The AI Times
EN
Volume I · Daily editionDaily roundup of AI newsMCPMD
EN
THE AI TIMES
Gemini · Character.ai · Privacy

New capabilities, uncertain services and the price of our data

Gemini is starting to roll out custom skills and bringing the Argon model to paid users. Meanwhile, Character.ai is looking for a buyer and may shut down without one. Today’s roundup also covers AI for mental health issues, app discovery and shopping. MIT Technology Review raises the question of where work archives go when a company closes.

Gemini · Features13 October 2026

Gemini starts getting custom skills. Workspace waits

Saved skills for recurring tasks are starting to roll out in the Gemini app. The Workspace launch is delayed.

Skills rollout · date and statusGoogle Workspace Updates · official source
Skills in the Gemini app13 October 2026Mid-NovemberRollout beginsExpected completionSkills in WorkspaceDelayed until further noticeOnly users over 18

Google began rolling out skills in the Gemini app on 13 October 2026 and expects to finish in mid-November. They are intended for both Google Workspace customers and people with personal Google accounts. Skills directly in Workspace, however, are delayed until further notice; there is no new date yet. In the coming weeks, Google plans to add sharing of skills from the Gemini app and integration with Google Drive. The original Workspace plan called for completing the rapid rollout on 12 October and starting the scheduled rollout on 19 October 2026. In Workspace, skills will be available only to users over 18, while in the Gemini app they will be available to people of all ages. There are no admin controls for skills in the Gemini app. Google Workspace Updates

Character.ai · Availability13 October 2026

Character.ai seeks a buyer. Without one, the service will shut down

Chats are working and subscriptions are still being billed. But the company has entered a court-⁠supervised sale.

Character.ai has voluntarily filed for bankruptcy under US Chapter 11 to sell the company under court supervision. All services continue for users for now, and data protection rules remain unchanged. If no viable offer emerges that allows operations to continue, the platform will gradually shut down. Character.AI

Gemini · Models13 October 2026

Gemini 4 Argon arrives in the app for paying users

The new model is starting to become more widely available, but a paid account is still required.

Gemini 4 Argon is starting to roll out in the Gemini app and Antigravity. According to Ars Technica, it is currently intended for paying users, and even lower paid tiers lack full access. For the average user, availability within their subscription matters more than the announcement of a new model itself. Google Gemini Blog · Ars Technica AI · 9to5Google · 9to5Google

AI · Health13 October 2026

Czechia prepares guidance on AI and mental health issues

The Ministry of Health wants to address the safe use of AI for mental health problems.

According to Lupa, the ministry announced that it is preparing guidance on the safe use of AI for mental health issues. It has also included the topic in the national action plan. The initiative thus targets people who already discuss sensitive personal problems with a chatbot. Spring research cited by Lupa says that more than half of people aged 18 to 34 had used AI for mental health in the previous 30 days. Nearly a third of users had discussed a mental health crisis or acute stress with AI. A similar share had used it at least occasionally instead of professional help. Lupa.cz

SOURCElupa.cz
Google Play · Features13 October 2026

Google Play expands AI app search

Ask Play is set to recommend and compare apps; private sharing of apps built with AI is also planned.

According to Ars Technica, Ask Play is expanding from questions about a specific app to general searches for apps and games. Users ask a question and receive recommendations that may include detailed comparison tables. The store thus adds another way to search by need rather than just by name. Google Play is also expected to offer private hosting and sharing of custom apps built with AI soon. Ars Technica AI

Facebook · Recommendations13 October 2026

Facebook will let users adjust topics recommended by AI

Your Algorithm gives users a direct choice of topics for the main feed and Reels.

According to The Verge, Facebook will let people add and manage topics they want to see in recommendations. Your Algorithm uses “interested” and “not interested” options. The outlet reports that Meta promises recommendations will respond to changes immediately. Interest and disinterest buttons may also appear directly below posts. Your Algorithm came to Instagram earlier this year, and Meta introduced a similar feature, Dear Algo, on Threads. The Verge AI

Higgsfield · Video13 October 2026

Higgsfield extends AI video beyond ten minutes

OpenAI describes a shift in video creation from one minute to more than ten.

According to an OpenAI post, Higgsfield is bringing creators longer videos with the GPT-6 Astra model. The source headline describes a shift from 1 minute to more than 10 minutes. This is an update for people creating video with AI. The post mentions GPT-Image-2.5 for precise image processing. It describes a workflow with GPT-Live for making films by voice. OpenAI News

What experts say

Simon Willison

developer, creator of LLM and Datasette

Redesigning source control at Anthropic

Willison describes why Anthropic replaced source control with its own system: the volume of changes made with AI rose sharply. He also highlights the limits of reviews in which one model looks for another’s mistakes. Experienced people still need to prevent AI from needlessly complicating solutions over hypothetical concerns.

simonwillison.net
What experts say

Jack Clark

Anthropic co-founder, author of Import AI

Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space

Clark examines a test of whether AI understands what an image merely implies. The results show a substantial gap behind humans, but Clark also sees capabilities that some considered beyond conventional language models. A common error is overlooking a decisive detail or misidentifying an object.

Disclosureco-founded Anthropic and also writes about its competitors

importai.substack.com
AI · Privacy13 October 2026

Work emails from closing companies head to AI labs

Companies shutting down are selling work archives for AI training. Employees generally receive nothing from the sales.

Work archive sales · share and countMIT Technology Review · reporting
SimpleClosure · work data sales47% of dealsInclude code or data salesMore than 150 dealsBrokered by Asset HubEmails · chats · documents · code

MIT Technology Review describes a growing trade in code, emails, chats and documents from startups that are shutting down. For employees, the key point is that records of everyday work can also be sold. SimpleClosure added data sales to its services in April, and its Asset Hub has already brokered more than 150 deals. Code or data sales now feature in 47% of its deals. Proceeds typically go to preferred creditors and investors rather than employees. MIT Technology Review AI

Shopify · Shopping13 October 2026

Shopify will open shopping to more AI assistants

Customers are set to be able to entrust purchases to assistants without a special partner integration.

According to The Verge, Shopify is expanding access for AI assistants to its catalogue and to purchases on users’ behalf. Such an assistant, sometimes called an agent, can find products and act for a person. Customers are set to be able to configure their own shopping assistants. The Verge AI

NVIDIA · Gaming13 October 2026

DLSS 5 tones down facial changes ahead of its debut

New previews suggest that AI image adjustments in Hogwarts Legacy will be less pronounced.

According to The Verge, DLSS 5 is set to debut in Hogwarts Legacy next month in a toned-down form. The technology uses AI to alter game visuals. What matters to players is how it changes characters’ appearance, which drew criticism in the first previews. The outlet reports that NVIDIA has released new comparison images showing subtler facial changes. At the original reveal, the author criticised how a fifteen-year-old Hogwarts Legacy character suddenly looked much older. He compares the new changes to the subtle adjustments in NBA 2K27. The Verge AI

Apple TV · Siri AI13 October 2026

Apple unveils a new Apple TV 4K with Siri AI

Apple’s AI is also coming to devices for watching television at home.

Apple announced the next generation of Apple TV 4K, linking it to Siri AI in the announcement’s title. The announcement was published on 13 October 2026. Apple Newsroom In its hands-on report, Engadget describes new audio features and Siri as a viewing companion. Engadget

SOURCEapple.com

VerifiedMyth of the day

“An alleged timestamped surveillance camera image from James Talarico’s house is circulating on X.”
False

Lead Stories identifies the image as AI-generated rather than real surveillance footage. Its editors used verification tools that indicated an artificial origin and found no matching reports from established media in news search engines.

Verified by: leadstories.com

Repo of the dayGitHub Trending

MCPClaude CodeReverse engineering

morluto/rea

REA: Reverse Engineer Anything

A single MCP interface through which an AI agent (Claude Code, Codex, Cursor or Gemini CLI) can explore an application without source code: native binaries, JavaScript and Electron applications, .NET and websites. It adds evidence and limitations to each conclusion, and the analysis runs on your own computer.

Total stars
68 211
Forks
14 360
Language
TypeScript
What it does
  • Explains how a feature in another application works, shows evidence and builds a similar one in your project.
  • Connects to Ghidra, Hopper or IDA to analyse native code.
  • Analyses JavaScript and Electron without an additional tool; the same features are also available in the terminal.
Get started
npx rea-agents setup

Repository on GitHub →

Tip to tryTip of the day

Midjourney

How to get Midjourney to fix an image with Thinking mode

Difficulty
Beginner
Time
5 minutes

Based on: updates.midjourney.com

  1. Open the Midjourney test website alpha.midjourney.com.
  2. Open the image you want to improve in the preview (lightbox).
  3. Click “Rerun (Thinking)” at the bottom right. Midjourney will regenerate the image with “thinking”.
Tip

According to Midjourney’s tests, the mode mainly helps with prompt accuracy, text in images and scene consistency. The mode is still being tested, and Midjourney may offer it to everyone later.

Events in CzechiaNext 14 days · until 27 October

  1. ConferencePrahaprice on website
  2. AI HotSpot

    AI HotSpot 2026 – an overview and networking for AI-⁠led meetings.

    ConferencePrahaprice on website
  3. Rakathon 2026

    Help change the future of cancer care with AI!

    Česká asociace AI

    Prahaprice on website
  4. AI in culture: Why and at what cost?

    A conference on the practical use of artificial intelligence in museums and cultural institutions.

    ConferencePrahaprice on website
  5. AI Monday Praha #16

    Join an open community of people who actually use AI.

    MeetupPrahafree
  6. Experience sharing for CMOs: AI in B2B marketing – how to fill your pipeline with quality leads effectively

    MeetingBrnofree
  7. Automation & AI Clinic

    An event for companies that want to discuss a specific automation, digitalisation or AI problem directly with specialists in small expert zones.

    WorkshopPrahafree
  8. AI Restart 2026

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

    Česká asociace AI

    ConferencePrahaCZK 4,290

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. Muse Spark 1.31,60 $
  9. Grok 4.73,74 $
  10. MiMo-⁠V2.6-⁠Pro0,13 $

Best image models

  1. GPT Image 2.5 Sunburst (max)1199
  2. GPT Image 2.5 Flare (max)1192
  3. GPT Image 2 (high)1172
  4. Nano Banana 2.11159
  5. Grok Imagine Image 2.01155
  6. MAI-⁠Image-⁠2.61152
  7. Nano Banana 2 (Gemini 3.1 Flash Image)1125
  8. Muse Image1115
  9. FLUX 3 Image1115
  10. GPT Image 1.5 (high)1109

Data: Artificial Analysis · Intelligence Index, higher is better, cost per completed task (Intelligence Index task average) and Text to Image Arena · as of Oct 14, 2026