Nuntia — must readTechnical AI news: what was released, and what is worth your time.https://news.naklitechie.com/feed.xml2026-11-13T03:36:47.083ZNuntiaOpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.…https://news.naklitechie.com/#x%3A21205351875872687772026-11-11T15:12:41.000ZOn Continual-ARC, our new benchmark at @letsunifyai , none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both. — @DanielLenton1 on X · 1391 likesFew-Step Streaming Video Generation via Condition Shiftinghttps://news.naklitechie.com/#arxiv%3A2611.140332026-11-09T19:00:00.000ZAPEX extends a streaming video generation method by replacing a separate fake-score network with condition shifting to estimate the generated distribution’s velocity field, improving instruction following and visual aesthetics while reducing memory and iteration time relative to DMD. — Deyuan Liu + 10 authors · 88 upvotestencent/Qwen3.8-27B-DSpine-b16https://news.naklitechie.com/#hf%3Atencent%2Fqwen3.8-27b-dspine-b162026-11-07T14:20:29.000ZDSpine drafts a whole block of tokens in one pass, passing each position's predicted features to the next position at every layer, so each draft token is conditioned on its predecessors without a second step. — text-generation · licence apache-2.0 · 210 likes · 3.7k downloadsToday, we're releasing the highest accuracy Qwen3.8-27B GGUFs.…https://news.naklitechie.com/#x%3A21205369467747698602026-11-11T15:19:40.000ZOur new Dynamic V3.1 GGUFs lead by 5–10% on average across Terminal Bench, DeepSWE & more. We’re working on even better version for Qwen4 Run on 12GB RAM. GGUF: https://huggingface.co/unsloth/Qwen3. — @UnslothAI on X · 3259 likesDon't Just Trust Your Memory: Evidence-Linked Memory Reduces Memory-Induced Degradation in LLM Agentshttps://news.naklitechie.com/#arxiv%3A2611.143202026-11-09T19:00:00.000ZWikiTrace links each distilled lesson to its source trajectory and rationale, enabling LLM agents to verify and correct memories before acting, improving performance and reducing memory-induced degradation across agentic benchmarks. — Yukyeong Lee + 3 authors · 67 upvotes · code releasedTree-Ising: Repairing Language Models with Bit-Level Ising Editshttps://news.naklitechie.com/#arxiv%3A2611.142102026-11-09T19:00:00.000ZTree-Ising repairs reasoning loops in LLM weights by solving sparse Ising/QUBO problems over BF16 parameters, improving math benchmark accuracy with fewer weight edits and no gradient or LLM-token cost. — Hadi Salloum + 1 authors · 63 upvotes · code releasedDoes a harness learn from an agent's failures, or from the search around them? A matched-control test on a frozen 9B agenthttps://news.naklitechie.com/#arxiv%3A2611.113422026-11-08T00:00:00.000ZArkid MitraPersistent Agents are the next unlock of AI usage Our spend on AI is now 45% persistent agents Unfortunately for Anthropic, Claude Tags…https://news.naklitechie.com/#x%3A21206177041011552702026-11-11T20:40:34.000Zsucks OpenAI's Slack agent is even worse We primarily use Perplexity Computer, which uses a lot of different models Perplexity's Harness mogs — @dylan522p on X · 1150 likesOur initial public offering: https://ipo.anthropic.com Legal disclaimer: Anthropic has filed a registration statement (including a…https://news.naklitechie.com/#x%3A21208417313021547082026-11-12T11:30:46.000Zprospectus) with the SEC for the offering to which this communication relates. Before you invest, you should read the prospectus in that registration — @AnthropicAI on X · 9538 likesKinetic — an exact-output local inference engine (Velocity)https://news.naklitechie.com/#gh%3Aveloresearch%2Fkinetic2026-11-10T00:00:00.000ZVelocity (Jakub Kozica, CEO and founder, author of the technical report)