Developer shares hard lessons from building an AI tutoring platform with talking avatar

·1 views

A developer has built COSMIQ, an AI-powered K-12 tutoring platform featuring a talking avatar professor named Dr. Lynch who teaches on an interactive whiteboard and grades student responses in real time. The builder found that offering a no-signup demo directly on the homepage was the single most effective conversion tool, as users needed to experience the product before trusting it. Voice and avatar technology proved costly, prompting a tiered approach that prioritized quality audio for learners while capping overall spend rather than restricting features behind paywalls. With no budget for paid ads, the team relied on programmatic SEO — generating hundreds of pre-rendered pages using public school and district data — as their primary growth channel. Building for children also imposed strict requirements around privacy, content filtering, and data handling that ruled out shortcuts common in typical SaaS development.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How GitHub Actions Pipelines Actually Work, Explained Simply

A developer education piece published on DEV Community uses a fictional senior-junior engineer dialogue to break down what happens after a developer runs git push. When code is pushed, GitHub stores the new commits, updates the branch pointer, and fires an internal push event to notify any registered listeners. GitHub Actions operates on a similar principle to webhooks, where an event triggers a configured receiver to take action. The system then checks the repository for YAML workflow files inside the .github/workflows/ directory to determine whether any pipeline should run. If no matching workflow files exist or none are configured to trigger on a push event, no pipeline starts and the process ends silently.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Open-Source Slack Template Lets Teams Run Five AI Marketing Agents via One Lead

A Vercel Labs open-source project called the 'marketing team eve template' lets users manage five specialist AI marketing agents — covering content, SEO, social media, email, and product positioning — through a single team lead in Slack. Users message the lead agent, which briefs the appropriate specialist and returns deliverables such as Notion pages, Typefully drafts, or Resend email campaigns. The architecture is intentionally one level deep, limiting handoffs to reduce context loss, with the exception of a newsletter workflow where two specialists collaborate in sequence. All irreversible actions, such as sending emails or publishing drafts, require explicit human approval via a Slack button before execution. A shared brand context document, maintained solely by the product marketer agent, ensures consistent messaging across all specialist outputs.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

React useInfiniteScroll Hook Simplifies Infinite Scrolling for Feeds and Chats

A custom React hook called useInfiniteScroll, available from the @reactuses/core library, aims to replace complex manual scroll-listener logic with a single function call. The hook accepts a ref to a scrollable element, a load-more callback, and an optional configuration object, handling edge detection, cleanup, and SSR safety automatically. Developers can configure scroll direction — bottom, top, left, or right — making it suitable for chat history loading and horizontal carousels, not just standard feeds. A distance option allows content prefetching before the user reaches the scroll boundary, while a preserveScrollPosition flag prevents viewport jumps when new content is inserted above the current view. The hook is written in TypeScript and built on top of an underlying useScroll utility, keeping its core implementation under 50 lines.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How developers can apply Shopify accessibility fixes without losing theme updates

A common challenge for Shopify developers is that customizing a theme effectively forks it, meaning every future upstream update must be manually reconciled. This tension often discourages teams from making accessibility improvements to live themes, since fixes to semantic HTML and landmark structure require direct edits to Liquid templates. A proposed workflow addresses this by isolating styling and scripting changes into separate custom asset files that rarely conflict with updates, while keeping Liquid edits minimal and clearly marked. Developers are advised to tag every Liquid change with a consistent comment marker, making it easy to locate all modifications with a single search command. Maintaining a duplicate of the live theme disconnected from version control also helps preserve access to Shopify's built-in update notifications.

0 comments Read more at DEV Community