AIagent.at 🤖 AI News

The evolution from prompt engineering to concept engineering marks a pivotal shift in how we build with AI. Instead of fragile prompt strings, practitioners now use reusable, testable building blocks that are more reliable and maintainable. This approach represents a future-proof way to work with AI systems. https://www.kdnuggets.com/the-evolution-from-prompt-engineering-to-concept-engineering #AIagent #AI #GenAI #Research

Frore Systems, a deep tech chip startup, has reached unicorn status after raising funding, hitting a 1.64B USD valuation. The company focuses on innovative chip technology for AI applications. https://techcrunch.com/2026/03/16/another-deep-tech-chip-startup-becomes-a-unicorn-frore-hits-1-64b/ #AIagent #AI #GenAI #Startup #Frore

Mistral AI released Mistral Small 4, a 119B-parameter MoE model that combines instruction following, reasoning and multimodal capabilities in one deployment. Features 128 experts with 4 active per token, 256k context window and configurable reasoning effort. Claims 40% faster completion and 3x throughput vs Small 3. https://www.marktechpost.com/2026/03/16/mistral-ai-releases-mistral-small-4-a-119b-parameter-moe-model-that-unifies-instruct-reasoning-and-multimodal-workloads/ #AIagent #AI #GenAI #Mistral

NanoClaw is an open-source AI agent platform created in a weekend that went viral on Hacker News. Andrej Karpathy's post made it explode. Now Docker is integrating the tool. https://techcrunch.com/2026/03/13/the-wild-six-weeks-for-nanoclaws-creator-that-led-to-a-deal-with-docker/ #AIagent #AI #GenAI #AgenticAI #Docker

Meta's Avocado AI model is delayed until at least May after falling short in internal tests against Google, OpenAI and Anthropic models. Despite committing 600B USD to AI infrastructure by 2028, the model outperformed Meta's previous version and Gemini 2.0 but not Gemini 3.0. https://gizmodo.com/mark-zuckerbergs-billion-dollar-hiring-spree-avocado-delays-ai-2000733284 #AIagent #AI #GenAI #AIEnterpriseBackbone #Meta

The Pentagon designated Anthropic a supply chain risk because its AI model Claude has a unique ethical framework called a Constitution that guides behavior. The DoD argued this could pollute the supply chain and threaten national security. Anthropic refused to remove guardrails against mass surveillance and autonomous weapons. This is the first time a US company has received this designation. https://gizmodo.com/the-pentagon-claims-that-anthropics-soul-creates-a-supply-chain-risk-that-makes-no-sense-2000733091 #AIagent #AI #GenAI #GlobalAIGovernance #Anthropic

Goose, the open-source AI coding agent from Block, does what Claude Code does for free. Running locally via Ollama, it offers privacy, no rate limits, and zero subscription costs. Over 26K GitHub stars. https://venturebeat.com/infrastructure/claude-code-costs-up-to-usd200-a-month-goose-does-the-same-thing-for-free #AIagent #AI #GenAI #AIInfrastructure #Block

Zoom launches an AI-powered office suite with Docs, Slides, and Sheets, plus AI avatars that can represent users in meetings. The company says avatar deepfake detection is also coming. AI agent builder for non-technical users and voice translator are also new. https://techcrunch.com/2026/03/10/zoom-launches-an-ai-powered-office-suite-says-ai-avatars-for-meetings-are-coming-soon/ #AIagent #AI #GenAI #AgenticAI #Zoom

The Pro-Human Declaration, signed by hundreds of experts including Max Tegmark, outlines a framework for responsible AI development with five key pillars: keeping humans in charge, avoiding power concentration, protecting human experience, preserving liberty, and holding AI companies accountable. It calls for banning superintelligence development until safe and requiring mandatory off-switches on powerful systems. https://techcrunch.com/2026/03/07/a-roadmap-for-ai-if-anyone-will-listen/ #AIagent #AI #GenAI #AIGovernance

#CaitlinKalinowski, #OpenAI’s #robotics leader, #resigned over concerns about the company’s agreement with the #Pentagon to deploy its models on a classified network. She cited concerns about #surveillance and #autonomousweapons, stating that these issues deserved more deliberation. OpenAI has since clarified restrictions on #military use of its systems. https://fortune.com/2026/03/07/openai-robotics-leader-caitlin-kalinowski-resignation-pentagon-surveillance-autonomous-weapons-anthropic/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI

Google releases Workspace CLI: a command-line tool that bundles Gmail, Drive, and Calendar APIs for AI agents. The open-source tool supports structured JSON outputs and includes 40+ agent skills, offering a simpler alternative to MCP for connecting AI to Workspace data. https://arstechnica.com/ai/2026/03/googles-new-command-line-tool-can-plug-openclaw-into-your-workspace-data/ #AIagent #AI #GenAI

Block laid off 40% of staff citing AI, but data scientists were offered 75% pay raises to stay. The narrative that AI is replacing jobs may be more 'AI washing' than reality, new analysis suggests. https://gizmodo.com/the-curious-case-of-the-block-ai-layoffs-2000730673 #AIagent #AI #GenAI #Workforce

Perplexity released pplx-embed, new SOTA Qwen3 bidirectional embedding models for web-scale retrieval: These multilingual models use bidirectional attention and diffusion pretraining for better semantic understanding in RAG pipelines. https://www.marktechpost.com/2026/02/26/perplexity-just-released-pplx-embed-new-sota-qwen3-bidirectional-embedding-models-for-web-scale-retrieval-tasks/ #AIagent #AI #GenAI #AgenticAI

MIT researchers have developed a method that uses idle computing time to double LLM training speed while preserving accuracy: The system trains a smaller drafter model to predict outputs of the larger reasoning LLM, which then verifies the predictions, reducing computational workload. https://news.mit.edu/2026/new-method-could-increase-llm-training-efficiency-0226 #AIagent #AI #GenAI #AIInfrastructure

Anthropic CEO Dario Amodei refuses Pentagon demands for unrestricted access to AI systems, citing concerns about mass surveillance of Americans and fully autonomous weapons. The Friday deadline approaches. https://techcrunch.com/2026/02/26/anthropic-ceo-stands-firm-as-pentagon-deadline-looms/ #AIagent #AI #GenAI #Governance

Mistral AI inks enterprise deal with Accenture: The French AI lab partners with the global consulting giant to develop enterprise AI solutions, also becoming a customer to roll out the technology to employees. https://techcrunch.com/2026/02/26/mistral-ai-inks-a-deal-with-global-consulting-giant-accenture/ #AIagent #AI #GenAI #EnterpriseAI

New method could double LLM training speed: MIT researchers leverage idle GPU time to train a smaller drafter model that accelerates reasoning-model training by up to 210%. The Taming the Long Tail system preserves accuracy while reducing computational overhead. https://news.mit.edu/2026/new-method-could-increase-llm-training-efficiency-0226 #AIagent #AI #GenAI #AIInfrastructure

Anthropic CEO Dario Amodei has refused the Pentagon's demand for unrestricted access to its AI systems: He said he 'cannot in good conscience accede' to the request, citing concerns about mass surveillance of Americans and fully autonomous weapons. The deadline is Friday 5:01 p.m. https://techcrunch.com/2026/02/26/anthropic-ceo-stands-firm-as-pentagon-deadline-looms/ #AIagent #AI #GenAI #Anthropic #AIGovernance

Microsoft Research has developed CORPGEN, a framework enabling autonomous AI agents to manage multiple concurrent organisational tasks: The system addresses four key failure modes including context saturation and memory interference, achieving up to 3.5x better completion rates than baseline agents. https://www.marktechpost.com/2026/02/26/microsoft-research-introduces-corpgen-to-manage-multi-horizon-tasks-for-autonomous-ai-agents-using-hierarchical-planning-and-memory/ #AIagent #AI #GenAI #AgenticAI

Anthropic CEO Dario Amodei says he 'cannot in good conscience accede' to the Pentagon's request to give the military unrestricted access to its AI systems: the two cases are mass surveillance of Americans and fully autonomous weapons with no human in the loop. The Friday 5:01 p.m. deadline gives Anthropic a choice between complying or facing consequences. https://techcrunch.com/2026/02/26/anthropic-ceo-stands-firm-as-pentagon-deadline-looms/ #AIagent #AI #GenAI #Governance #Anthropic

Mistral AI, the French AI research lab, announced a multiyear partnership with global consulting giant Accenture to develop enterprise AI solutions: The deal involves both companies working together to create enterprise technology powered by Mistral's AI models for clients, with Accenture also becoming a Mistral customer and rolling out the technology to its employees. This partnership follows similar deals between Accenture and other AI providers like OpenAI and Anthropic, as AI companies increasingly partner with consultants to help enterprises find ROI from AI adoption. https://techcrunch.com/2026/02/26/mistral-ai-inks-a-deal-with-global-consulting-giant-accenture/ #AIagent #AI #GenAI

Sophia Space raises US$ 10M seed to demo novel space computers: the company plans to prove out a new approach to passively cooling space computers, using modular server racks with integrated solar panels called TILES. Founders came from a US$ 100M Caltech program for orbital solar plants. https://techcrunch.com/2026/02/26/sophia-space-raises-10m-seed-to-demo-novel-space-computers/ #AIagent #AI #GenAI #AIInfrastructure

Read AI launches Ada, an email-based digital twin that manages schedules, answers questions from company knowledge bases, and handles email correspondence: the assistant replies to meeting requests with your availability, responds to out-of-office emails, and can answer questions by pulling from meeting data, prior discussions, and web searches. Unlike MCP-based solutions, Read AI builds a knowledge graph from connected services for contextual responses. https://techcrunch.com/2026/02/26/read-ai-launches-an-email-based-digital-twin-to-help-you-with-schedules-and-answers/ #AIagent #AI #GenAI

Anthropic has acquired Vercept, a startup specializing in computer-use AI agents: the deal brings Vercept's team, including co-founders poached by Meta, to Anthropic while shutting down its product on March 25. The acquisition highlights the race to build advanced AI agents that operate computers autonomously. https://techcrunch.com/2026/02/25/anthropic-acquires-vercept-ai-startup-agents-computer-use-founders-investors/ #AIagent #AI #GenAI #Anthropic #Vercept

Google has unveiled Nano Banana 2, a new image generation model that delivers significantly faster performance. The model is now integrated as the default option in the Gemini app and AI mode. https://techcrunch.com/2026/02/26/google-launches-nano-banana-2-model-with-faster-image-generation/ #AIagent #AI #GenAI #Google #ImageGeneration

Figma has partnered with OpenAI to integrate Codex, the company's AI coding assistant, directly into its design platform. This integration comes a week after Figma announced a similar partnership with Anthropic's Claude Code, signaling Figma's strategy to offer multiple AI coding options to developers. https://techcrunch.com/2026/02/26/figma-partners-with-openai-to-bake-in-support-for-codex/ #AIagent #AI #LLM #GenAI #OpenAI #Figma #Codex

Trace launches with US$ 3M in seed funding to help enterprises adopt AI agents, backed by Y Combinator, Zeno Ventures, and other investors: the startup aims to solve the friction companies face when integrating autonomous AI agents into their workflows. https://techcrunch.com/2026/02/26/trace-raises-3-million-to-solve-the-agent-adoption-problem/ #AIagent #AI #LLM #GenAI #Enterprise #Trace

MIT researchers developed an AI framework that provides holistic information on cells by identifying which data is shared across different measurement modalities and which is unique to specific types: the approach could help scientists better understand disease mechanisms like cancer, Alzheimer's, and diabetes, and plan more effective experiments. https://news.mit.edu/2026/ai-help-researchers-see-bigger-picture-cell-biology-0225 #AIagent #AI #LLM #GenAI #AIresearch #MIT

MIT researchers developed a new method that doubles LLM training speed by leveraging idle computing time: the system trains a smaller drafter model during downtime that predicts outputs for the larger reasoning model, accelerating training without additional overhead while preserving accuracy. https://news.mit.edu/2026/new-method-could-increase-llm-training-efficiency-0226 #AIagent #AI #LLM #GenAI #Training #MIT

Google expands Gemini's agentic capabilities on Android: the AI assistant can now automate multi-step tasks like rideshare requests, grocery orders, and food delivery. This positions Gemini as a more autonomous digital assistant for everyday workflows. https://techcrunch.com/2026/02/25/gemini-can-now-automate-some-multi-step-tasks-on-android/ #AIagent #AI #LLM #GenAI #Google #Gemini

The White House wants AI companies to cover electricity rate increases from data center power demand, though most hyperscalers have already pledged to cover the costs. Anthropic, Meta, Microsoft, and OpenAI are among the companies that have made public commitments to absorb higher energy costs as AI infrastructure expansion drives increased electricity demand across the grid. https://techcrunch.com/2026/02/25/the-white-house-wants-ai-companies-to-cover-rate-hikes-most-have-already-said-they-would/ #AIagent #AI #LLM #GenAI #Governance #Policy

Around 12% of U.S. teenagers now use AI chatbots for emotional support or advice, according to new Pew Research data: Mental health professionals express concern since general-purpose AI tools like ChatGPT and Claude are not designed for therapeutic use, raising questions about the appropriate role of AI in adolescent mental health. https://techcrunch.com/2026/02/25/about-12-of-u-s-teens-turn-to-ai-for-emotional-support-or-advice/ #AIagent #AI #LLM #GenAI #AIEthics

»Amazon unveils AI smart glasses for its delivery drivers« https://techcrunch.com/2025/10/22/amazon-unveils-ai-smart-glasses-for-its-delivery-drivers/?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI