Blog

AI (Artificial Intelligence)

IT Today: Essential Technology News (November 2026)

In November, we’re putting human–AI collaboration (dare we call it symbiosis?) under the microscope. The gains come from a smart division of labor: AI takes on the analysis and the repetitive tasks, while people decide what’s worth doing, check the results, and take ideas further.

Two new studies, summarized by Jakob Nielsen, show how researchers use AI in their day-to-day work. A team led by Mihai Codreanu, from Google, Google DeepMind, and MIT FutureTech, analyzed 15 million interactions with Gemini and surveyed 637 scientists. The result: 47% use AI daily, and the average respondent saves almost 7 hours a week – nearly a full working day.

The second study, carried out by Anthropic on more than 207,000 research sessions with Claude and Claude Code, shows that scientists turn to AI mostly where it is most useful, from data analysis to writing. In other words, specialists quickly figure out where the technology helps them and use it exactly there.

The same pattern shows up in other news this month. At Google, AI agents cleaned up thousands of outdated references in the codebase, and human reviewers approved more than 95% of the changes.

For those who want to work in tech, this is an opportunity: anyone who knows how to use AI as a partner – which tasks to delegate, how to configure it, and how to check its results – can gain back hours every week. You can learn to work this way, for example, in the AI Automation Specialist course at NewTech Academy.

We cover these topics, along with the latest launches in the field, in detail below.

AI News

Claude Fable 5.2 and Mythos 5.2: Faster at the Toughest Technical Tasks

On November 9, Anthropic launched Claude Fable 5.2 and Mythos 5.2, two versions of the same model: Fable, available to everyone, and Mythos, reserved for vetted organizations. The new versions respond up to 30% faster, reach 69.9% on the FrontierSWE v2 benchmark, and formalized Fermat’s Last Theorem in 30 hours, coordinating around 1,000 agents.

Grok Bot Can Make Phone Calls For You, But You’re Still the One Responsible

SpaceXSI has enabled phone calls in Grok Bot: the AI agent can wait on hold, book appointments, and negotiate bills on U.S. numbers, and when it doesn’t know an answer, the user types it in during the call. The terms of service, however, make the user solely responsible for the calls and for complying with the law, and the bot has to say it is a virtual assistant.

Gemini 4 Hydrogen, Google’s Fast and Affordable Model for AI Agents

On November 12, Google launched Gemini 4 Hydrogen, a model built for everyday agentic tasks. It costs $1.50 per million input tokens and $7.50 per million output tokens, is about 2.5 times faster than Gemini 3.8 Flash and, on the Artificial Analysis index, narrowly beats even the flagship Gemini 4 Argon.

Digital Marketing News

Google’s Third Core Update of 2026 Arrives Right Before Black Friday

On November 10, Google began rolling out the November core update, its third major update this year. The global rollout, covering all languages, takes about two weeks and overlaps with Black Friday, so online stores may see fluctuations. The advice: don’t make changes based on the first few days, and analyze the results a week after the rollout ends.

AI Visibility: Being Recommended Matters More than Being Mentioned

Patrick Reinhart of Conductor says in a conversation with Search Engine Journal that the most important metric in AI search is whether an LLM recommends your brand, since a mere mention can also be negative. He recommends a fixed set of prompts tracked across several models, log file analysis, and investment in video: YouTube is the most-cited source.

Holidays 2026: Shoppers Use AI, but Authenticity Makes the Difference

According to Marketing Dive, Deloitte estimates that American households will spend an average of $1,654 over the holidays, 4% more than last year. More than a third of consumers are willing to shop with AI’s help, but they also consult more sources. To stand out, Levi’s and e.l.f. bet on stop-motion and handmade creative work.

Cybersecurity News

Hackers Defrosted the Fridges in Hundreds of French Supermarkets

On the night of October 11–12, attackers remotely triggered forced defrosting in the refrigeration systems of chains such as Intermarché, Coopérative U, Carrefour and E.Leclerc. Temperatures rose as high as 50 °C, products had to be thrown away, and fires broke out. The attack exploited a Danfoss vulnerability, and the affected units were running outdated firmware.

Through Optimizely, Malicious Code Reached the Websites of 178 Organizations

On November 16, attackers exploited a vulnerability in Optimizely, a website testing tool, and slipped malicious code onto the pages of 178 organizations, including Zoom, BBC, CNN and HSBC. The code could steal card details and hijack crypto payments. Optimizely has fixed the issue, and visitors are advised to update their iPhones and check their transactions.

The Microsoft 365 Outage, Explained: a Hidden Bug Spread by Automatic Recovery

According to a preliminary report, the November 12–13 Microsoft 365 outage, which affected Outlook and other services in the suite, started with a defect in a fix that had been live since July and got worse on newer servers with more cores. Tests, run on less powerful machines, didn’t catch it. Automatic recovery then moved connections around and spread the problem globally, leaving more than 80% of servers unresponsive.

Programming News

TIOBE Index: Dart Enters the Top 20 Programming Languages for the First Time

In the November TIOBE index, Dart entered the Top 20 for the first time, at number 20, boosted by Flutter, Google’s framework for building apps for multiple platforms from a single codebase. Python remains the leader at 20.86%, followed by C and Java. Rust dropped out of the Top 10, making room for Go, and Assembly returned to the Top 20.

Windows 11 Records Which AI Coding Agents You Use

Researcher Xusheng Li discovered a Windows 11 component that notes which AI coding agents you run, based on names such as codex.exe, claude.exe, or cursor.exe. The report contains only yes/no flags and is sent only if optional diagnostic data is turned on. Microsoft hasn’t explained the purpose, and the option can be turned off under Settings > Privacy & security > Diagnostics & feedback.

Forgotten TODO Comments Send AI Agents Down the Wrong Path

Google engineers found that AI coding agents get sidetracked by TODO comments pointing to old bugs: in 6% of the cases where they come across them, they start fixing problems nobody asked them to, using 77% more tokens. The fix: agents that cleaned up thousands of references, with more than 95% of the changes approved by people.

UI/UX and Graphic Design News

Context Is the New Interface: Design Moves Between Apps

On UX Collective, Aurélie Radom argues that the future of design is playing out between products: AI assistants can start from the user’s intent and connect information from messages, maps, and calendars. For designers, context becomes working material: what the system remembers, when it asks for confirmation, which tools it uses, how it recovers from a mistake, and how much autonomy it gets.

Designers Remain More Original than AI, a Benchmark Shows

A Kombai benchmark, summarized by Jakob Nielsen, compared 15 AI models with designers on 160 briefs. The top models met 99% of the requirements, but their designs looked very similar to one another, while the designers came up with more varied solutions tailored to each domain. Nielsen’s advice: ask AI for several variations and combine the best ideas yourself.

“The Age of Agile Must End”: Strategy Before Speed

On UX Collective, Michael Burnett argues that Agile and Scrum are incompatible with research, design, and scalable development, because they put delivery speed ahead of strategy. In his view, people aren’t machines, and a product can’t be defined by a two-week sprint. The solution: long-term strategy, incremental releases and investment in design systems.

***

On the fragility of systems and what makes us resilient

This month we talked about agents that make calls on our behalf, about models that write code and create designs faster than we can check them, and about studies showing how much time people who work alongside AI save. The same month, however, reminded us how fragile the systems all of this relies on really are. In a single week, a hidden bug took Microsoft 365 down globally, a software error disrupted online services at several U.S. banks, and a website testing tool became the entry point for an attack on 178 organizations.

The Microsoft outage kept millions of people away from email and online meetings for a whole day and, according to Reuters, generated more than 7.4 million reports on Downdetector. Government agencies in Australia, Japan and the U.S. were affected, as was Britain’s NHS, which fell back on paper records, while a Taiwanese financial firm confirmed trades by fax. On Monday, November 16, customers of Chase, Wells Fargo, Bank of America, Truist and U.S. Bank reported problems with their online accounts and transfers. The cause, according to AP, was a software bug in a JPMorgan Chase hardware system, not a cyberattack, but the effects also reached customers of other banks trying to transact with Chase.

The common thread is the concentration of power: more and more work runs through a handful of providers, and a single defect can spread globally – sometimes through the very mechanisms built to protect us, as Microsoft’s automatic recovery did. Resilience, however, doesn’t come only from more technology. It comes from tests that resemble reality, updates applied on time and backup plans that people know how to use. Last but not least, resilience is built before the crisis, and it depends on people who know how to keep going when the system stops.


Find out all the details about our courses!
Fill in the fields below, and we will contact you within the next 24 hours.

    We're waiting for you at NewTech Academy