@arXiv_csOS_bot@mastoxiv.page2026-06-03 07:36:32
Agent libOS: A Library-OS-Inspired Runtime for Long-Running, Capability-Controlled LLM Agents
Yingqi Zhang
https://arxiv.org/abs/2606.03895 https://arxiv.org/pdf/2606.03895 https://arxiv.org/html/2606.03895
arXiv:2606.03895v1 Announce Type: new
Abstract: Large language model (LLM) agents are evolving from request-response assistants into long-running software actors: they maintain state across model calls, fork subtasks, wait for external events, request human authority, generate tools, and perform side effects that must be resumed and audited. This paper presents Agent libOS, a library-OS-inspired runtime substrate for LLM agents. Agent libOS runs above a conventional host operating system; it does not implement hardware drivers, kernel-mode isolation, or a POSIX-compatible operating system. Instead, it treats an agent as an AgentProcess: a schedulable execution subject with process identity, parent-child lineage, lifecycle state, a tool table derived from an AgentImage, typed Object Memory, explicit capabilities, human queues, checkpoints, events, and audit records. Its central design rule is tools are libc-like wrappers; runtime primitives are the authority boundary. Filesystem access, object access, sleeps, human approval, JIT tool registration, and external side effects are checked at primitive boundaries under explicit capabilities and policy.
We describe the design, threat model, Python prototype, and safety-oriented evaluation. The current prototype implements async scheduling, namespace-local Object Memory, runtime-integrated human approval, one-shot permission grants, per-process working directories, shell and image-registration primitives, Deno/TypeScript JIT tools over a libOS syscall broker, filesystem/object bridge tools, an injectable Resource Provider Substrate, deterministic demos, real-model smoke scripts, and 123 regression tests at the time of writing. Rather than improving planner accuracy, Agent libOS demonstrates a runtime substrate in which long-running LLM agents can be scheduled, authorized, resumed, and audited without treating tool dispatch as the trust boundary.
toXiv_bot_toot
@andres4ny@social.ridetrans.it2026-05-26 09:50:11
War started: line goes up. War on pause: line goes up. War resumes: line goes up. Oil prices rise: line goes up. "Peace" talks: line goes up. Talks break down, more bombing: line goes up. Thermonuclear war threats: line goes up. Ceasefire summit: line goes up. Ceasefire starts: line goes up. Ceasefire actually completely made up by the US: line goes up. Bombings continue: line goes up.
Rational markets, baby!
@georgiamuseum@glammr.us2026-07-31 16:32:08
“Moving to work at the museum was a stark contrast to anything I had ever done before,” she said. “Overall, I’ve gained a wider knowledge of art.” Now, when asked about her favorite styles and artists, she tends to lean more towards contemporary art, an aesthetic she has developed an appreciation for over the years.
4/
@heiseonline@social.heise.de2026-06-22 13:28:00
Zweirad-Industrie startet neue Fahrradmesse in Köln
Mit der „towards tomorrow - European Bike Show“ startet 2027 eine neue Fahrradmesse in Köln. Die Branche setzt auf frische Impulse und Konkurrenz zur Eurobike.
https:…
@memeorandum@universeodon.com2026-05-29 20:20:48
Taryn Thomas was a committed member of the pro-Palestine movement. Then she went to Israel (Lianne Kolirin/Telegraph)
https://www.telegraph.co.uk/news/2026/05/27/received-death-threats-abandoning-pro-palestine-movement/
http://www.memeorandum.com/260529/p86#a260529p86
@aredridel@kolektiva.social2026-05-31 14:48:16
RE: https://mastodon.social/@yaelwrites/116647387354163140
This is really great "actual use"
But read with context: this what an investigative journalist does. This is very prudent in their line of work, and the threats that come with being a journalist who can incur the ire of powerful people and entities are in play here.
You really have to have a bit of a threat model to make good choices. A few things are universal, with little to no downside, but remarkably few.
@arXiv_physicscompph_bot@mastoxiv.page2026-07-02 09:00:03
Replaced article(s) found for physics.comp-ph. https://arxiv.org/list/physics.comp-ph/new
[1/1]:
- How Physical Dynamics Shape the Properties of Ising Machines: Evaluating Oscillators vs. Bistable...
Abir Hasan, Nikhil Shukla
https://arxiv.org/abs/2603.06843 https://mastoxiv.page/@arXiv_physicscompph_bot/116204003660873932
- Pulgon-tools: A toolkit for analysing and harnessing symmetries in quasi-1D systems
Yu-Jie Cen, Sandro Wieser, Georg K. H. Madsen, Jes\'us Carrete
https://arxiv.org/abs/2603.28435 https://mastoxiv.page/@arXiv_physicscompph_bot/116322896642584491
- MADField: Multi-fidelity Amortized Density Field for Adsorption in Nanoporous Materials
Yoonho Kim, Seongsu Kim, Sungsoo Ahn, Honghui Kim
https://arxiv.org/abs/2606.21284 https://mastoxiv.page/@arXiv_physicscompph_bot/116798531764581123
- Monte Carlo Physics-informed Neural Networks for Inverse Multiscale Heat Conduction Problems via ...
Qingyi Lin, Chuang Zhang, Xuhui Meng, Zhaoli Guo
https://arxiv.org/abs/2606.25793 https://mastoxiv.page/@arXiv_physicscompph_bot/116809736857979012
- TRIDS: AI-native molecular docking framework for accelerating high-throughput virtual screening w...
Liu, Zhang, Xue, Hao, Bing, Chai, Zhao, Yang, Zhang, Gao
https://arxiv.org/abs/2510.24186 https://mastoxiv.page/@arXiv_physicschemph_bot/115456702015641743
- Absence of poor local minima in matrix product states
Hao-Kai Zhang, Chenghong Zhu, Shuo Liu, Shi-Xin Zhang, Tao Xiang
https://arxiv.org/abs/2606.09988 https://mastoxiv.page/@arXiv_quantph_bot/116724766789822682
toXiv_bot_toot
@simon_brooke@mastodon.scot2026-05-25 15:10:19
Hauling logs across to the sawmill. I've got a certain amount done this morning, but it's now too hot to work.
#TheJoyOfCrofting
@cdarwin@c.im2026-07-24 16:42:54
Towards a Fossil-Free Future:
Seed the Commons invites you to a two-part webinar series this weekend so that you can join the movement to build a truly regenerative food system and a fossil-free future.
Date & Time
Jul 26, 2026
10:00 AM Pacific Time
https://us06…
@heiseonline@social.heise.de2026-07-07 06:11:00
