Editor's note
Today's issue leans toward day-to-day operations. The scheduled Next.js security release and Google's revised wording for the ‘Crawled – currently not indexed’ status are both things site owners can act on right away. On the AI side, Salesforce's voice-AI acquisition and OpenAI's everyday-task benchmark point to a shift from raw capability toward practical, measurable usefulness. It's a day where patching, checking and evaluating matter more than headline model launches.
WebNext.js Blog
Next.js ships November security release: seven fixes in 16.4.5 and 15.5.29
On November 3, Next.js published its pre-announced security release, fixing seven vulnerabilities in v16.4.5 (Active LTS) and v15.5.29 (Maintenance LTS). Issues include cross-tenant cache disclosure with Cache Components and use cache (one rated High), cache poisoning and response mix-ups in self-hosted setups, cookie exposure in Edge routes, and an SVG-rejection bypass in Image Optimization. The team recommends upgrading to the patched versions.
Context
Next.js now announces security releases about a week in advance and then ships patches on a set date, and this release follows that pattern. Most of the issues only apply under specific conditions — Cache Components, self-hosting with next start, or particular reverse-proxy setups — so not every site carries the same risk. But for multi-tenant or logged-in sites that rely on caching, these are the kind of bugs that can show one user's data to another, which raises the priority.
HaLVision's take
For the Next.js sites we maintain, the plan is simple: check the version and move both the 16.4 and 15.5 lines to the latest patch. Even for a mostly static company site, the upgrade is small enough that ‘it probably doesn't apply to us’ isn't a good reason to wait. If your site separates cached content by member, locale or location, spot-check that pages render correctly after updating.
Read source ↗WebGoogle Search Console Help
Google revises Search Console's ‘Crawled – currently not indexed’ description
On November 3, Google updated the Page indexing report help text in Search Console for the ‘Crawled – currently not indexed’ status. The new wording says Google doesn't index every page on the web, that this is a common message for many sites, and that it doesn't necessarily indicate an issue needing attention. It also points to the URL Inspection tool and a form for reporting suspected site-wide indexing bugs.
Context
This status is one of the most frequently questioned items in Search Console, and many site owners read it as a sign they'd done something wrong. The old text was brief — the page may or may not be indexed later, no need to resubmit. The revision makes it explicit that non-indexing is normal, while giving clearer steps for when to dig deeper.
HaLVision's take
When a small-business site shows this status, the practical move is to sort pages by whether you actually need them indexed. If key pages — service pages, contact, location info — are affected, look at thin or duplicated content and weak internal linking. If it's tag archives or parameter URLs, it's usually not worth worrying about.
Read source ↗AISalesforce Newsroom
Salesforce to acquire voice-AI company Cresta for Service Cloud and Agentforce
On November 2, Salesforce announced a definitive agreement to acquire Cresta, a voice-first AI platform for customer service. Cresta's AI agents can handle complex calls end-to-end, coach human reps in real time and analyze conversations. The deal is expected to close in Q1 of fiscal 2028, after which Cresta will be integrated with Service Cloud and Agentforce.
Context
Compared with chat, phone calls involve interruptions and rephrasing, which has made natural AI handling harder. Salesforce wants to connect CRM data such as customer history with live conversation intelligence on a single platform. Reports put the price at about $2.7 billion, though Salesforce did not officially disclose terms.
HaLVision's take
This is an enterprise deal, but how much phone handling can be delegated to AI is a live question for local shops and clinics too. As these capabilities become standard in major CRMs, we expect cheaper SMB-grade versions to follow within a few years. A useful step now is to write down your common customer questions and answers as an FAQ — that groundwork pays off whichever tool you eventually pick.
Read source ↗AIOpenAI
OpenAI releases LifeVal, a benchmark for everyday-task help
On November 2, OpenAI released LifeVal, a benchmark measuring how well AI helps people with everyday tasks such as finding childcare, managing household finances and handling government paperwork. Tasks are derived from the US Bureau of Labor Statistics' American Time Use Survey, the Multinational Time Use Study and a survey of roughly 4,000 people. Models are scored on whether they combine sources, respect user constraints and give guidance that actually gets the task done.
Context
Most AI benchmarks focus on specialist skills like math or coding, which critics say don't reflect what ordinary people ask AI to do. LifeVal tries to close that gap by building tasks from how people actually spend their time. It is, however, a benchmark created and reported by the model developer itself, so cross-vendor comparisons and independent validation are still to come.
HaLVision's take
The scores matter less than the evaluation idea: not just listing information, but getting the user to a finished task within their constraints. That's a good lens for SMBs piloting AI too — test whether a tool can actually finish your own common tasks, on a small scale. Treat vendor-reported numbers as a starting point and compare tools on your own data and workflows.
Read source ↗AIGoogle Blog
Google turns Gemini and Search off by default for K-12 student accounts, expands Guided Learning
On November 3, Google announced education updates under which Gemini, Gemini Notebook, Gemini in Classroom and Google Search will be off for K-12 student accounts until school administrators explicitly turn them on. Google also made Guided Learning — which coaches students step by step instead of giving answers — globally available in Classroom, with a Search pilot starting this month in select US schools. Schools will get rollout details in the coming months.
Context
Schools have increasingly asked for control over how AI is used in class, alongside concerns about learning outcomes. This change bundles AI features and Search into an opt-in model chosen by each school. Because Search is included, students at schools that keep AI off would also lose Search access, which has drawn debate.
HaLVision's take
Schools are the direct audience, but the ‘off by default, admin decides’ pattern is a sensible model for any organization rolling out AI. Inside a company, enabling AI tools team by team after agreeing on scope tends to prevent problems. For tutoring and education clients, we'd suggest preparing a clear explainer page for parents.
Read source ↗