How to Pick the Right Web AR Delivery Route Before You Start Building

·1 views

Developers building augmented reality experiences for the web face a critical early choice between three delivery routes: handing off to the device's native viewer (AR Quick Look or Scene Viewer), running a full AR session inside the browser via WebXR, or shipping a native app. Each route involves distinct trade-offs in reach, interactivity, and asset delivery, and defaulting to whichever a vendor already built is a common mistake. Unlike native apps, which download assets once over Wi-Fi after a deliberate install, web AR pages must re-download assets every time a link is opened, often on poor connections with no user commitment. This imposes a hard ceiling on scene complexity, making unglamorous optimisation work — compressed geometry, GPU-friendly texture formats, and progressive loading — non-negotiable rather than optional. The article urges developers to feature-detect browser AR capabilities at build time and always provide a meaningful fallback rather than an apology page.

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)

Contradictory wiki data won't make AI agents hallucinate — but it will break them

A developer built a controlled testbed to examine how ingest quality affects an AI agent that navigates a wiki using search and read tools, rather than relying on vector databases or RAG. The agent scored 4 out of 4 on questions when given a clean wiki, but its correct-answer rate dropped to near zero when contradictory or stale duplicate pages were introduced. Crucially, the agent did not hallucinate confidently wrong answers — instead, it detected conflicting information and refused to commit to a single response. The real cost of poor ingest quality is therefore not misinformation but a collapse in the wiki's authority, forcing the agent to read multiple pages, hedge its answers, and consume more resources per query. The researcher concluded that this failure mode is easy to overlook because no single answer is technically wrong, yet the system's core value — delivering one trusted answer — is effectively destroyed.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Engineer launches open resource to help developers use AI coding tools responsibly

A software engineer has launched loveyourclanker.org, a free, non-profit open web resource aimed at helping developers interact more intentionally with AI coding tools and agents. The site outlines various patterns engineers can consciously adopt to stay in control, maintain code quality, and improve efficiency without over-relying on automation. The creator was motivated by concerning trends in the developer community, including token leaderboards, engineers automating away human roles, and others abandoning AI tools entirely due to stress. The goal is to normalize open discussion about how these tools are used and encourage approaches that preserve human agency in the development process. The project is fully open-source, and contributions via pull requests are welcome.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Cloud vs. Local LLMs for Scheduled Curation: A Practical Operational Tradeoff

Scheduled LLM curation jobs run silently overnight to deduplicate, summarize, and re-rank agent memory without any human oversight, making reliability and failure modes especially critical. Unlike interactive workflows where errors are visible, headless cron jobs can fail silently — hanging on unanswered prompts or losing data on pod restarts. Running curation against hosted cloud APIs is quick to set up and benefits from frontier model quality, but costs scale with memory size and every run sends potentially sensitive data off-premises. Pointing the same workload at a locally hosted model addresses both privacy and recurring token costs, but introduces a new layer of infrastructure complexity including GPU management, node affinity, and image maintenance. The right choice depends on data sensitivity, curation complexity, and an organization's operational capacity to manage local model deployments.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Insufficient source content to generate accurate headline

The provided article contains no readable text beyond metadata links and URLs. No factual summary can be responsibly produced without fabricating information. Please provide the full article text for accurate processing.

0 comments Read more at Hacker News