#deep-research — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #deep-research, aggregated by home.social.
-
#MissKittyPolitics #AI #Research
I'm attempting to run the research prompt that is not that complicated of a prompt with no attachments in #Google #Gemini 3.1 Pro #DeepResearch and this is the 5th time that it's told me that I can leave the chat and it will execute the prompt. -
#MissKittyPolitics #AI #Research
I'm attempting to run the research prompt that is not that complicated of a prompt with no attachments in #Google #Gemini 3.1 Pro #DeepResearch and this is the 5th time that it's told me that I can leave the chat and it will execute the prompt. -
Spektrografie #digitalidentity Daniela Beránka: transdisciplinární #EEAT #analyza a sémantická dekonstrukce; by #Gemini #DeepResearch
-
Los nuevos límites de Gemini fallaron y Google tuvo que dar marcha atrás
Apenas una semana después de presentar en el Google I/O 2026 su nuevo sistema de límites de uso basado en poder de cómputo, Google se vio obligado a aplicar parches de urgencia tras una avalancha de quejas: usuarios suscriptores pagos reportaban que un solo prompt de generación de video consumía toda su cuota antes de que el video siquiera terminara de generarse.
https://twitter.com/joshwoodward/status/2060171610922058142
Google tuvo una semana complicada con Gemini. En el Google I/O 2026, la compañía presentó un nuevo sistema de límites de uso basado en poder de cómputo, reemplazando el modelo anterior que contabilizaba cada prompt por igual. La idea sonaba razonable sobre el papel: medir cuánto procesamiento consume cada solicitud en lugar de contar interacciones individuales. Luego los usuarios lo pusieron a prueba, y el rechazo fue inmediato.
Los casos más extremos resultaron difíciles de ignorar. Un suscriptor del plan Google AI Pro publicó un video como prueba mostrando cómo un único intento de generación de avatar consumió toda su cuota de cinco horas antes de que el video terminara de generarse. No fue el único: las primeras reacciones al cambio fueron mayoritariamente negativas, con usuarios cuestionando incluso la forma en que Google comunicó el nuevo sistema, que comparaba el plan Pro contra el nivel gratuito en lugar de explicar qué ofrecía en relación a lo que el propio plan Pro ofrecía antes.
El nuevo esquema, que entró en vigor el 20 de mayo, funciona de la siguiente manera: los límites se renuevan cada cinco horas hasta alcanzar un tope semanal, y el cálculo de consumo toma en cuenta la complejidad del prompt, las funciones utilizadas y la extensión de la conversación. Las tareas que más cuota consumen incluyen generación de imágenes, videos y música, Deep Research, el modelo Pro y las funciones de pensamiento extendido como Deep Think.
La respuesta de Google llegó rápido. Josh Woodward, líder del equipo de Gemini, anunció públicamente que Google está aplicando un tope al máximo de cuota que un solo prompt puede consumir al usar Gemini 3.1 Pro, de modo que una solicitud pesada no pueda agotar el presupuesto completo de una sesión. Además, Google confirmó que en el futuro los usuarios de la app de Gemini podrán comprar créditos de IA de pago por uso para continuar trabajando una vez alcanzado el límite, sin tener que esperar el reinicio de la cuota.
El contexto de fondo es relevante: el cambio llega menos de un mes después de que GitHub reformulara su plan Copilot, también abandonando un modelo de solicitudes fijas para adoptar «AI Credits» basados en tokens realmente utilizados. La tendencia es clara: los planes de tarifa plana para IA generativa están llegando a su límite natural frente a la creciente demanda de funciones computacionalmente intensivas. La pregunta que queda es si los usuarios pagarán más o simplemente utilizarán menos.
#AIUltra #cómputo #DeepResearch #gemini #GeminiPro #google #GoogleAI #GoogleIO2026 #IA #InteligenciaArtificial #JoshWoodward #limitesdeuso #PORTADA #Suscripciones #tecnologia -
Googles-new-deep-research-and-deep-research-max-agents-can-search-the-web-and-your-private-data venturebeat.com/technology/goo… #AI #Google #Gemini #DeepResearch #MCP
-
Googles-new-deep-research-and-deep-research-max-agents-can-search-the-web-and-your-private-data venturebeat.com/technology/goo… #AI #Google #Gemini #DeepResearch #MCP
-
Google, Deep Research ve Deep Research Max'i duyurdu! Yapay zeka ve derin öğrenme alanında çığır açacak bu modeller, teknoloji dünyasında büyük yankı uyandıracak. Araştırma süreçleri yeniden tanımlanıyor.
-
meta #analyza #EEAT analýz #SEOspecialsitaAcopywriterDanielBeranek by #Gemini 3 #DeepResearch
-
🔍 Tired of paying thousands for market research that takes weeks?
Meet Grep — your on-demand team of AI research experts.
Forget hiring consultants. Grep gives you access to 19 PhD-level specialists across 16 categories — from market #research and competitive intelligence to customer vetting — all working simultaneously, all available right now.
#AIResearch #MarketResearch #DeepResearch
🧠 What makes Grep different:
- Deep competitor research without the analyst price tag🧵 👇
-
🔍 Tired of paying thousands for market research that takes weeks?
Meet Grep — your on-demand team of AI research experts.
Forget hiring consultants. Grep gives you access to 19 PhD-level specialists across 16 categories — from market #research and competitive intelligence to customer vetting — all working simultaneously, all available right now.
#AIResearch #MarketResearch #DeepResearch
🧠 What makes Grep different:
- Deep competitor research without the analyst price tag🧵 👇
-
Deep Research, Shallow Agency: What Academic Deep Research Can and Can't Do" | Aaron Tay's Musings about Librarianship
Predefined handcrafted #deepsearch system vs agentic adaptative #deepresearch
#academic #scientific #ai
https://aarontay.substack.com/p/how-agentic-are-academic-deep-research -
Deep Research, Shallow Agency: What Academic Deep Research Can and Can't Do" | Aaron Tay's Musings about Librarianship
Predefined handcrafted #deepsearch system vs agentic adaptative #deepresearch
#academic #scientific #ai
https://aarontay.substack.com/p/how-agentic-are-academic-deep-research -
OpenDeepResearchに学ぶDeepResearchの設計思想と実務適用のポイント
https://qiita.com/jw-automation/items/5af853dc315be0daa7b0?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items -
OpenAI poliert die KI-Recherche: Deep Research in ChatGPT läuft jetzt auf GPT-5.2, liefert Vollbild-Reports mit Inhaltsverzeichnis & Quellen und lässt gezielt Websites priorisieren. Mehr Kontrolle statt „Wischi-waschi“ – aber kein Freibrief für unkritische KI-Outputs. #DeepResearch #KI #ChatGPT | t3n https://t3n.de/news/update-deep-research-chatgpt-1729113/?xing_share=news
-
OpenAI poliert die KI-Recherche: Deep Research in ChatGPT läuft jetzt auf GPT-5.2, liefert Vollbild-Reports mit Inhaltsverzeichnis & Quellen und lässt gezielt Websites priorisieren. Mehr Kontrolle statt „Wischi-waschi“ – aber kein Freibrief für unkritische KI-Outputs. #DeepResearch #KI #ChatGPT | t3n https://t3n.de/news/update-deep-research-chatgpt-1729113/?xing_share=news
-
https://winbuzzer.com/2026/02/11/chatgpt-deep-research-gpt-52-upgrade-xcxwbn/
ChatGPT Deep Research Gets GPT-5.2 Model Upgrade
#AI #ChatGPT #OpenAI #GenAI #AIAssistants #AIResearch #AITools #DeepResearch
-
https://winbuzzer.com/2026/02/11/chatgpt-deep-research-gpt-52-upgrade-xcxwbn/
ChatGPT Deep Research Gets GPT-5.2 Model Upgrade
#AI #ChatGPT #OpenAI #GenAI #AIAssistants #AIResearch #AITools #DeepResearch
-
OpenAI hat Deep Research auf GPT-5.2 migriert. Das Update bringt Site-Specific Search, um Quellen auf verifizierte Domains zu beschränken. Zudem erlauben App Connectors den Zugriff auf Daten aus Drittanbieter-Anwendungen. Nutzer können laufende Suchprozesse nun in Echtzeit stoppen oder korrigieren. Die Ausgabe erfolgt in einer neuen Vollbildansicht. #OpenAI #DeepResearch #GPT52
https://www.all-ai.de/news/news26top/openai-deepresearch-new -
OpenAI hat Deep Research auf GPT-5.2 migriert. Das Update bringt Site-Specific Search, um Quellen auf verifizierte Domains zu beschränken. Zudem erlauben App Connectors den Zugriff auf Daten aus Drittanbieter-Anwendungen. Nutzer können laufende Suchprozesse nun in Echtzeit stoppen oder korrigieren. Die Ausgabe erfolgt in einer neuen Vollbildansicht. #OpenAI #DeepResearch #GPT52
https://www.all-ai.de/news/news26top/openai-deepresearch-new -
OpenAI just upgraded ChatGPT’s Deep Research mode with a built‑in document viewer and a jump‑to‑section table of contents. Now you can scroll through PDFs, slide decks, or web pages and hop straight to the part you need—making AI‑assisted research smoother than ever. Find out how this changes the workflow for developers and scholars alike. #ChatGPT #DeepResearch #AIResearch #DocumentViewer
🔗 https://aidailypost.com/news/chatgpt-deep-research-tool-adds-built-in-viewer-jump-to-section-toc
-
OpenAI just upgraded ChatGPT’s Deep Research mode with a built‑in document viewer and a jump‑to‑section table of contents. Now you can scroll through PDFs, slide decks, or web pages and hop straight to the part you need—making AI‑assisted research smoother than ever. Find out how this changes the workflow for developers and scholars alike. #ChatGPT #DeepResearch #AIResearch #DocumentViewer
🔗 https://aidailypost.com/news/chatgpt-deep-research-tool-adds-built-in-viewer-jump-to-section-toc
-
Deep Research without Deep Pockets.
I pulled apart the premium “Deep Research” tools to see what they actually do, then built the same workflow locally without needing a huge model or huge spend.
The trick: make the pipeline do the hard work (search + reduce + evidence), so the LLM mostly just writes.
Part 5 of the DocSummarizer series: https://www.mostlylucid.net/blog/doomsummarizer-deep-research
What’s your best technique for reducing “model made it up” without just throwing a bigger model at it?
#rag #llm #deepresearch #ai #llm #lucene -
Deep Research without Deep Pockets.
I pulled apart the premium “Deep Research” tools to see what they actually do, then built the same workflow locally without needing a huge model or huge spend.
The trick: make the pipeline do the hard work (search + reduce + evidence), so the LLM mostly just writes.
Part 5 of the DocSummarizer series: https://www.mostlylucid.net/blog/doomsummarizer-deep-research
What’s your best technique for reducing “model made it up” without just throwing a bigger model at it?
#rag #llm #deepresearch #ai #llm #lucene -
🚧 DoomSummarizer (PREVIEW / ALPHA)
I’ve been distilling the lucidRAG principles (hybrid search, entity extraction, knowledge graph + evidence-grounded synthesis) into a console-first, local-first research assistant + personal knowledge base.Think 'big boy's Deep Research running on a laptop with local llms'.
It’s a CLI that can:
Scroll: fetch + rank news/search into a digest / deep-dive / newsletter
Ask: interactive Q&A over your stored evidence
Crawl: index any site (incremental with ETags)
Page: summarise a single URL
Long-form: multi-section articles with grounding + validation
MCP server: expose KB/search/entity graph to agents
Runs fully offline after first model download (no API keys needed for default sources). Cloud LLM/search providers are optional + budget controlled.
End of this post is DoomSummarizer… summarising its own README. 🙂https://github.com/scottgal/lucidrag/releases
doomsummarizer page "https://www.mostlylucid.net" --name doom
doomsummarizer scroll "Tell me about DoomSummarizer" --name doom
-
🚧 DoomSummarizer (PREVIEW / ALPHA)
I’ve been distilling the lucidRAG principles (hybrid search, entity extraction, knowledge graph + evidence-grounded synthesis) into a console-first, local-first research assistant + personal knowledge base.Think 'big boy's Deep Research running on a laptop with local llms'.
It’s a CLI that can:
Scroll: fetch + rank news/search into a digest / deep-dive / newsletter
Ask: interactive Q&A over your stored evidence
Crawl: index any site (incremental with ETags)
Page: summarise a single URL
Long-form: multi-section articles with grounding + validation
MCP server: expose KB/search/entity graph to agents
Runs fully offline after first model download (no API keys needed for default sources). Cloud LLM/search providers are optional + budget controlled.
End of this post is DoomSummarizer… summarising its own README. 🙂https://github.com/scottgal/lucidrag/releases
doomsummarizer page "https://www.mostlylucid.net" --name doom
doomsummarizer scroll "Tell me about DoomSummarizer" --name doom
-
ИИ-агенты: как мы сделали DeepResearch по корпоративным данным и кодовой базе
ИИ‑агенты — очень горячая тема. Кажется, все их делают, но также кажется, что реальную пользу приносит только небольшая часть. Один из основных удачных примеров — DeepResearch, глубокий поиск, отвечающий на сложные вопросы. Многие им пользуются в ChatGPT или Perplexity, но у внешних решений нет доступа к нашим корпоративным данным, поэтому мы сделали свой DeepResearch и сэкономили время сотрудников компании. Меня зовут Сергей Скородумов, я руководитель отдела поисковых сервисов. В статье расскажу про ИИ‑агентов в целом, как мы делали своего, за счёт чего растили его качество и какие главные выводы сделали.
-
Google sprząta w Gemini. Koniec z bałaganem w plikach Deep Research i Canvas
Gigant z Mountain View powoli, ale skutecznie zmienia Gemini z prostego chatbota w kombajn do pracy biurowej.
Najnowsza aktualizacja webowej wersji usługi rozwiązuje jeden z najbardziej irytujących problemów „power userów”: bałagan w wygenerowanych plikach. Sekcja „Moje rzeczy” zyskała właśnie dedykowany widok dla dokumentów.
Do tej pory wszystko, co stworzyliśmy w Gemini – od obrazków, przez kod, po długie raporty – lądowało w jednym worku, wyświetlanym jako siatka zaokrąglonych kafelków. Wyglądało to ładnie, ale przy dłuższych tytułach dokumentów było kompletnie nieczytelne. Google wreszcie to dostrzegł.
Pomóż nam rozwijać iMagazine – ruszyło badanie czytelnictwa 2026
Lista zamiast kafelków
W najnowszej odsłonie, po kliknięciu w panel boczny „Moje Rzeczy”, zobaczymy ponoć (widzą to już Amerykanie, u nas w redakcji mamy jeszcze stary widok) podział na dwie logiczne sekcje:
- Media: tu nadal znajdziemy wygenerowane obrazy i wideo w starym układzie siatki.
- Documents: to nowość stworzona z myślą o użytkownikach funkcji Deep Research oraz Canvas.
Raporty z głębokiego researchu, fragmenty kodu czy teksty są teraz prezentowane w formie przejrzystej listy. Dzięki temu wreszcie widać pełne nazwy plików, a nowe ikony po lewej stronie pozwalają błyskawicznie odróżnić raport badawczy od projektu programistycznego. To mała zmiana UI, która drastycznie poprawia UX – zwłaszcza jeśli używacie Gemini do generowania dziesiątek materiałów tygodniowo.
Na razie tylko w przeglądarce
Zmiany są obecnie widoczne w webowej wersji Gemini. Aplikacje na Androida i iOS wciąż czekają na aktualizację. Warto jednak odnotować inną nowość w świecie mobilnym – w menu konta w aplikacji Gemini pojawił się skrót do NotebookLM.
Co ciekawe, kliknięcie w niego nie otwiera dedykowanej aplikacji (nawet jeśli mamy ją zainstalowaną), a przenosi nas do pełnej, desktopowej wersji strony internetowej narzędzia. Wygląda na to, że Google wciąż szuka sposobu na idealne spięcie swoich rosnących zasobów AI w jeden ekosystem.
#aktualizacjaGemini #Canvas #DeepResearch #GoogleGemini #news #NotebookLM #sztucznaInteligencja #UIUXTrzęsienie ziemi w Cupertino. Apple potwierdza: Gemini od Google fundamentem nowej Siri
-
ZombieAgent atakuje ChatGPT – kolejna luka w systemie AI ujawniona
Czy można naprawić sztuczną inteligencję, która z natury chce każdemu dogodzić? Nowy atak na ChatGPT pokazuje, że kiedy stawiamy wyższą barierkę, ktoś po prostu znajduje dłuższą drabinę.
Czytaj dalej:
https://pressmind.org/zombieagent-atakuje-chatgpt-kolejna-luka-w-systemie-ai-ujawniona/#PressMindLabs #chatgpt #deepresearch #radware #shadowleak #wyciekdanych
-
Hi, folks!
I gave Gemini Deep Research Agent a try. I used it to research the history of knowledge graphs, and it did an amazing job.
I also extracted an interactive semantic knowledge graph from the research text.
You can read the research and explore the knowledge graph at this website:
https://kg-history-research.netlify.app/
Code available here in a Colab notebook:
https://github.com/Ifeanyi55/GephiDatasets/
Please star 🌟 the repository to support its continued maintenance 🙏
-
Hi, folks!
I gave Gemini Deep Research Agent a try. I used it to research the history of knowledge graphs, and it did an amazing job.
I also extracted an interactive semantic knowledge graph from the research text.
You can read the research and explore the knowledge graph at this website:
https://kg-history-research.netlify.app/
Code available here in a Colab notebook:
https://github.com/Ifeanyi55/GephiDatasets/
Please star 🌟 the repository to support its continued maintenance 🙏
-
I asked GPT 5.2 deep research to do an analysis of my intellectual style by exploring a sample of my blog posts
It’s slightly unsettling it has written this in the first person! This is all GPT 5.2 rather than me:
What follows are not positions so much as habits of thought.
1. Take the narrow problem frame, then widen it until the stakes show up
A recurring pattern is to start from the immediate debate, assessment integrity, detection, social media engagement, and then widen the frame to institutional adaptation, labour, incentives, and harms. It is a refusal of the convenience of technical fixes.
2. Treat technologies as structures with powers, not tools with uses
This is the move that underpins most of my work on platforms and now on LLMs. Once you treat the object as structure, you start asking different questions: what does it condition, what does it incentivise, what does it make easy, what does it make costly, and who bears those costs.
3. Look for mechanisms that connect political economy to inner life
Distraction, cognitive triage, communicative escalation, epistemic flooding: these are not just descriptive flourishes. They are the conceptual bridge between macro conditions and what it feels like to think and act in a given environment.
4. Refuse the pro anti binary, without collapsing into fence sitting
I often try to hold ambivalence, because technologies can enable solidarity and flatten hierarchies while also generating toxicity and new burdens. The move is to specify conditions and distributions rather than deliver a verdict.
5. Use time as an analytic dimension, not just a background variable
I keep returning to biographies, trajectories, and the temporal conditions for critique. Acceleration matters because it changes the conditions under which reflection is possible, and it changes what kinds of work can be done well.
6. Convert everyday scenes into theoretical probes
A film, a management practice, a platform feature, a cultural artefact becomes a way of getting at a mechanism. This is how the blog often works: theory is tested against scenes, and scenes become legible through theory.
7. Treat institutions as moral economies, not neutral containers
Universities are not just rule sets. They are systems of responsibility allocation, blame distribution, and risk management. A lot of my writing on AI in higher education is really about this, who gets trusted, who gets suspected, who gets protected, and who is made to absorb uncertainty.
8. Read “inexorability” talk as a political manoeuvre
When someone says “this is inevitable,” I tend to hear a bid to foreclose alternatives and dissolve responsibility. The counter move is to insist on mediation: incentives, infrastructures, and choices.
9. Treat the epistemic environment as an object in its own right
Instead of focusing only on claims and counterclaims, I focus on the conditions under which claims circulate, stick, and become actionable. Epistemic chaos and epistemic flooding are ways of shifting attention from content to environment.
10. Translate abstract theory into a vocabulary that can travel
The blog repeatedly works as a translation machine: taking Archer, Habermas, psychoanalysis, and pulling out handles that can be used in everyday description. If it cannot travel, it cannot do explanatory work outside a small circle.
11. Turn reflexivity itself into an object of inquiry
Rather than treating reflexivity as a given capacity, I treat it as variable, patterned, and historically conditioned, and I keep asking what kinds of environments cultivate what kinds of inner dialogue.
12. Use the blog as a method of serial thinking
There is a recurring practice of working through a book, a concept, or a problem in public, chapter by chapter or post by post, because writing clarifies. This is a methodological commitment, not just a communication choice.
13. Keep returning to dependence
From early writing on the self as an object of control and dependence, through social media as moral psychology, to LLMs as embedded in lifeworld activity, the recurring question is what we become dependent on, what dependency does to agency, and how dependencies are organised.
14. Treat degradation as a process, not an event
This is why enshittification matters as a concept, and why I care about how concepts themselves get degraded by platform circulation. The analytic focus stays on how, not merely that.
15. Look for feedback loops
Generative systems feed platforms with more content, which intensifies attention competition, which worsens epistemic conditions, which increases the appeal of tools that promise relief, which further embeds those tools. A lot of my recent writing is trying to name these loops early, before they harden into “common sense.”