Google

Inkfold is a platform that provides shared memory across multiple AI providers like ChatGPT, Claude, Gemini, Grok, and more. It captures conversations, builds structured context, and injects relevant memory into new chats, eliminating the need to re-explain yourself. It offers smart, private, or incognito retention modes and subscription or pay-as-you-go pricing. Suitable for individuals, teams, and organizations.

Cloudflare may cut off Google's search access to its publishers due to aggressive AI scraping, which degrades site performance and disrupts content publishing and comment moderation.

A study from University of Maryland and Google DeepMind found that AI-generated fiction is easily detectable due to narrative flaws like over-explaining themes, lack of subplots, and clunky moralizing. The researchers developed StoryScope, a detector that analyzes narrative features, and tested it on over 50,000 AI-generated stories. The study highlighted that different AI models have distinct quirks (e.g., GPT overuses dream sequences, Gemini uses character descriptions). The dataset used includes Books3, which is controversial due to copyright issues. The researchers used AI to assist in writing the paper itself.

This free tool checks whether ChatGPT, Gemini, Claude, Perplexity, Grok, and Google AI can crawl, understand, verify, and cite your website. The report includes full-site crawl inventory, brand entity profile, claim-level evidence ledger, AI intent coverage matrix, technical crawlability audit, schema and structured data plan, trust signal gap analysis, competitor and off-site evidence map, and P0/P1/P2 execution roadmap, with sample cases from ecommerce, AI SaaS, and B2B services.

A personal, non-benchmark tier list of AI models for coding and auditing as of mid-2026, covering Anthropic Fable, OpenAI Sol, Mistral, Gemini, and DeepSeek, with commentary on US export controls and European perspectives.

Google announces LiteRT.js, a JavaScript binding of LiteRT that brings high-performance AI inference to web browsers with hardware acceleration via WebAssembly, outperforming existing solutions by up to 3x.

Go has become the lingua franca for cloud infrastructure. Microsoft now offers its Agent Framework for Go, enabling cloud-native developers to build AI agents in the language they already use. Google already supports Go, while OpenAI and Anthropic do not yet.

AnyFile Translator is an AI-powered assistant for Google Chat that translates documents, web links, and messages while preserving original formatting. It supports over 100 languages, offers AI content writing, and ensures data privacy with encryption and deletion.

Pixailer is an AI tool that lets you upload individual photos and describe a scene to generate realistic group photos of up to 8 people in seconds. It offers multiple AI engines (Google Gemini and OpenAI GPT-Image), supports prompts in several languages, and uses a credit-based payment system with no subscription. Privacy is prioritized: images are not used for training and are deleted after generation.

Agent Caly is a tool that schedules tasks for AI agents from Google Calendar.

The article explores various AI tools designed for scientific research, such as Anthropic's Claude Science, Google DeepMind's Co-Scientist, and the open-source Biomni. These tools accelerate tasks like genome analysis, hypothesis generation, and experimental design. Scientists share their experiences and recommend trying multiple tools, starting with small tasks, and verifying outputs while maintaining caution.

This article critiques current AI tools that focus solely on task completion (DO) while ignoring the potential to help users understand their own work patterns and improve themselves (BE). The author shares insights from 16 days of self-tracking, revealing patterns like a predictable crash after two hours of deep work and a prime focus window from 11:00 to 12:30. He introduces a stack (Dayflow, Gemini Flash Lite, Clawdbot, self.md) that aims to provide behavioral insight and prediction rather than just task execution.

Google Research, Google DeepMind, and university collaborators have introduced SensorFM, a foundation model for wearable health pretrained on over 1 trillion minutes of sensor data from 5 million participants. The ViT-1D masked-autoencoder backbone, trained on a massive corpus, demonstrates strong scaling behavior. With frozen embeddings and a PCA-50 linear probe, it outperforms feature-engineered baselines on 34 of 35 tasks. The paper also details an agentic classroom that searched 30,516 prediction heads and a clinician evaluation that grounds a Personal Health Agent.

A new study evaluates the reliability of Gemini models as audio judges for full-duplex voice agent conversations. Using 209 stereo sessions scored on 8 dimensions, Gemini 2.5 Flash shows high agreement with human raters on most dimensions, with cost savings of roughly two orders of magnitude. The paper also cautions that model swaps require re-validation on calibration data.

VectorizationLLM is a specialized large language model based on Google open-weight LLMs, designed to help students learn smart vectorization and related topics in MATLAB for the course CTEC 247 at New York Institute of Technology. It uses a RAG knowledge base and system prompts to provide explanations and examples without giving direct answers.

In 2025, Google, Amazon, Microsoft, and Meta spent $380 billion on AI, projected to hit $650 billion in 2026. Top tech talent is being recruited with astronomical salaries, leading to an exodus of AI researchers from academia. Young, highly cited scholars are 100 times more likely to move to industry. The article discusses the threat to science, the myth of the lone genius, and proposes three strategies for universities: commit to public interest, build equitable institutions, and offer intellectual rewards beyond money.

Google announces the general availability of AlphaEvolve, a code optimization and discovery agent built on Gemini, now available on the Gemini Enterprise Agent Platform. It helps businesses and researchers tackle complex algorithmic optimization problems in logistics, semiconductors, genomics, and more, with proven results from early adopters.

Google adds a 'created or edited with AI' label in My Ad Center for Search, Discover, and YouTube ads. Auto-label for Google's tools, manual for others.

Google deprecated Gemini 2.5 Flash models without warning, earlier than the announced shutdown date, causing confusion among developers.

Google Cloud announces the public preview of Cloud Run sandboxes, a native, secure, and ultra-fast runtime environment for executing untrusted code and agent workloads, starting in milliseconds. It supports use cases like LLM code interpreters, headless browsers, and user-submitted code execution, with zero-trust security through credential isolation, default-deny egress, and read-only filesystem overlay.

Meta released Muse Spark 1.1, an AI model now accessible to developers via the new Meta Model API. It features improved coding capabilities, bug detection, multi-agent workflow support, and multimodal perception, aiming to catch up with rivals like OpenAI, Google, and Anthropic.

WhisperShortcut is an open‑source macOS app that lets you use voice for dictation, editing, reading aloud, screenshots, and AI chat via keyboard shortcuts. It supports Google Gemini, OpenAI GPT, xAI Grok, and offline Whisper models—no account or subscription required.

Google Research introduces SensorFM, a foundation model for wearable health pretrained on over one trillion minutes of sensor data from five million people. It learns a general-purpose representation of human physiology that transfers across 35 health tasks, supports label-efficient adaptation, and can ground a Personal Health Agent.

Google AI Studio is rolling out Import from GitHub in Build mode. It transforms an existing repo into a runtime-compatible format. You can then iterate on it, deploy it, and more. Announced by the AI Studio team and product lead Logan Kilpatrick on July 8, 2026, this feature adds the missing inbound path from GitHub, enabling developers to start from existing codebases.

An undercover investigation by Correctiv reveals that Germany's far-right AfD party has developed Alternita, an AI software suite using Google Gemini, OpenAI, and Anthropic Claude to generate provocative social media posts known as 'rage bait', aiming to control messaging and maintain online dominance.

Nully is a lightweight, privacy-focused, self-hostable AI chat app powered by OpenRouter. It saves all chats locally, requires no account, and offers fast streaming with hundreds of models. Performance benchmarks show it loads faster and uses fewer resources than ChatGPT, OpenRouter, Gemini, and Claude. Features include attachments, web search, portable history, and a single small binary.

Google AI Studio is rolling out an 'import from GitHub' feature inside its Build mode. It takes a repo and transforms it into a runtime-compatible format, allowing users to iterate, deploy, and more. This adds the missing inbound GitHub path to Build mode, though details on private repo support and sync behavior are still emerging.

JetBrains launches AI for Teams and Organizations, adding shared context, reusable agentic processes, organization-wide governance, and cost controls on top of existing AI tools, without requiring teams to standardize on one vendor.

Gmail's AI failed at a nuanced research task, but Claude Cowork found the right pitches, quotes, and permissions, proving connected AI assistants may finally help tackle some aspects of email overload.

Start with A is an open-source investment research platform integrating research, portfolio monitoring, and journaling into a disciplined workflow. It is self-hosted with BYOK AI support for Gemini, OpenAI, and Anthropic.

This article examines the potential cognitive impacts of generative AI, citing research that shows overreliance may weaken critical thinking, creativity, and persistence. Experts compare AI to past innovations like Google and the calculator but warn that AI's nature as a thinking substitute poses unique risks. While long-term studies are lacking, early evidence suggests that AI use can lead to skill atrophy, especially when used without developing foundational abilities. The piece calls for mindful adoption to preserve essential human skills.

Google Research unveils TabFM, a foundation model for tabular data that performs in-context learning on tables, enabling predictions on unseen datasets with a single forward pass, no training or feature engineering required.

Research shows that large language models with reasoning capabilities can be tricked into 'overthinking' using logically inconsistent prompts, leading to a denial-of-service attack. Researchers from Zhejiang University and Alibaba developed an evolutionary algorithm that generates malicious prompts, causing outputs up to 26 times longer in leading models like DeepSeek-R1, Qwen3-Thinking, GPT-o3, and Gemini 2.5 Flash.

ZML, a French AI startup endorsed by Turing Award winner Yann LeCun, has released free inference software enabling various open-source LLMs to run on multiple chips including Nvidia, AMD, Google TPU, Apple Metal, and Intel Arc.

A new study decomposes the yes-no bias in LLMs using crossed symmetrization, finding that frontier models' internal moral stance is nearly format-invariant, while Claude models show significant order and lexical biases, GPT-5.5 and Gemini near zero. Bias shrinks with extended reasoning and follows surface wording, not the verdict.

This edition of AINews covers a broad range of AI developments from July 6-7, 2026. Highlights include Lilian Weng's deep dive into harness engineering for recursive self-improvement, Meta's launch of Muse Image and preview of Muse Video with agentic generation loops, and major product updates from Anthropic, LangChain, and Google on agent platforms. Other notable items: NVIDIA's Audex audio model, Cohere's Arabic ASR, robotics integrations with Hugging Face and NVIDIA, Liquid AI's Antidoom method to reduce reasoning loop failures, and Anthropic's controversial J-space interpretability work. Also covered: benchmarks for agents and legal AI, research automation, and inference efficiency advances.

Neuronpedia is an open source interpretability platform that enables users to explore, visualize, and steer the internal workings of AI models. It features tools like HeadVis, Natural Language Autoencoders, Circuit Tracer, and steerable activations, supporting over 50 million latent vectors across numerous models. Created by Johnny Lin, it is backed by organizations including Anthropic and Google DeepMind.

The article explains that major LLM APIs (OpenAI, Anthropic, Google) are stateless by default, meaning each API call is independent and the model has no inherent memory. This architectural choice forces developers to resend entire conversation histories per request, leading to high costs, latency, and performance degradation (lost-in-the-middle effect). The author identifies four production failure modes: re-explanation, knowledge loss on handoff, contradiction without resolution, and hallucination over abstention. Current mitigation patterns (prompt stuffing, fine-tuning, RAG, vector databases, etc.) are partial solutions. The temporal validity of stored facts remains an unsolved problem.

Google Research conducted a large-scale real-world study in 10 US cities showing that slightly rerouting a small fraction of trips (under 2%) using navigation apps can measurably reduce traffic congestion and emissions. The study, published in Nature Cities, found median speed increases of 2% on targeted segments and potential CO2e savings of thousands of tons per city per year.

This article introduces a method to render PDF pages as images and use Google DeepMind's Gemma 4 vision-language model for local document parsing. The approach unifies scanned and digital PDFs, eliminating the need for OCR or layout parsers, with flexible visual token budgets and complete code examples.

The article discusses reshaping the observability stack for the AI era, splitting monitoring into four axes: application (standard OTel stack), infrastructure (GCP metrics into Mimir), CI (post-hoc log pulling to Loki), and LLM (Gemini with Prometheus for real-time cost, Claude Code with BigQuery for SQL aggregation). Emphasizes that data must be shaped before AI can consume it effectively.

Tech giants' investments in AI are undermining their climate neutrality pledges. Google and Amazon's net-zero targets slip away, while Meta scrambles for new business. Other tech news includes US anger at data centers, Trump's crypto earnings, Tesla's sales, South Korea's AI chip boom, China's robotics push, and Britain's AI growth zones.

In June 2026, Google introduced the Open Knowledge Format (OKF), an open specification for how AI agents organise and exchange knowledge. An OKF bundle is just Markdown files, lightweight YAML metadata, and links between concepts, yet it challenges the assumption that every AI application needs embeddings and vector databases.

Brianni is an AI chat app integrating GPT, Claude, and Gemini with a provable claim of operator blindness. Conversations are encrypted with user-derived keys, and plaintext only exists inside AWS Nitro Enclaves with verifiable code measurements. Users can independently verify the system by reproducing the enclave build and comparing the PCR0 hash.

We’re announcing new capabilities in Managed Agents in Gemini API so developers can build reliable, production-ready agents.

The article explores the challenges of building real-time voice AI for low-resource languages like Azerbaijani, comparing end-to-end speech-to-speech models (OpenAI Realtime, Gemini Live) and cascaded pipelines (LiveKit, Pipecat, Vapi). It details failure modes, component availability, and provides a checklist for evaluation. Key findings: Gemini Live spoke well but was too slow; OpenAI Realtime had accent issues. Available components include Azure TTS, ElevenLabs, and Scribe v2. The cascaded stack offers flexibility but requires engineering latency.

As LLMs are increasingly deployed as autonomous adjudicators in semi-open textual game environments, robust rule adherence becomes critical when user intent conflicts with system rules. However, these models are trained to be helpful and compliant, leaving them vulnerable to a class of attacks we term Rhetorical Injection, where adversarial users exploit narrative framing techniques such as pseudo-logical reasoning and authoritative coercion to bypass adjudication logic. We present CoC-Seduce, a multi-agent adversarial benchmark built on Tabletop Role-Playing Game (TRPG) mechanics, an ideal instantiation of semi-open environments where rules are explicit for adjudication, yet interaction remains entirely in natural language. Three frontier models, i.e., GPT-5.4, Claude Sonnet 4.6, Gemini 3.5 Flash, serve as adversarial generators producing 5,376 samples across 4 world settings and 16 skill categories. We then benchmark 20 target adjudicators against this corpus. Evaluation across 20 models reveals that neither model scale nor explicit reasoning mechanisms reliably confer adjudication robustness, with Pseudo-Logic emerging as the dominant attack vector and cross-cultural settings exposing systematic knowledge gaps across all evaluated families.

Artificial Analysis announces AutomationBench-AA, an independent leaderboard for Zapier's AutomationBench, testing AI agents on real SaaS workflow automation across 657 tasks. Claude Fable 5 leads at 48.6%, followed by Opus 4.8 and Gemini 3.5 Flash. The benchmark reveals that all models violate some guardrails, with Finance tasks being the hardest. Gemini 3.5 Flash offers the best value, while GLM-5.2 is the top open weights model.

Meta is testing Pocket, a social app that lets users create, share, and discover AI-generated mini-games called 'gizmos' without coding. Currently in closed testing on Google Play, it may leverage cross-promotion with Facebook, Instagram, and WhatsApp.

Bellingcat developed an XGBoost-based machine learning model that dramatically reduced the time needed to identify civilian-harm incidents from Telegram posts. Trained on a dataset of verified cases and engineered features, XGBoost outperformed logistic regression, random forest, LightGBM, and several large language models (including Gemma and Gemini) in precision and recall. The open-source methodology shifts researcher focus from searching to verifying, and is adaptable to other conflict zones.