Create

Sign in to ReadmeX

Sign in to join communities, post, vote and chat.

New here?

AI News

‹Wed, October 7
All dates
619
Industry

Pentagon's Tradewinds uses 5-minute videos to speed AI 'kill chain' buys

Tradewinds, a procurement marketplace run by the Pentagon's Chief Digital and Artificial Intelligence Office (CDAO), lets companies submit product videos of no more than five minutes describing how their AI offerings address periodically updated "strategic focus areas"; a panel reviews submissions at least monthly, and accepted products gain "post-competitive" status that makes it easier to buy them via vehicles such as Other Transaction agreements. A DOD official says some awards have been made in under a week, and OpenAI, Anthropic and Google are all listed as participants. WIRED reports that 2026 topics seek AI that can "enhance force lethality" and "unleash" AI-enabled agents for joint targeting, including so-called kill chain execution, while earlier pitches on a "culture of responsible AI" and guardrails are no longer sought.

WIRED AI·
6210
Open source

Rapid-MLX: open-source LLM server for Apple Silicon tops GitHub Trending

Rapid-MLX is an Apache-2.0 local LLM inference server built on MLX for Apple Silicon, exposing both OpenAI (/v1/chat/completions, /v1/responses) and Anthropic Messages (/v1/messages) APIs and focusing on reliable tool calling for coding agents. The project claims up to 4× faster decoding than Apple's mlx-lm on identical weights: a 1.50× median across 18 streaming tasks with Qwen3.5-9B 4-bit on an M4 Pro Mac mini, and up to 4.27× on a whole-file code edit (46.4 to 198.4 tok/s), plus 3.0× Ollama's aggregate decode at 8 concurrent streams on Qwen3.6-35B-A3B. All figures are the project's own measurements, and much of its comparison table is drawn from other projects' own documentation.

GitHub Trending · Python·
639
Industry

Tony Fadell on why the first wave of AI gadgets failed

Speaking at MIT Future Fest, Tony Fadell — credited with the iPod, the iPhone and as founder of Nest — said the first wave of AI hardware (Rabbit R1, Humane AI Pin, Limitless pendant) failed because it “didn’t meet any kind of need” and was just interesting technology for geeks. He argued trust and safety are paramount for a personal AI assistant, noting Meta’s Muse had a serious vulnerability found soon after launch, and that Apple has the hardware, chips and privacy credibility but no world-class model of its own — the new Siri AI runs on custom versions of Google’s Gemini. He predicted a successful agent will have to run on-device only, for privacy and to stay lightweight.

TechCrunch AI·
649
Industry

Kuaishou forms enterprise AI productivity unit; Yu Yue named Kling AI CEO

DoNews reports that Kuaishou has set up a virtual organization called "Enterprise AI Productivity" to build its office foundation, general-purpose agents and information systems in a unified way. The report also says Cheng Yixiao will concurrently lead the community science line, while Yu Yue moves to become CEO of Kling AI.

DoNews·
657
Industry

Applied Materials and Intel Expand Collaboration on AI Chip Manufacturing

Applied Materials and Intel announced a collaboration at Applied Materials’ EPIC Center and Intel’s Hillsboro research campus to advance transistor, interconnect, and advanced packaging technologies for AI computing. The work includes materials and process development for advanced logic nodes, interconnect scaling, and packaging for Intel’s Foveros 3D stacking; the companies say it is intended to accelerate the path from research to U.S. production.

科技新报·
669
Open source

Stacklok's Mecatl: a Kubernetes-style cloud-native harness for agents

Stacklok, founded by Kubernetes creators Craig McLuckie and Joe Beda, pivoted from software supply chain security to Kubernetes-based agentic infrastructure. Its open source “cloud-native harness” Mecatl, started in June, separates the agent loop from tool execution, session management and memory; the companion open source project ToolHive runs and governs MCP servers on Kubernetes. The company says its AI Gateway is not yet open sourced but is on the roadmap, with an “enterprise spine” control plane as the commercial product.

Latent Space·
678
Products

Google prepares Antigravity for Gemini 4 Argon and a Concierge voice agent

TestingCatalog spotted new labels in Google's Antigravity coding app around usage limits, subscription tiers and prompts to upgrade to Google AI Ultra, plus UTM links carrying an "argon_model_selector" campaign parameter — suggesting Ultra subscribers may get Gemini 4 Argon first. Per Google's September 30 announcement, Argon already powers internal workflows and reportedly scores 77.9% on DeepSWE v1.1 for longer software-engineering tasks, with rollout starting from trusted cyber defenders to paid API customers and Ultra subscribers. Also found in development: an internal agent prototype called Concierge with real-time voice, an Auto mode that may reduce permission prompts, and UI skins — none with a confirmed release timeline.

TestingCatalog·
688
Industry

Ex-Tesla Optimus AI lead launches industrial robot startup Intelligent Machines

Ashish Kumar, a former AI lead for Tesla's Optimus humanoid robots, has co-founded Intelligent Machines, a San Francisco startup building non-humanoid robots for tasks normally done by highly skilled workers in manufacturing and aerospace, according to The Information, citing two people with direct knowledge. The company has hired several former members of Tesla's Optimus team and is raising a seed round that could end up around $100 million. It currently focuses on robot software, using reinforcement learning to teach complex tasks, though its leaders have discussed designing hardware as well.

The Information·
698
Products

Kore.ai launches Autoloop to keep optimizing enterprise AI agents after go-live

Kore.ai introduced Autoloop, an optimization engine for agents built on its Kore.ai Agent Platform: customers set goals such as task completion, adherence to business rules, cost and accuracy, and Autoloop adjusts the agents automatically, including after deployment. It drafts each agent from a company's existing operating procedures, then uses Kore.ai's StateTrace evaluation layer to spot goal misses against the full production record and Agent Blueprint Language to rewrite only the responsible part. Autoloop is available now to customers on the platform's Artemis edition.

SiliconANGLE AI·
708
Products

Cisco brings Dialog agentic framework to Webex enterprise platform

Cisco is updating its Webex cloud collaboration platform with a new "agentic harness" called Dialog, which it says directs multiple AI agents to handle customer interactions and keep working on the customer's behalf after a conversation ends, coordinating across people, agents and backend systems. The refreshed Webex also lets workspace members invite AI agents directly into spaces, meetings and calls to execute multi-step tasks, such as pulling up statistics during a presentation and updating a PowerPoint live. Cisco says it is working with data analytics company Splunk for a safety-forward experience, though these capabilities and reliability claims are not independently verified.

Engadget AI·
7114
Industry

General AI Models Move Upstream in Video Creation

An analysis describes two ways general-purpose models are entering AI video: Claude Opus 5.5 renders programmable videos through code, while GPT-6 Astra plans scenes and camera movements before handing rendering to specialist video models. It argues that ByteDance, MiniMax and Kuaishou are responding by expanding from standalone generation into end-to-end creator workflows, so specialist video models may retain an advantage in production reliability and integration.

创业邦 科技·
728
Industry

North American startup funding falls 35% QoQ to $92B in Q3, with $61B to AI

North American startups raised $92B in Q3, down 35% quarter over quarter but up 50% year over year, according to Crunchbase News. AI startups took $61B of that total, and the quarter saw 11 acquisitions worth $1B or more, while IPOs stayed quiet.

Techmeme·
738
Industry

Rivercell raises $25M to build an AI virtual cell model

Rivercell has raised a $25M seed round led by HV, with HCVC, Alven and Bpifrance Digital Venture participating. The money will fund its data platform, a larger wet lab in Paris and its AI Virtual Cell programme, which the company describes as a world model of the cell that predicts how a cell changes after treatment. Rivercell says no dataset of comparable scale exists for how cells respond to drugs over time, so it plans to generate its own by filming living cells through each experiment and reading their gene activity, with the model choosing the next experiment and the company building its own instruments.

The Next Web·
747
Industry

Taiwan outlines ‘National Cybersecurity Protection 2.0’

Taiwan’s National Institute of Cyber Security has outlined “National Cybersecurity Protection 2.0,” a plan to use AI for source-code vulnerability detection, automated threat-intelligence analysis, AI-driven red-team testing and a controlled cyberattack sandbox. According to the institute, AI red-team exercises for government agencies are scheduled to run from April through October next year, subject to the plan’s implementation.

iThome 台湾·
758
Industry

Report: Anthropic May Seek $100B in IPO Funding

A Cyzone article, reprinted from Jiedian AI, claims Anthropic has confidentially filed an S-1 and may seek up to $100 billion at a valuation above $2 trillion after the U.S. midterm elections. It presents the reported plan alongside alleged 2025 revenue of $4.6 billion, net losses of $42 billion, and $518 billion in future compute and infrastructure commitments, while warning that the figures and valuation depend on rapid growth and high utilization.

创业邦 科技·
768
Industry

Sereact Robots Sort Zalando Returns in Germany and Poland

Sereact’s two-armed robots are sorting Zalando fashion returns at CEVA Logistics sites in Greven, Germany, and Świebodzin, Poland. The companies say Sereact’s Cortex software can grasp, identify and sort unfamiliar, crumpled items without product-specific training, while workers shift toward supervision, exception handling and quality checks.

The Next Web·
778
Models

Fine-Tuned Nemotron Reports Gold-Level IOI and IMO Results

A Hugging Face post reports gold-level results after fine-tuning Nemotron for the International Olympiad in Informatics (IOI) and International Mathematical Olympiad (IMO). The provided source contains no further details about the fine-tuning method or results.

Hugging Face·
788
Industry

Bridge Neurotech Raises $13.5M for Ultrasound BCI

Bridge Neurotech has emerged from stealth with $13.5 million to develop a non-invasive brain-computer interface that uses functional ultrasound imaging to measure brain activity through the skull. The startup also signed a multi-year technology and hardware agreement with Butterfly Network, while its first human study is enrolling healthy adults in California.

The Next Web·
7911
Open source

thedotmack/claude-mem

Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More https://claude-mem.ai 🇨🇳 中文 • 🇹🇼 繁體中文 • 🇯🇵 日本語 • 🇵🇹 Português • 🇧🇷 Português • 🇰🇷 한국어 • 🇪🇸 Español • 🇩🇪 Deutsch • 🇫🇷 Français • 🇮🇱 עברית • 🇸🇦 العربية • 🇷🇺 Русский • 🇵🇱 Polski • 🇨🇿 Čeština • 🇳🇱 Nederlands • 🇹🇷 Türkçe • 🇺🇦 Українська • 🇻🇳 Tiếng Việt • 🇵🇭 Tagalog • 🇮🇩 Indonesia • 🇹🇭 ไทย • 🇮🇳

GitHub Trending（每日）·
808
Products

Sesame updates voice agents, each with its own computer

Sesame released a fresh preview of its voice assistant in its iOS and Android apps, powered by a new voice model. Its agents are described as smarter and each has access to its own computer, letting users speak hands-free, have them plan trips or research hobbies and get an update when work is done, use tools, coordinate other agents, connect to everyday services starting with several Google apps, and attach their own MCPs. The company also laid out a three-part plan covering hand-made Japanese frames, a voice-first assistant for smart glasses it says are coming in 2027, and characters designed to be less sycophantic; developer tooling is due later this year.

TestingCatalog·
Next →← Previous