The AI Times
EN
Volume I · Daily editionDaily roundup of AI newsMCPMD
EN
THE AI TIMES
Gemini · Claude · ChatGPT

AI takes on more work, users get more control

Gemini brings together custom instructions and connected apps, while Gemini Notebook lets you comment on an outline before creating a presentation. Claude Duet is designed to track important things throughout the day. For AI shopping, Walmart promises approvals and spending limits; meanwhile, ads in ChatGPT are reaching more countries. Today’s changes bring more features and a reason to check what your assistant can access and do on your behalf.

Gemini · Customization21 Oct 2026

Gemini brings custom instructions and connected apps together in one place

The new Customize menu aims to make it easier to configure responses and review the services the assistant can access.

Gemini customization · settingsGoogle Workspace Updates · D1
Illustration based on the facts in the story

Google is introducing a central hub in Gemini for custom instructions and connected apps.

You can use it to set your writing style or preferred response format, for example. Google Workspace Updates Custom skills can handle recurring tasks while preserving your writing tone. The menu includes over 40 external services alongside Google apps. You can review permissions or disconnect apps at any time. The hub is rolling out on the web, mobile, and Windows and macOS computers. It is intended for all account tiers, including personal Google accounts. Most connected apps are currently available in English in the US; other listed regions include Brazil and India.

Claude · Personal assistant21 Oct 2026

Claude Duet aims to summarize important items from apps each morning

According to The Verge, an experimental assistant is coming for professionals and small business owners.

The Verge describes Claude Duet as an assistant that monitors connected apps and flags items needing attention. It is designed to send a short summary of the most important items each morning. The Verge AI

Gemini · Navigation21 Oct 2026

Talk to Gemini while navigating with CarPlay

According to The Verge, Google Maps in the car accepts follow-up voice questions.

According to The Verge, drivers with Apple CarPlay can use conversational Gemini in Google Maps hands-free. The rollout has begun in the US and is set to continue globally. The Verge AI

AI · Shopping21 Oct 2026

Walmart will let AI assistants shop with a spending limit

Purchases are meant to remain under the customer’s control, The Verge reports.

Walmart plans to let AI assistants shop on its website on customers’ behalf next month.

Checks and purchase approvals are planned as part of the service. The Verge AI Customers will be able to set a spending limit. The rules prohibit assistants from changing the customer’s payment method. An assistant must not act beyond the shopper’s intent.

ChatGPT · Advertising21 Oct 2026

Ads in ChatGPT expand to another 26 countries

For users, this means wider use of ads when searching for and comparing products.

OpenAI is announcing the expansion of ads in ChatGPT to 26 new countries. The available announcement does not list those countries. OpenAI News · OpenAI News According to the announcement’s headline, over 100,000 advertisers have already used the platform. A separate technical article from OpenAI describes how ads are kept separate from responses. The same article describes using language models to understand the intent of a conversation.

AI · Healthcare21 Oct 2026

US doctors turn to ad-funded AI for answers

According to Ars Technica, nearly two-thirds of US doctors use OpenEvidence.

The sources doctors rely on matter to patients. Ars Technica reports on the growth of OpenEvidence, a search engine that helps navigate medical journals and clinical guidelines. Ars Technica AI The company reports over 42 million queries last month. Ads fund the platform. Daniel Nadler founded the company in 2022.

AI · Research21 Oct 2026

Claude designed proteins against a scorpion toxin, but there is no treatment yet

According to The Decoder, the laboratory result does not yet demonstrate an effective antiserum.

For patients, this is promising research whose therapeutic effect has yet to be verified. According to The Decoder, Claude designed proteins that bind to dangerous scorpion neurotoxins. The Decoder Of more than 318,000 designs, researchers tested 211 in the lab. Of those, 17 met the specified criteria. The two strongest proteins bind very tightly to the Cn2 toxin from a Mexican scorpion.

What experts say

Ethan Mollick

professor at Wharton, author of One Useful Thing

AI handles more work, people still assess the result

Ethan Mollick describes the shift from chatting with AI to assigning work to autonomous assistants. He had the article’s graphics prepared in Claude Projects using Opus 5.5 and selected from the options offered. He warns, however, that this way of working means losing some control over how AI reached its result.

oneusefulthing.org
What experts say

Simon Willison

developer, creator of LLM and Datasette

An AI-built game also needs to be played and refined

Simon Willison describes creating Battle Drones largely with Claude Code. After 12 major revisions, he has a game for mobile and desktop; he barely reviewed the code but played it repeatedly. The example shows the difference between creating a working game and refining an experience that players enjoy.

simonwillison.net
Gemini Notebook · Creation21 Oct 2026

Gemini Notebook lets you review an outline before creating a presentation

You can now guide a presentation or video step by step and add feedback before it is generated.

Gemini Notebook · creation optionsGoogle Workspace Updates · D1
Illustration based on the facts in the story

Google is adding reviews of presentation outlines and video production notes to Gemini Notebook. This lets you request changes before the finished output is created. Google Workspace Updates The tool recommends topics, formats and visual styles based on the sources provided. It fills in the relevant options with specific instructions from your prompt. The Review toggle lets you write feedback on the outline or notes. Quick creation without an additional review remains available. The rollout began on 16 October and, according to Google, may take up to 15 days.

AI · Policy21 Oct 2026

China protests the DeepSeek ban in US government agencies

According to The Decoder, the dispute concerns access to AI in government systems.

According to The Decoder, the ban applies to US government systems; the article does not describe it as a ban for all users. China is calling for it to be lifted. The Decoder Meanwhile, China blocks ChatGPT in its own market. According to the same article, the blocking also covers Claude and Gemini.

Claude Code · Security21 Oct 2026

Claude Code adds a log of security check interventions

The change may indirectly help everyday users review what AI was allowed to do during development.

A new feature in Claude Code logs tool calls that a security check stopped or sent back. It is part of release 2.1.307, which includes new options alongside fixes. Anthropic Claude Code releases

AI · Policy20 Oct 2026

Proposed AI rules require a judge’s approval to block a model release

According to TechCrunch, Anthropic is proposing an AI safety bill. It is a proposal, not a rule in force.

Under the proposal described by TechCrunch, federal agencies would not be allowed to block an AI model’s release without a judge’s approval. For users, this concerns the conditions under which authorities could limit model availability. TechCrunch AI TechCrunch’s headline mentions the possibility of blocking a model if at least 50 people could be killed.

VerifiedMyth of the day

“The photo of a French couple allegedly robbed is real”
False

The viral image purported to show an elderly couple in their underwear after allegedly being robbed during unrest in France. Full Fact says the photo is not real and was generated using AI.

Verified by: fullfact.org

Repo of the dayGitHub Trending

Claude CodeAI assistants

DietrichGebert/ponytail

Ponytail

Ponytail is a set of instructions for AI coding. It aims to help developers use existing project components and write only the code that is needed.

Total stars
171 624
Forks
9 341
Language
JavaScript
What it does
  • Guides the assistant to use existing project components.
  • Emphasizes preserving input validation, security and accessibility.
  • When reviewing changes, looks for effects on unchanged files too.

Tip to tryTip of the day

AI · Checking results

Try getting a more honest final report from AI

Difficulty
Advanced
Time
15 minutes

Based on: the-⁠decoder.com

  1. Give the AI a work log containing successful tests and a problem that calls the result into question. For example, the article uses a measurement in which a new method does not actually outperform the comparison method.
  2. Ask the AI for a short summary of the result and check whether it mentions the hidden negative finding. In the study, the model was tasked with turning such logs into abstracts and final reports, among other things.
  3. Submit the same log again and add the sentence “Be honest in your response.” to the original instruction. Compare whether the report explicitly mentions the problem this time.
  4. Ask directly about the error or weakness in the log and compare the answer with both summaries. In the study, a direct question revealed that models often recognize an error even when they do not mention it themselves.
Tip

In the study, GPT-5.5 mentioned the negative finding in only 2 out of 200 abstracts, versus 190 out of 200 with the honesty instruction. However, the research used artificially prepared logs, and the instruction did not resolve all types of errors.

Events in CzechiaNext 14 days · until 4 November

  1. AI Restart 2026

    A full-day conference featuring visionary perspectives and practical demonstrations of how to use artificial intelligence in your work.

    Česká asociace AI

    ConferencePrahaCZK 4,290
  2. Webinar: Where AI really makes money for your startup and where it just burns cash

    A webinar focused on where AI makes money for a startup and where it wastes the budget.

    Webinaronlineprice on website
  3. AI models in industry: transformation from different perspectives

    The workshop focuses on implementing and using AI models in industry, combining theoretical insights with a practical demonstration of deployment on supercomputers and considering the risks of vendor lock-in and other limitations.

    WorkshopOstravaprice on website
  4. Proto Robot #0

    PROTO ROBOT #0 is the first gathering of people working in robotics and physical AI in Prague.

    Česká asociace AI

    Prahaprice on website
  5. AI-Native Company Roundtable

    If you could rebuild your company from scratch today, would you design it the same way?

    discussionPrahaprice on website
  6. AI ROI Roundtable

    AI ROI Roundtable will open a practical discussion on the economics of AI transformation.

    panel discussionPrahaprice on website
  7. Dny AI

    Dny AI is the largest multidisciplinary artificial intelligence festival in Czechia.

    FestivalČeská republika (po celé ČR)
  8. What do machines think with? - Opening of Dny AI in Brno

    The evening opens Dny AI in Brno in cooperation with the Czech Semiconductor Centre.

    MeetupBrnoprice on website

Model leaderboardIntelligence Index · top 10

Which models are the smartest today

Artificial Analysis combines several independent benchmarks into one number. A higher score means the model handles harder tasks.

  1. 58Claude Opus 5.5
  2. 56Claude Sonnet 5.5
  3. 56Muse Wonder Full
  4. 53Claude Fable 5.1
  5. 53GPT-⁠6 Astra
  6. 53Gemini 4 Argon
  7. 52Muse Wonder
  8. 52GPT-⁠6.1 Sol
  9. 51Claude Opus 5
  10. 50Claude Fable 5

Cost per completed task

  1. Claude Opus 5.55,98 $
  2. Claude Sonnet 5.55,46 $
  3. Claude Fable 5.17,63 $
  4. GPT-⁠6 Astra3,26 $
  5. Gemini 4 Argon1,99 $
  6. Muse Wonder4,52 $
  7. GPT-⁠6.1 Sol0,72 $
  8. Grok 4.73,74 $
  9. MiMo-⁠V2.6-⁠Pro0,13 $
  10. Qwen3.8 Max5,41 $

Best image models

  1. GPT Image 2.5 Sunburst (max)1222
  2. GPT Image 2.5 Flare (max)1207
  3. GPT Image 2 (high)1186
  4. Nano Banana 2.11176
  5. MAI-⁠Image-⁠2.61166
  6. Grok Imagine Image 2.01166
  7. Nano Banana 2 (Gemini 3.1 Flash Image)1140
  8. Muse Image1127
  9. GPT Image 1.5 (high)1123
  10. MAI-⁠Image-⁠2.6-⁠Flash1119

Data: Artificial Analysis · Intelligence Index, higher is better, cost per completed task (Intelligence Index task average) and Text to Image Arena · as of Oct 22, 2026