AI News

Mistral says its new model can afford organizations a kind of control over its cyber defenses that closed models can't.

FENGARDE is an open-source, self-hosted security pipeline for IT, AI and OT. Press replay: a recorded lab attack runs through the real console and the departure board announces it.

NVIDIA's GeForce RTX 5080 has seen a significant increase in its listed price over the past few weeks, with this high-end GPU now costing nearly $2,000. Due to price hikes throughout the supply chain and the GPU's 16 GB of GDDR7 memory, which is currently in short supply, the listing pric

Altera announced its new Agilex 2X series with the higher-end Agilex 27 with features like LPDDR6 and 1.6TbE support coming in 2027

Label-free reinforcement learning with verifiable rewards (RLVR) rewards a language model for agreeing with its own majority vote. It improves reasoning at first but eventually collapses: the self-assessed reward goes to one while accuracy falls, often until the model gives the same answer to every problem. The collapse appears only when training uses many problems. We explain this dataset-size dependence as a collective instability. In a solvable many-problem model in which problems share parameters, a population whose problems prefer different answers amplifies any shared perturbation for as long as the population-averaged slope $\langle F'\rangle$ of the single-problem self-reinforcement drift $F$ is positive. This holds while problems are still uncertain about their answers. Per-problem sharpening, slowed by $1/N$ for $N$ training problems, races against the shared mode, which grows at a rate $\proptoλ^2$ independent of $N$, where $λ$ is the strength of the shared coupling. This yields a critical number of problems $N_c\simeq\ln(1/δ)/(λ^2G)$, where $δ$ seeds the shared mode and $G$ decreases with the initial confidence. Simulations of the model confirm the predictions: a sharp transition, $N_c\proptoλ^{-2}$ up to a logarithmic correction, a single critical amplification $λ^2NG$ across couplings, confidence profiles and training variants, a logarithmic dependence on batch size, and no collapse for confident initial models. A kernel version of the theory identifies a protective coupling (generalization of sharpening) and a marginal-entropy bonus that stabilizes the uniform collective mode. In label-free RL of a small Transformer, removing the shared direction prevents collapse entirely, restoring it causes collapse, and collapse grows with training-set size and base-model entropy.

Hy-Nomas

Towards Free-form Interactive World Generation

paper: https://t.co/by82MMznL9
app: https://t.co/clt7ItswBD

The cloud security startup raised a Series C at a $3.8B valuation, instead of selling itself for the same price.

Simplismart says its annualized revenue run rate has grown fourteenfold in six months, with the U.S. accounting for more than half its revenue.

Disruptive has made headlines in recent years for pouring money into companies like chip startup Groq and the AI voice-generation startup ElevenLabs.

ChinaEU e UBTECH puntano sulla robotica in Europa. Gambardella: progetti pilota in Italia, integrazione industriale e nuove opportunità

Open-source canvas for orchestrating Claude Code, Codex, and other CLI agents. Run agents in parallel, delegate tasks, and let AI build reusable workflows. - chen-985211/cleancode

Multi-model AI aggregation system with auto-fallback, CCR closed-loop agent, and Self-IDE. Supports DeepSeek, NVIDIA, SiliconFlow, ZenMux, Zhipu, Kimi, MiMo, Mistral. - Alanzj/SelfClaw

Give your agents a home 🏠. Govern, deploy, monitor & manage AI agents. All SDKs & frameworks welcome - trytilde/tilde

Enframe is a drop-in sidecar that wraps any OpenAPI spec into a live MCP server. - erwinkramer/enframe

Source code of Local Code, a private AI agent for Mac, iPad and iPhone. Runs DeepSeek 4.1 on device, with Python, Node.js and a shell built into the app. - drawthingsai/local-code-community

CXMT is expected to start mass production of its first-ever HBM memory in the HBM3E form factor. According to an exclusive report from Reuters, this mass production is anticipated to begin within the next two months, meaning CXMT will officially start by the end of the year. Back in August

Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control helps and how much is needed remains incomplete. We analyze norm control in Muon over the entire training trajectory using the \emph{accumulated effective learning rate}, $C_T=\sum_{t Based on this insight, we propose MuonGS, which uses Gram--Schmidt (GS) projection to remove the radial component of each Muon update without fixing the weight norm. With bounded update directions and a constant nominal learning rate, we prove that the weight norm grows at most proportionally to $\sqrt{t}$ after $t$ steps. Consequently, MuonGS guarantees $C_T=Ω(\sqrt{T})$, with accumulation sustained even in later training. Further, for smooth scale-invariant objectives we establish its $O(T^{-1/2})$ stationarity guarantee. In pretraining experiments with 160M- and 400M-parameter dense models as well as a 135M active- and 1B total- parameter MoE model, MuonGS achieves lower validation perplexity than MuonH at long training horizons. In particular, for the dense model it achieves an estimated token-equivalent speedup up to about $15\%$ at matched validation loss to MuonH.

Join the discussion on this paper page

Highly recommended read. AI-written text is everywhere on the web now. This study measured how much.

Lots of interesting insights here.

They found that across more than 5 million crawlable pages from 2019 to 2026, about 24% contained AI-generated or AI-edited text in August 2026, up from near zer…

Gemini 4 Carbon is our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

The manufacturing accelerator bills itself as "America’s answer to Shenzhen."

General Compute is betting on inference chips from outside the Nvidia ecosystem.

L’intelligenza artificiale agentica sta evolvendo: non basta più automatizzare singole attività aziendali, occorre sviluppare agenti capaci di comprendere le specificità dei diversi settori e contribuire al raggiungimento di risultati misurabili. In questo video, Giuseppe Mascoli, CEO di Alterna, e Mirco Giorgini, COO dell’azienda, raccontano la nuova strategia del partner Microsoft, che punta sulla vertical AI con suite di agenti specializzati per manifatturiero, fashion, food & beverage, utility e servizi professionali. Un cambio di prospettiva che coinvolge non soltanto l’offerta tecnologica, ma anche l’organizzazione interna, la collaborazione con i clienti e il modello commerciale. Dalla co-creazione degli use case alla realizzazione dei prototipi, fino al passaggio da un pricing basato sulle giornate uomo a un approccio value-based e outcome-based, Alterna punta a legare il valore dell'AI ai risultati ottenuti sul campo. Quali vantaggi può offrire la vertical AI alle imprese? E come cambia il ruolo di un partner tecnologico nell'era degli agenti intelligenti? Guarda il video per scoprire la strategia di Alterna e le prospettive dell'intelligenza artificiale agentica specializzata. #VerticalAI #IntelligenzaArtificiale #AgenticAI #AI #Microsoft #Alterna #GenerativeAI #DigitalTransformation #InnovazioneDigitale #AI4Business

Contribute to DeepLink-org/pvisor development by creating an account on GitHub.