Skip to content
Vol. IIssue 15

Educational AI Daily

Artificial intelligence, as told by three teaching communities

Back to the issue

In the print dialog, choose “Save as PDF” to keep it as a file.

https://diarioiaeducativa.github.io/en/2026-11-15/declarar-uso-ia-transparencia.html

Debates 16 November

Academic integrity

Declaring AI use: a teacher asks schools to clarify which uses are acceptable

A teacher asks whether schools, very focused on detecting AI-generated work, explain clearly enough which uses are acceptable and how to declare them.

The usual debate in schools is how to detect work done with AI. In ChatGPT-IA-edu, a teacher proposed shifting it towards transparency of the process, based on “Verdad a medias: ¿hay que declarar cómo hemos utilizado la IA?” (“A half-truth: do we have to declare how we have used AI?”), an instalment of his series “Familias con inteligencia”. Its author is Chema Falcó, a secondary and sixth-form teacher; the paper has already reported on other instalments of the series, on homework done with AI and on the percentage given by a detector.

The instalment starts from two first-year Bachillerato students, Laura and Pablo, on a piece of work in which AI may be used if it is declared. Laura used it for ideas and to check requirements, but her report presents the work as individual; Pablo declared minimal use and later admitted that he had relied on it extensively. According to the author, the cases serve to work with students on academic integrity, authorship, responsibility and the traceability of learning, and he argues that declaring does not detract from merit but makes authorship verifiable. He proposes a three-sentence template (“I used AI to…”, “Afterwards I verified…”, “The research and the writing were mine”) and points to the INTEF guide, which includes a model declaration.

In the group, the author added the underlying question for schools and teachers: whether it is being explained clearly enough which uses are acceptable, which are not and how they should be declared. He is concerned, he said, that students who want to do the right thing perceive that declaring may count against them. The message ended by asking whether a declaration of AI use is required and of what kind; there is no record of the group replying that day.

The same idea of declaring permitted uses and asking for evidence of the process is the one put forward in Sunday’s feature on AI detectors.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-15/modern-web-guidance.html

Vibe coding 16 November

Programming with agents

Chrome publishes Modern Web Guidance 1.0, a guide to the current web for coding agents

The guide brings together expert-verified solutions and compatibility data to correct the outdated knowledge of models.

In Vibe Coding Educativo it was mentioned that models that generate code may be trained on outdated web development and infrastructure data. As a possible remedy, Modern Web Guidance 1.0 was shared, published on 10 November by Chrome for Developers.

According to the article, it includes an expanded base of guides with cross-browser solutions, each reviewed by experts, and compatibility data based on Baseline. It adds a modernisation workflow that analyses existing code (HTML, CSS, JavaScript and package.json) in search of safe improvements; in this first version, with static analysis of a limited number of patterns. It also offers a public evaluation dashboard that compares results with and without the guide. It is installed with `npx modern-web-guidance@latest install`, or from the agent plugin marketplaces such as Claude Code, Codex or Cursor. Firefox and Edge join Chrome in supporting the project.

In the group, one person passed the article to an AI assistant, which recommended consulting the documents but not installing the program in case it sends data to Google. The article does not mention what data it does or does not send, so that warning could not be confirmed. Another person said that their assistant had suggested the same thing and an alternative: downloading the documentation as .md files and adding it to the project folder or as an agent skill.

Read the original conversation in Vibe Coding Educativo

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in Vibe Coding Educativo

Checked in

https://diarioiaeducativa.github.io/en/2026-11-15/enfonsar-la-flota.html

Vibe coding 16 November

New in the repository

Battleships in Catalan, against the machine or online with a room code

“Enfonsar la flota” is a web version of the classic battleships game, with a 10×10 board, which works on coordinates and strategy in primary and secondary school.

“Enfonsar la flota” is a Catalan version of the well-known battleships game, created by Felip Sarroca i Gil and shared in the Vibe Coding Educativo resource repository. The interface, in Catalan, opens directly in the browser, with no sign-up screen, and can be used on a computer or a phone.

The board is 10×10, with rows from A to J and columns from 1 to 10. Each player places five ships, horizontally or vertically, and turns alternate with shots reported as water, hit or sunk. The first person to sink the opponent's five ships wins. A “Regles” button explains how the game works within the page itself.

The start screen offers two modes. In “Partida en línia” a room is created and its code or a QR code is shared so that another person can join; a code received from someone else can also be entered. In “Contra la màquina” there are three difficulty levels, called Grumet (easy), Capità (medium) and Almirall (hard). According to the entry and its author's presentation, the mode against the machine also works offline, and the page offers a button to install it as an app on the device.

For primary and secondary teachers, the game can be useful in logical-mathematical thinking: locating squares with a letter and a number, reasoning from what is known about the board and planning a strategy. The online mode lets students play in pairs from different devices, while the mode against the machine lets each student practise on their own. Because it is in Catalan, it also fits work on the game's vocabulary in that language.

The page states that the code is published under the AGPL v3 licence and that the contents are CC BY-NC-SA 4.0. The buttons and texts shown are in Catalan only; the entry does not mention other languages.

See the Vibe Coding Educativo resource repository

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

See the Vibe Coding Educativo resource repository

Checked in

https://diarioiaeducativa.github.io/en/2026-11-15/gemini4-hydrogen-orchestrator.html

Institutions & Industry 16 November

Google Antigravity

Antigravity adds Gemini 4 Hydrogen and an Orchestrator mode that shares out the work

Orchestrator combines Hydrogen with more powerful models; access depends on Google’s paid plan.

A member of Vibe Coding Educativo said they still had to try the Orchestrator mode of Google Antigravity, announced on 12 November together with Gemini 4 Hydrogen.

According to Google, Hydrogen is for everyday programming and agent tasks, with four reasoning levels (low, medium, high by default and a new “xhigh”). Orchestrator is an option in the model drop-down that combines it with more powerful ones, such as Gemini 4 Carbon: Carbon explores and plans, Hydrogen edits and tests, and Carbon takes over again if Hydrogen gets stuck. These are the company’s figures: in its Terminal-Bench 4.1 test, Orchestrator comes close to Carbon’s success rate with 24% less quota and 19% less time on average.

Hydrogen is rolling out gradually to the paid Google AI Pro and Ultra subscriptions. Orchestrator with Carbon or with Argon is for Google AI Ultra, and Orchestrator with Argon also for paid Google AI Pro. Businesses can access both with Gemini for business. The paper reported on 29 October that Carbon was presented only for cybersecurity specialists; the Antigravity page does not explain how this has been widened, so it is worth checking in the tool itself what each plan offers.

Read the original conversation in Vibe Coding Educativo

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in Vibe Coding Educativo

Checked in

https://diarioiaeducativa.github.io/en/2026-11-15/seedream-55.html

Multimedia 16 November

AI imaging

Seedream 5.5 promises to keep an image stable across 100 rounds of editing

ByteDance presents a version of its image generator designed for retouching the same image again and again without changing what is left untouched, and which generates transparent backgrounds.

ByteDance Seed presented Seedream 5.5 on 9 November, which was shared in EduMedia-IAG. According to the company, it preserves unedited areas at pixel level and reduces colour shifts and loss of detail, allowing work to continue on the same image for up to 100 rounds of editing; its demonstration chains 100 edits on the same image.

The second novelty is native image generation with an alpha channel, with full and partial transparency, so there is no need to remove the background separately. It may be useful for preparing characters and elements for classroom materials. ByteDance adds improvements in light, skin and materials, and in text within the image; these are figures from its own evaluation, with no details of sample or method.

The page links to access via Dola and the BytePlus Playground, but gives no rollout schedule or says in which regions it will be available, so its availability in Europe is not confirmed. It also announces a Pro version of the series, with no date.

Read the original conversation in EduMedia-IAG

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in EduMedia-IAG

Checked in

https://diarioiaeducativa.github.io/en/2026-11-15/contrastar-varias-ia-model-council.html

A year ago 16 November 2025

Checking what AI says

Cross-checking several AIs was a homemade trick; today Perplexity does it in a single query

In November 2025, the advice for trusting an answer was to ask two or three artificial intelligences and use another to look for the discrepancies. In February 2026, Perplexity turned that method into a feature.

On 16 November 2025, in ChatGPT-IA-edu, people were discussing an article about the risks of taking what AI says at face value without a critical eye. The reply was a method: when accuracy matters, make the same request to two or three different models, hand the answers to another to point out where they disagree, and then review the whole process in person. It was also suggested asking the model to explain what it said, especially if it did not match expectations: that is where you see whether it slips.

On 5 February 2026, Perplexity introduced Model Council: a single query goes at once to three models, such as Claude Opus 4.6, GPT-5.2 or Gemini 3.0, according to its announcement, and another model gathers the answers, resolves conflicts when it can and points out where they agree and where they do not. It launched on the web only and for subscribers to its Max plan.

What once meant opening several tabs and copying and pasting now fits into a single question. The last part of the method, however, is not replaced by any feature: reviewing the result is still the teacher’s job.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

Chalk and Algorithm's comic strip 16 November

In the staff room, drafting the AI policy.We need to clarifywhich uses of AIare acceptable andhow to declarethem.I'll draft thepolicy in asecond!
Done. Point three: everyuse of AI must bedeclared, no exceptions.
Then let's start with thepolicy: “I used AI towrite this policy…”

Based on Declaring AI use: a teacher asks schools to clarify which uses are acceptable.

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.