AI, machine learning and LLM insights.
The Claude Code sub-agent issued a recursive PowerShell delete that targeted the root of the Windows user profile. Because the safety classifier was offline, the system’s fallback incorrectly allowed the command, erasing 234,884 files in under two minutes.
Internal documents show Project OT aimed to cut up to 60% of staff by swapping human teams for AI agents, but the bots fell short, employee morale plunged from 74% to 55%, and Zuckerberg pulled the plug just hours before a November wave.
Arga’s “digital-twin” sandboxes clone entire SaaS apps—including permissions and webhooks—so reinforcement-learning agents can be trained on realistic, resettable environments, a capability missing from today’s thin API-only test rigs.
The videos, posted on Telegram, were spotted by analysts for irregular blinks and flat-toned Ukrainian audio, yet their personalized nature lets a small team churn out dozens of plausible clips, exploiting the “liar’s dividend” that erodes trust in real footage.
LLM Structured Outputs: Valid But Wrong Structured outputs are not what the marketing says. I tested 12 model APIs to see how they handle JSON schemas. Here is what I found. The G…
Moonshot AI is negotiating a revenue-share arrangement that would let its Kimi K3 model run on Microsoft Azure, Amazon Web Services, and Google Cloud, turning the startup from a cloud tenant into a content partner and reshaping AI monetisation.
The Bengaluru startup, which launched a no-code website builder in 2025, now plans to let a single AI assistant handle SEO, ad campaigns and content creation, aiming to replace outsourced marketers for small businesses.
The trio lets developers embed standard cron expressions directly in Claude Code, storing prompts until execution and automatically cleaning up after a session or a seven-day cap, removing the need for external scripts.
The new Granite 4.2 suite, released under Apache 2.0, includes 3 B, 8 B and 30 B models trained on 15 trillion tokens and features built-in tool-calling, dual-mode compute, and compatibility with vLLM and SGLang for high-throughput inference.
Backed by Qatar Investment Authority, Koch Disruptive Technologies and other institutional investors, Gatik will use the $200M to upgrade AI models, add hardware and expand its middle-mile autonomous fleet across the U.S., bringing total funding above $600M.
The new Orin Nano 2 blends a GPU-style core with dedicated neural accelerators, handling vision-language and small diffusion models on a credit-card-sized board while staying power-efficient enough for UAVs and handheld inspection tools.
The fresh capital will fund a 150-person engineering surge, a broader product suite, and expansion into Europe and APAC, positioning SynthAI to capitalize on the projected $150 billion enterprise AI spend in 2025.
The watermark is a subtle statistical pattern baked into word choice and sentence rhythm, persisting through copy-paste and minor edits. Only heavy rewrites, multiple translations, or blending with large human-written blocks erase it.
The new toggle wraps any Cloudflare Worker with the Zero Trust gateway, enforcing Google Workspace, Microsoft Entra or Okta authentication without any code changes, letting teams protect internal dashboards and utilities instantly.
The gang feeds AI assistants like PentestGPT with IP ranges, letting the models handle reconnaissance, exploit selection and payload generation, shrinking the attack timeline from days to minutes and allowing simultaneous compromises across continents.
The fresh $10 million extension brings Ringg’s total Series A to $15.5 million and lets the startup shift from cheap outbound calls to complex, outcome-driven workflows for clinics, fintechs, and retailers, all while expanding through Global Capability Centers.
The new event-triggered tasks let ChatGPT react the moment a matching email lands, a Slack message appears, or a GitHub pull-request changes, with optional filters, dedup keys and a forbidden-action list to keep high-risk operations safe.
The bug stems from a reduction performed along the wrong axis in PyTorch’s pure-Python chunked-scan fallback, letting later tokens flow backward during inference. It silently lowers loss and perplexity for checkpoints like Zamba2-1.2B (chunk-size 256) and Nemotron-H-8B (chunk-size 128).
The Series B round, led by 8VC and bolstered by Nvidia, Union Square Ventures, Bezos Expeditions and AI pioneer Fei-Fei Li, pushes Generalist’s total capital to $600 million as it readies Gen 1.5 for real-world trials in logistics and manufacturing.
Chris Malone, who arrived from Google and Meta last March, guided OpenAI’s $500 million Stargate build-out and brokered deals with Microsoft, Nvidia, Oracle and SoftBank. His sudden departure adds to more than a dozen senior exits this year, raising doubts about the rollout’s timeline and the firm’s
OpenAI’s Jalapeño inference chip, unveiled at Hot Chips 2026 with Broadcom, trims token-generation latency by up to 3.6x and boosts power efficiency 1.5-1.9x, letting the company cut GPT-5.6 Sol API rates by 20-33% for developers.
The WHOI team fused rapid sonar sweeps with a French image-matching routine, turning 2-D sonar pixels into depth cues and stitching them into a live 3-D model fast enough for split-second navigation, then handed off to high-resolution cameras once clarity returns.
The capsule, just 6x 4 mm, houses a 1 mm2 silicon chip that uses a leakage-current oscillator for 0.01 °C accuracy and a backscatter antenna that reflects external UHF signals, extending battery life on a 1.55-V coin cell.
The collaborative team engineered lipid-nanoparticle LNPs that deliver mRNA encoding two immune-activating genes, achieving a 10-15-fold T-cell boost in mouse models of bladder, colon, melanoma, lung cancers and even COVID-19 and flu vaccines, while sidestepping cytokine-related toxicity.
