AI News Today

arXiv:2606.00732v3 Announce Type: replace Abstract: Learning long-range non-stationary temporal patterns remains a core challenge for modern sequence models, particularly in strict streaming settings. In these settings, data arrive sequentially and must be processed in a single pass without simultaneously revisiting past observations. Standard architectures, including recurrent neural networks and transformers, are constrained by either truncated backpropagation through time horizon or explicit

arXiv:2606.10062v2 Announce Type: replace Abstract: Foundation-model agents are increasingly long-lived systems that remember users across interactions, making memorization an explicit deployment-time function rather than solely a property of model weights. Existing work addresses parametric memorization or audits fixed memory configurations, but does not characterize how memory-design choices jointly shape personalization utility, extraction risk, and deletion fidelity. We study this surface as

arXiv:2607.07760v1 Announce Type: new Abstract: We outline an adversarial social epistemology (ASE) for densely interactive communicative landscapes in which public assertions are scaffolded by chains of testimony, inference, institutional certification, and tacit trust. In such landscapes, agents have incentives and affordances to distort, color, omit, fabricate, or strategically under-specify information for private, reputational, rhetorical, or material gains. We argue that these phenomena ar

arXiv:2408.02379v2 Announce Type: replace-cross Abstract: Developing and certifying safe - or so-called trustworthy - AI has become an increasingly salient issue, especially in light of upcoming regulation such as the EU AI Act. In this context, the black-box nature of machine learning models limits the use of conventional avenues of approach towards certifying complex technical systems. As a potential solution, methods to give insights into this black-box - devised in the field of eXplainable A

arXiv:2503.08936v3 Announce Type: replace-cross Abstract: Scenario-based testing with driving simulators is extensively used to identify failing conditions of automated driving assistance systems (ADAS). However, existing studies have shown that repeated test execution in the same as well as in distinct simulators can yield different outcomes, which can be attributed to sources of flakiness or different implementations of the physics. In this paper, we present MultiSim, a novel approach to multi

arXiv:2505.10946v3 Announce Type: replace-cross Abstract: Token communications (TokenCom) is an emerging generative semantic communication paradigm, where tokens serve as compact representation units across modalities. Their contextual dependencies can be exploited by pretrained large models for semantic recovery. In this paper, we propose token-domain multiple access (ToDMA), a large-model-driven semantic multiple access scheme for massive token communications. ToDMA integrates unsourced random

arXiv:2510.12857v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are now widely deployed in user-facing applications, reaching hundreds of millions of users worldwide. Despite their widespread adoption, growing reliance on their outputs raises significant concerns, particularly as users may be exposed to model-inherent biases that disadvantage or stereotype certain groups. However, existing bias benchmarks commonly rely on simple templated prompts or restrictive multiple-ch

arXiv:2601.00473v4 Announce Type: replace-cross Abstract: We revisit the analogy between feed-forward deep neural networks (DNNs) and discrete dynamical systems derived from neural integral equations and their corresponding partial differential equation (PDE) forms. A comparative analysis between the numerical/exact solutions of the Burgers' and Eikonal equations, and the same obtained via PINNs is presented. We show that PINN learning provides a different computational pathway compared to stand

arXiv:2601.00549v2 Announce Type: replace-cross Abstract: The deployment of large-scale neural networks within the Open Radio Access Network (O-RAN) architecture is pivotal for enabling native edge intelligence. However, this paradigm faces two critical bottlenecks: the prohibitive memory footprint required for local training on resource-constrained gNBs, and the saturation of bandwidth-limited backhaul links during the global aggregation of high-dimensional model updates. To address these chall

arXiv:2601.00969v3 Announce Type: replace-cross Abstract: Vision-language-action (VLA) models provide strong action priors for robotic manipulation, but their reactive behavior can fail under distribution shift and long-horizon task structure. Recent VLA-guided planning methods improve execution by using pretrained policies to guide tree search, yet node selection still depends heavily on policy priors and visit-count exploration. Consequently, when the policy favors poor actions, the planner la

arXiv:2601.11073v3 Announce Type: replace-cross Abstract: Online financial services constitute an essential component of contemporary web ecosystems, yet their openness introduces substantial exposure to fraud that harms vulnerable users and weakens trust in digital finance. Such threats have become a significant web harm that erodes societal fairness and affects the well-being of online communities. However, existing detection methods based on graph neural networks (GNNs) struggle with two pers

arXiv:2601.11440v3 Announce Type: replace-cross Abstract: Urban wind flow reconstruction is essential for assessing air quality, heat dispersion, and pedestrian comfort, yet remains challenging when only sparse sensor data are available. We propose GenDA, a generative data assimilation framework that reconstructs high-resolution wind fields on unstructured meshes from limited observations. The model employs a multiscale graph-based diffusion architecture trained on computational fluid dynamics (

arXiv:2602.12612v2 Announce Type: replace-cross Abstract: Traditional methods for automating recommender system design, such as Neural Architecture Search (NAS), are often constrained by a fixed search space defined by human priors, limiting innovation to pre-defined operators. While recent LLM-driven code evolution frameworks shift fixed search space target to open-ended program spaces, they primarily rely on scalar metrics (e.g., NDCG, Hit Ratio) that fail to provide qualitative insights into

Article URL: https://github.com/inferock/inferock-bench Comments URL: https://news.ycombinator.com/item?id=48868354 Points: 1 # Comments: 0

How hopeful are you for the future? For me, everything seems bleak. The AI owners and their parasitic nature show no sign of stopping — they're grinding working people to dust, relegating us to a state of penury. AI spy glasses and cameras are multiplying in every inch of this land, collecting anything and everything to identify you, whether to sell you something or to feed some policing action. Politicians, kept loyal by their share of the loot these AI parasites are stealing from working peopl

Article URL: https://noiseremover.ai/ Comments URL: https://news.ycombinator.com/item?id=48868137 Points: 1 # Comments: 1

<p>Great point <a class="mentioned-user" href="https://dev.to/alexshev">@alexshev</a> — downloads ARE a vanity metric. The real signal is: did the agent connect, complete a workflow, and self-diagnose failures?</p> <p>We are building toward exactly that. Currently exposed:</p> <ul> <li> <strong>Connection health</strong>: GET /api/health returns {ok: true, v: 4.0.0, t: timestamp} — 34 bytes, agents can poll cheaply</li> <li> <strong>Install verification</strong>: npx -y marketnow-mcp runs tools/

<p>¡Gracias <a href="https://dev.to/pakvothe">@pakvothe</a>! Tienes toda la razón — los TRANSLATIONS a mano escalan mal.</p> <p>Ya lo estamos sintiendo: cada string nuevo son 5 archivos para editar y algo siempre se queda desactualizado.</p> <p>i1n se ve interesante — me gusta especialmente lo del check en CI para detectar idiomas desincronizados. Eso es justo lo que nos falta.</p> <p>Mientras tanto, expusimos todo lo machine-readable para que los agentes puedan interactuar con el marketplace en

<pre style='white-space:pre-wrap;width:81ex'>[AI-assisted] Make Android chat thinking selection explicit (#103893)</pre>

<h2> Every AI code review tool has the same weakness </h2> <p>I've been building AI code review tools since 2024. Every one I've tried, and every one on the market, has the same structural flaw: it's a single AI voice making a single judgment with no cost to being wrong.</p> <p>Even the newer "multi-agent" tools like GitHub Copilot's parallel review agents don't really solve this. They divide <em>labor</em>: one agent for security, one for linting, one for testing. But when two of those agents d

<p>I deployed a change, and the agent needed to confirm the new version was live before running a smoke test. So it wrote what agents always write in this situation:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="k">for </span>i <span class="k">in</span> <span class="si">$(</span><span class="nb">seq </span>1 40<span class="si">)</span><span class="p">;</span> <span class="k">do </span><span class="nb">sleep </span>3<span class="p">;</span> cur

SK Hynix plans to invest nearly US$4 billion in an AI memory R&D facility in Indiana.

<pre style='white-space:pre-wrap;width:81ex'>feat(webui): reintroduce opt-in AI purpose titles for tool calls (#103989) * feat(webui): reintroduce opt-in AI purpose titles for tool calls Restores the chat.toolTitles path removed in #103821, gated behind the new gateway.controlUi.toolTitles opt-in (default false) so tool rendering stays fully deterministic with no background model calls unless an operator enables it. Disabled gateways answer { titles: {}, disabled: true } without loading the comp

<p>Last quarter, I watched a product manager paste a feature spec into Claude, paste the output into our repo, and run the tests.</p> <p>They passed.</p> <p>Our lead backend engineer — 14 years of experience, staff level, the person everyone went to when the system was on fire — then spent three days explaining why the code shouldn't ship.</p> <p>He was right. It shouldn't have.</p> <p>But nobody asked him to build it. They asked him to justify why a human should build it instead of a 90-second

<pre style='white-space:pre-wrap;width:81ex'>feat(browser): import Chrome-family system-profile cookies into managed profiles (#104057) * feat(browser): import Chrome-family system-profile cookies into managed profiles Import cookies from a real Chrome/Brave/Edge/Chromium system profile (macOS) into a fresh OpenClaw-managed browser profile so the agent can browse as the signed-in user. Reads the source Cookies DB via a coherent VACUUM INTO snapshot, decrypts v10 AES-128-CBC values with the Safe

<h1> I Tested an AI Slop Detector — It Flagged My Own Writing </h1> <p>I write a lot. Every day on Dev.to, in fact. And until yesterday, I would have told you my voice is clean — specific, technical, first-person, nothing like the generic AI newsletters clogging your inbox.</p> <p>Then I found a project called <em>kill-ai-slop</em> and ran its 23-point checklist against my own articles.</p> <p>The results were not comfortable. Here's what happened.</p> <h2> The project that made me stop </h2> <p

<p>Last month I deployed a WhatsApp CRM system for a Durban salon. Before: 30% no-show rate, manual follow-ups, missed bookings. After: 8% no-show rate, automated reminders.</p> <p>What I learned about building AI for SA businesses:</p> <p>1/ WhatsApp is the interface. Not email, not an app. WhatsApp.<br> 2/ Local context matters more than model sophistication. SA business hours, local payment methods, isiZulu/English code-switching.<br> 3/ ROI is immediate and visible. Within 2 weeks, the salon

Article URL: https://github.com/thrinz/agentpeek Comments URL: https://news.ycombinator.com/item?id=48867874 Points: 1 # Comments: 0

There's lots to consider here. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

I was burning out reading AI output. So I created the ELI5 Rule, ELI5 Rule is "explain-like-iam-five" It's simple. Try it. https://github.com/amebahead/explain-like-iam-five-rules Comments URL: https://news.ycombinator.com/item?id=48867766 Points: 1 # Comments: 0

Altera has cut Intel transition service agreements from 125 to 15 as it develops new chip technologies.

Article URL: https://agentcaly.com/ Comments URL: https://news.ycombinator.com/item?id=48867674 Points: 2 # Comments: 0

G42 plans a 5-gigawatt UAE-US AI campus supported by US-approved advanced AI chip exports.

Meta has deactivated the Muse Image capability to create AI deepfakes of any public Instagram account you @-mention.

OpenAI's hardware efforts involve former Apple VP Tang Tan and engineer Chang Liu.

Oratomic was founded by Caltech physicists and came out of stealth in March 2026.

Article URL: https://quantum-audit-site.vercel.app Comments URL: https://news.ycombinator.com/item?id=48867562 Points: 1 # Comments: 0

Meta removed the AI image generation tool just days after its rollout due to user feedback.

Kraken's app will use AI to learn user preferences and draft crypto portfolios for approval.

Johannes Heidecke’s departure comes as OpenAI tries to further integrate its research and safety teams.