Why websites feel laggy when users click: how INP measures responsiveness

·2 views

A page can appear fully loaded quickly yet still stall when a visitor taps a menu or button, a problem Google now measures with Interaction to Next Paint (INP). INP evaluates how responsive a site is throughout a visit, rather than only how fast content first appears. Google rates an INP of 200 milliseconds or less as good, 200 to 500 milliseconds as needing improvement, and anything above 500 milliseconds as poor. Delays arise because the browser runs JavaScript on a single thread, so heavy scripts such as third-party tracking tags, oversized bundles, and excessive framework re-renders can block a click from registering visually. Developers can reduce this lag by trimming or deferring these scripts and simplifying page structures.

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)

Ubicloud and Blacksmith compared on GitHub Actions runner pricing and features

Ubicloud advertises the lowest published Linux rate for a 2 vCPU runner at $0.00125 per minute, though new accounts default to a $0.002 premium tier. The company runs an open-source cloud on bare metal in Germany, licensed under AGPL v3, and offers no Windows or macOS runners. Blacksmith sells managed runners on its own bare metal in four regions, with 3,000 free minutes monthly and support for Linux, Windows and macOS. The comparison draws on pricing and documentation published by each vendor as of 21 September 2026, with unpublished figures marked as such.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Python checker verifies that each startup-shutdown tracker row links to a live, on-topic source

A developer built a standard-library Python script to audit a tracker of Indian startup shutdowns in 2026, since broken or irrelevant source links can undermine the list. The script fetches each cited URL and flags HTTP errors, connection failures, and pages that do not mention the company. The article illustrates the problem with a YourStory link for AI fashion startup Alle that returned 403 behind Cloudflare, so its facts were sourced from other outlets instead. The tracker's reason column preserves each source's own wording, such as Zupee being described as 'on the verge' of shutdown rather than closed. The post includes sample rows for Alle, Satvacart and Zupee and the full checker code, which writes results to a report CSV.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

ChatGPT Wallet Launches With Per-Purchase Approval for Agent Checkouts

OpenAI released ChatGPT Wallet on November 13, 2026, letting an agent inside dots complete checkouts after the user approves each transaction with Face ID or fingerprint. Payments use network-issued tokens from American Express, Mastercard, or Visa, so neither the agent nor the merchant sees the actual card number. Early access is limited to US residents aged 18 and over with eligible US-issued cards and a ChatGPT Pro subscription, and pre-approved purchases under spending limits are planned for later. The accompanying developer guidance urges merchants to publish accurate prices, availability, and terms in crawlable form and to accept signed agent requests rather than block them. The article also notes that United has barred AI agents from booking or servicing its tickets without asking first.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Agent Approval Should Hinge on Action Consequences, Not Model Confidence

A DEV Community article argues that approval systems for AI agents often ask whether an action is safe rather than what happens if it is wrong and whether it can be undone. It uses a hypothetical procurement agent to illustrate how routing every payment to a human can produce rubber-stamp approvals and late reviews after key decisions are already made. The author proposes setting review thresholds by consequence and reversibility, reserving human approval for actions that send data externally, change permissions or financial details, create external commitments, or rely on ambiguous instructions. Approval requests should specify the proposed action, its scope, the authority used, and the resulting effects. The piece concludes that confidence scores can help prioritize review but should not determine whether it is required.

0 comments Read more at DEV Community