Tootfinder

OpenAI is rolling out GPT-5.4-Cyber to some participants in its Trusted Access for Cyber program, a week after Anthropic announced Mythos (Rachel Metz/Bloomberg)
https://www.bloomberg.com/news/articles/2026-04-14/openai-rele…

When the EIB grants a €20B loan for AI infrastructure while its yearly budget is €100B, that effectively means one-fifth of its annual budget goes to AI infrastructure.
Public transport? Green energy? Social infrastructure? No. Instead, data centres operated by private consortia to potentially sell private services. I hope it will be at least to produce open-source models and software.
#ai

»Matrix messaging gaining ground in government IT:
One-to-one and group messaging, encrypted VoIP calls, video conferencing – the open protocol handles them all«
As I see, @… is used peatically and therefore more popular. When will this become more popular in companies and in general? Any criticism of any state should not prevent this.
💬

Matrix is quietly becoming the chat layer for governments chasing digital sovereignty
FOSDEM 2026: One-to-one and group messaging, encrypted VoIP calls, video conferencing – the open protocol handles them all

An initial group of 28 malicious skills targeting Claude Code and Moltbot users were published to ClawHub and GitHub between January 27-29, 2026. A second larger group of 200 skills were published January 31-February 1, bringing the total number of malicious skills to 230 .
https://opensourcemalware.…

OpenSourceMalware.com - Community Threat Intelligence
Security professionals sharing intelligence on malicious packages, repositories, and CDNs to protect the open source ecosystem.

Fighting back is one piece; building the world we want is another. For The People will host an Open House on Feb. 19th for folks interested in strengthening their #libraries by getting involved in Friends of the Library & Foundation groups.

Applications Now Open! Rare Book School Summer 2026 https://networks.h-net.org/group/announcements/20139027/applications-now-open-rare-book-school-summer-2026

De #trike reed vandaag in opgevouwen toestand met de auto naar Dronten alwaar hij in Gerrit's Ligfiets Paradijs, ik bedoel natuurlijk Ligfietsshop Tempelman, is nagezien en gerepareerd aan een kabelbreukje.
Breken die dingen zomaar? Nee, nauwelijks, maar dit is een 2de hands die dat in de 25.000 km van zijn vorige bestaan al eens is overkomen, waarna de toenmalige eigenaar dat heeft gerepa…

Applications Now Open! Rare Book School Summer 2026 #acrel https://networks.h-net.org/group/announcements/20139027/applications-now-open-rare-b…

Calibrations for the Sasaki volume on odd spheres and the no-gap problem
Jonas Matuzas
https://arxiv.org/abs/2602.22961 https://arxiv.org/pdf/2602.22961 https://arxiv.org/html/2602.22961
arXiv:2602.22961v1 Announce Type: new
Abstract: For each odd sphere $S^{n}$ with $n=2m 1\ge 5$, we consider the Sasaki volume functional $\mathrm{Vol}^S(V)=\int_{S^{n}}\sqrt{\det(I (\nabla V)^{\top}(\nabla V))}\,d\mathrm{vol}$ on smooth unit tangent vector fields $V$. Using the Brito--Chacon--Naveira calibration $\omega=a\wedge\Theta$ on the unit tangent bundle $E=UTS^{n}$, we establish the universal calibrated lower bound $\mathrm{Vol}^S(V)\ge c(m;1)\,\mathrm{vol}(S^{n})$, where $c(m;1)=4^{m}/\binom{2m}{m}$. In the relaxed (integral-current) setting, we show that the section-constrained stable mass in $E$ equals the calibration value and is attained by an $\omega$-calibrated mass-minimizing integral $n$-cycle in the section class.
We also analyze the equality case on smooth graphs. If a smooth graph is $\omega$-calibrated on an open set, then it satisfies the rigidity system $\nabla_V V=0$ and $\nabla_X V=\lambda X$ for all $X\perp V$, hence is locally a radial distance-gradient field. In particular, for $m\ge 2$ there is no smooth unit field on $S^n$ whose graph is $\omega$-calibrated everywhere.
Finally, we construct an explicit smooth recovery sequence (presented in detail for $S^5$ and then extended to all odd dimensions) and prove a uniform nonvanishing estimate for the polar-shell normalization in the patching construction. As a consequence, $\inf_{V}\,\mathrm{Vol}^S(V)=c(m;1)\,\mathrm{vol}(S^{n})$, so there is no Lavrentiev gap.
toXiv_bot_toot

Marco DeepResearch: Unlocking Efficient Deep Research Agents via Verification-Centric Design
Bin Zhu, Qianghuai Jia, Tian Lan, Junyang Ren, Feng Gu, Feihu Jiang, Longyue Wang, Zhao Xu, Weihua Luo
https://arxiv.org/abs/2603.28376 https://arxiv.org/pdf/2603.28376 https://arxiv.org/html/2603.28376
arXiv:2603.28376v1 Announce Type: new
Abstract: Deep research agents autonomously conduct open-ended investigations, integrating complex information retrieval with multi-step reasoning across diverse sources to solve real-world problems. To sustain this capability on long-horizon tasks, reliable verification is critical during both training and inference. A major bottleneck in existing paradigms stems from the lack of explicit verification mechanisms in QA data synthesis, trajectory construction, and test-time scaling. Errors introduced at each stage propagate downstream and degrade the overall agent performance. To address this, we present Marco DeepResearch, a deep research agent optimized with a verification-centric framework design at three levels: \textbf{(1)~QA Data Synthesis:} We introduce verification mechanisms to graph-based and agent-based QA synthesis to control question difficulty while ensuring answers are unique and correct; \textbf{(2)~Trajectory Construction:} We design a verification-driven trajectory synthesis method that injects explicit verification patterns into training trajectories; and \textbf{(3)~Test-time scaling:} We use Marco DeepResearch itself as a verifier at inference time and effectively improve performance on challenging questions. Extensive experimental results demonstrate that our proposed Marco DeepResearch agent significantly outperforms 8B-scale deep research agents on most challenging benchmarks, such as BrowseComp and BrowseComp-ZH. Crucially, under a maximum budget of 600 tool calls, Marco DeepResearch even surpasses or approaches several 30B-scale agents, like Tongyi DeepResearch-30B.
toXiv_bot_toot