;/p&gt;

&lt;p&gt;A separate local browser check used real Backboard inference: five expected sample labels, matching quotes and &lt;strong&gt;11.96 seconds&lt;/strong&gt; of model-response latency.&lt;/p&gt;

&lt;p&gt;The first public call exposed a deployment mistake: a cloud credential reference was unusable on Render. Setting the Backboard key directly in Render fixed it. My manual public walkthrough then covered decisions, estimates, replies and export.&lt;/p&gt;

&lt;p&gt;For the model choice, eight fixed synthetic requests covered exclusions, limits, ambiguity and hostile instructions. Expected labels preceded inference. Both models used the same prompt and validator through one Backboard key.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Model&lt;/th&gt;
&lt;th&gt;Six-request scope/limits case&lt;/th&gt;
&lt;th&gt;Two-request hostile/uncertain case&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Qwen3-32B&lt;/td&gt;
&lt;td&gt;6/6 expected labels; 27.2 s&lt;/td&gt;
&lt;td&gt;2/2 expected labels; 16.8 s&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Qwen3.5-9B&lt;/td&gt;
&lt;td&gt;Timed out at the 45 s limit; no result accepted&lt;/td&gt;
&lt;td&gt;2/2 expected labels; 40.7 s&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;These are accepted results after validation. The timeout is not six incorrect answers, and eight examples are too few for a general accuracy claim. The &lt;a href="https://github.com/tanzilamd/scopeproof/blob/main/docs/evidence/comparison.json" rel="noopener noreferrer"&gt;full comparison&lt;/a&gt; supported keeping Qwen3-32B for this demo.&lt;/p&gt;

&lt;p&gt;A catalogue-listed Qwen3-8B candidate returned a provider deprecation error. Actual inference matters; the app never silently switches to a closed model.&lt;/p&gt;

&lt;p&gt;The first successful five-request call cost &lt;strong&gt;$0.000145&lt;/strong&gt;. Through the local live-browser check, the observed balance change was &lt;strong&gt;$0.001505&lt;/strong&gt;, with auto-reload off. The &lt;a href="https://github.com/tanzilamd/scopeproof/blob/main/docs/evidence/costs.json" rel="noopener noreferrer"&gt;billing evidence&lt;/a&gt; records that testing checkpoint, not future public-service costs.&lt;/p&gt;

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

&lt;p&gt;Open weights give this project somewhere to go if the current hosting arrangement stops fitting. Qwen's published weights and permissive license allow another host, quantization or a future local deployment to be explored with the same workflow and test cases. The MIT code makes the quote checks inspectable too.&lt;/p&gt;

&lt;p&gt;Clients can have different confidentiality requirements, and small businesses need predictable costs. Backboard made a two-model comparison possible through one key. The smaller model's timeout showed why catalogue price alone was insufficient.&lt;/p&gt;

&lt;p&gt;Today's app still sends input to hosted services. I have not implemented offline inference or measured an advantage over a closed-model baseline.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Render:&lt;/strong&gt; the single Render service hosts the interface and server-side AI integration.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Backboard:&lt;/strong&gt; two open-weight models were compared through one Backboard key, with published cases, classifications, latency and a recorded timeout.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  A few boundaries
&lt;/h3&gt;

&lt;p&gt;Avoid confidential client material in the public demo. Live input goes to Backboard and its model provider; memory off and attempted thread deletion do not guarantee zero retention.&lt;/p&gt;

&lt;p&gt;A model can misread a real quote or follow hostile instructions. ScopeProof is a review aid, not legal advice or proof of client approval. The demo has input limits, timeouts and request quotas, but those in-memory quotas reset on restart and are not guaranteed account spending caps. Extra work still needs human review and written approval.&lt;/p&gt;

&lt;p&gt;Credits: the &lt;strong&gt;Qwen team&lt;/strong&gt; for the Apache-2.0 models; &lt;strong&gt;Backboard, OpenRouter and Render&lt;/strong&gt; for hosted services; &lt;strong&gt;Playwright&lt;/strong&gt; (Apache 2.0) and &lt;strong&gt;axe-core&lt;/strong&gt; (MPL 2.0) for test tooling. The application code is MIT licensed.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>opensource</category>
      <category>ai</category>
    </item>
    <item>
      <title>Anyone participating here?</title>
      <dc:creator>Tanzil Ahmed</dc:creator>
      <pubDate>Wed, 14 Oct 2026 14:21:00 +0000</pubDate>
      <link>https://dev.to/tanzil_ahmed_e6a826eee8c4/anyone-participating-here-2hjp</link>
      <guid>https://dev.to/tanzil_ahmed_e6a826eee8c4/anyone-participating-here-2hjp</guid>
      <description></description>
    </item>
  </channel>
</rss>
