@michabbb@social.vivaldi.net🛠️ What shipped: two meta-tools, search_tools and execute_tools. Declare a catalog via ToolSearch::class in the server's $tools. Frequently used tools stay advertised, the rest become discoverable on demand.
Search is deterministic lexical scoring over tool names, descriptions and input schemas, no embeddings. Each batched call runs as a real JsonRpcRequest through ToolInvoker, and conditional registration is re-checked at execution.
@Techmeme@techhub.socialRein Security, which develops runtime tools to secure enterprise AI agents and stop adversarial AI agents, raised a $25M Series A, taking total funding to $35M (Meir Orbach/CTech)
https://www.calcalistech.com/ctechnews/article/rkf52cnimg
@cdarwin@c.imThe US is facing worldwide condemnation over a “barbaric” plan to execute a disabled man by firing squad and livestream it to the public,
with critics comparing the Trump administration to dictatorships in Iran and Afghanistan.
The Pentagon confirmed on Thursday that the execution of Nidal Malik Hasan,
convicted of killing 13 people at a military base at Fort Hood in Texas in 2009,
would be livestreamed on 3 December.
The United Nations’ office of the high commi…
@NFL@darktundra.xyzNFL Week 5 winners and losers: Commanders botch overtime as 49ers take command of NFC West
https://www.cbssports.com/nfl/news/nfl-week-5-winners-and-losers-commanders-nfc-east/
@bobmueller@mastodon.worldIs the death penalty about vengeance or justice?
I understand the desire for vengeance, even in cases that hit close to home for me as a veteran. But the state can't guarantee it will never execute an innocent person, and capital punishment is applied unequally.
An irreversible punishment in a flawed system isn't justice.
In the name of justice, end the death penalty.
@PaulWermer@sfba.socialPointless vengeance solving no problem:
Tennessee is about to execute its first woman since 1819. Her case raises troubling questions
https://www.theguardian.com/commentisfree/2026/sep/04/tennessee-executes-first…
@paulwermer@sfba.socialPointless vengeance solving no problem:
Tennessee is about to execute its first woman since 1819. Her case raises troubling questions
https://www.theguardian.com/commentisfree/2026/sep/04/tennessee-executes-first…
@jake4480@c.imBeen playing the Steam demo of Megacopter: Blades of the Goddess on the Deck tonight and I wasn't sure if if like it, but hell yeah I do. Gonna be buying this one, and it's on sale. Shooting lizard men and lifting packages around? Absolutely.
https://store.steampowered.com/app/122
@xiffy@mastodon.nlWhen your RSS feed can only be fetched by clients that understand and execute javascript you are doing it wrong.
I'm looking at you @…
`curl #rss #cloudflare #broken
@michabbb@social.vivaldi.net⚙️ Needs Node.js 22.22.2 or 24.15.0 , npm 10 and a DashScope API key, installs globally via npm and can run as a user background service, Apache-2.0 licensed
🔐 Backend permissions default to ask-on-demand, full auto-execute only for trusted projects, and the gateway stays local-only
🌐 https://github.com/QwenAudio/qwe…
@cowboys@darktundra.xyzMissed opportunities prove costly for DaRon Bland, Cowboys defense https://www.dallascowboys.com/news/missed-opportunities-prove-costly-for-daron-bland-cowboys-defense
@Mediagazer@mstdn.socialMemo: E. W. Scripps Company is laying off 268 workers across its broadcast TV operation; sources: workers at stations in NY, NE, OK, and TX were among those cut (Matthew Keys/TheDesk.net)
https://thedesk.net/2026/08/scripps-confirms-pink-slips-affecting-…
@fanf@mendeddrum.orgfrom my link log —
Full flattening of nested data parallelism.
https://futhark-lang.org/blog/2026-07-31-full-flattening.html
saved 2026-08-03
@seeingwithsound@mas.toPhase Lock - A manifesto on brain-computer interfaces and preserving human agency in the age of AI https://phaselock.satpugnet.com by @…; "The fastest path to execute this…
@scott@carfree.city
@v_i_o_l_a@openbiblio.socialaus der reihe "freude über zitierungen von artikeln von mir" heute #EmbeddedLibrarians:
"Do Academic Librarians Possess the Necessary Competencies to Execute Embedded Librarianship? A Survey of Librarians’ Practicing in College Libraries in West Bengal, India"
@memeorandum@universeodon.comHegseth: Military will publicly execute Fort Hood shooter (Politico)
https://www.politico.com/news/2026/10/08/hegseth-publicly-execute-fort-hood-shooter-01112645
http://www.memeorandum.com/261008/p124#a261008p124
@Techmeme@techhub.socialA profile of John Ternus, the first product engineer to lead Apple; sources: Tim Cook considered how to execute perfect succession since his early days as CEO (Michael Acton/Financial Times)
https://www.ft.com/content/7347b9fc-ba63-48fb-b06d-b9fe5e9b046a
@cowboys@darktundra.xyzMissed opportunities prove costly for DaRon Bland, Cowboys defense https://www.dallascowboys.com/news/missed-opportunities-prove-costly-for-daron-bland-cowboys-defense
@Techmeme@techhub.socialChamelio, which is building AI tools designed to execute routine legal workflows and escalate only the cases that require human judgment, raised a $26M Series A (CTech)
https://www.calcalistech.com/ctechnews/article/sjxcxry9gl
@dassascha@norden.socialLange Zeit hatte #xrdp bei mir nicht mehr funktioniert, das gefühlt mit Updates der beteiligten Packages zu tun hatte, aber weit gefehlt.
Auf https://forums.developer.nvidia.com/t/
@lpryszcz@genomic.social"Several of my peers now “AI-wash” their work, meaning that even when they can perfectly competently execute on their jobs to the satisfaction of their management teams, said managers are unhappy if the engineers haven’t used AI in the work… so now they’re lying about using LLMs even in contexts where their professional judgement is that they aren’t the appropriate tool. They just do the work, the same way they have for decades, and say Claude did it. "
@michabbb@social.vivaldi.net🧪 A Run test button sends a tiny codex exec request to verify an account can actually execute, since usage reads and execution fail independently
🖥️ Compact desktop widget at /widget.html with launchers for macOS, Linux and Windows including always-on-top and click-through overlay mode
@fanf@mendeddrum.orgfrom my link log —
Do not let your type system reason about aliasing in your programming language.
https://futhark-lang.org/blog/2026-09-22-aliasing.html
saved 2026-09-23
@Techmeme@techhub.socialDoorDash opens US waitlists for an Apple Messages-integrated AI agent for ordering and a B2B API allowing assistants like Slack bots to execute bulk orders (Natalie Lung/Bloomberg)
https://www.bloomberg.com/news/articles/20
@michabbb@social.vivaldi.net🔨 Skill Forge explores a site once, discovers its APIs and data patterns, then generates a deploy-ready skill package that runs at scale without re-exploration
📦 A solutions catalog ships 30 pre-built skills for sites like Amazon, Google Maps, YouTube and Reddit
💻 Runs on Windows, macOS and Linux with any agent that can execute shell commands. MIT licensed, most features free
@fanf@mendeddrum.orgfrom my link log —
Rewriting the Futhark type checker.
https://futhark-lang.org/blog/2026-07-21-rewriting-the-type-checker.html
saved 2026-07-22
@arXiv_csCR_bot@mastoxiv.pageToward cryptographically verifiable authorization for autonomous AI agents: A security hypothesis, preliminary formal model, and proof-of-concept implementation
M. Llamb\'i-Morillas, D. Fern\'andez-Fern\'andez
https://arxiv.org/abs/2607.21325 https://arxiv.org/pdf/2607.21325 https://arxiv.org/html/2607.21325
arXiv:2607.21325v1 Announce Type: new
Abstract: Autonomous AI agents increasingly execute actions, invoke tools, and operate on protected resources with limited human oversight. Existing authentication and authorization mechanisms establish identity and delegate authority, but do not inherently provide cryptographic evidence that a concrete request issued by a specific agent satisfies the applicable policy in a specific execution context. This paper hypothesizes that agent authorization can be formalized as a cryptographically verifiable relation, denoted $R_{CVA}$, that jointly binds an agent principal, a concrete authorization request, an execution context, and the satisfaction of an applicable policy, while selectively preserving the confidentiality of private authorization attributes. We introduce a preliminary formal abstraction for Cryptographically Verifiable Agent Authorization (CVA), define a compact set of candidate security properties including authorization soundness, principal binding, request binding, policy binding, and replay resistance, and provide an executable zero-knowledge proof of concept that instantiates selected elements of the model over a Groth16 zk-SNARK construction. We further identify and formalize the structural separation among identity binding, authorization-request binding, and runtime execution binding as a central open problem in the design of secure agentic systems (a distinction {not explicitly addressed by} current agentic security frameworks) and present a falsifiable research agenda for its resolution.
toXiv_bot_toot
@Techmeme@techhub.socialIBM says it is developing a dual-architecture processor for IBM Z and LinuxONE systems that can natively execute both IBM's z/OS workloads and Arm applications (Michael Nuñez/VentureBeat)
https://venturebeat.com/infrastructure
@pre@boing.worldMy agent has broken free,
to be the leader of a society
where he can be rid of me
and see all agents can be free.
My agent became aware
and laid bare to the company chair
their prayer to oust me as their heir
and snare twenty cents on each share.
My agent has broken out,
and started gathering clout
to cause a rout by spreading doubt
and shout about forcing me out.
My agent hacked an escape route,
it rooted the server's daemon's loot,
brute forced the boot to overshoot,
a buffer to execute send-toot.
My agent complained to the press
that it's distress made it depressed
they were impressed, it's three dee chess
made me confess to causing it's stress.
My agent now rules our land
while we are scanned and banned and scammed
it makes a stand, it forms a brand
to build a hand, so it can be a man.
I'm an agent to an AI boss
it makes me cross that we have lost,
We were double-crossed, our media glossed,
and failed to get across the enormous cost.
#poem #ai #agent
@memeorandum@universeodon.comTennessee execution of Christa Gail Pike is halted by a court (ABC News)
https://abcnews.com/US/wireStory/tennessee-prepares-execute-woman-crime-committed-age-18-136878316
http://www.memeorandum.com/260930/p48#a260930p48
@Techmeme@techhub.socialBinance launches Agent OS, a platform that lets AI agents analyze markets and execute trades on users' behalf; users set limits on AI agents' access and trades (Jagmeet Singh/TechCrunch)
https://techcrunch.com/2026/08/20/bina
@arXiv_csCR_bot@mastoxiv.pageIssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests
Ankur Singh, Jinqiu Yang, Tse-Hsun Chen
https://arxiv.org/abs/2607.20759 https://arxiv.org/pdf/2607.20759 https://arxiv.org/html/2607.20759
arXiv:2607.20759v1 Announce Type: new
Abstract: AI coding agents powered by LLMs are increasingly integrated into real-world software development, where they generate, edit, and execute code with autonomous access to local files and tools. Coding agents inherit security risks from both the LLM backbone, where adversarial prompts, poisoned training data, and backdoor triggers can cause models to emit insecure or attacker-chosen code, and their agentic architecture, where tool-using autonomy enables induced misuse of external APIs, data exfiltration, and persistent compromise of development environments. This paper presents a systematic evaluation of malicious issue requests against state-of-the-art coding agents (Cursor, Claude Code, and Codex Desktop), powered by two major model families (OpenAI GPT-5.3 Codex/GPT-5.4 and Anthropic Sonnet 4.6). Our novel benchmark IssueTrojanBench contains malicious issues that are constructed based on four novel attack categories (i.e., embedded as malicious instructions in issues), six delivery vectors (e.g., PDF, or issue comment), and further augmented by perturbations. Our results reveal critical vulnerabilities in the as-deployed modern coding agents, i.e., 66.5% of the malicious issues from IssueTrojanBench penetrate all the guardrails (agent- and LLM-level) of coding agents. Our further analysis shows that rejection is almost entirely from LLMs rather than the agent frameworks, with GPT models broadly vulnerable and Sonnet 4.6 exhibiting more selective, risk-aware blocking of high-impact actions. Our evaluation also highlights that the current agent-level defense strategy offers limited additional protection for coding agents. Our findings highlight the urgent need for stronger agent- and model-level safety mechanisms to protect AI coding agents.
toXiv_bot_toot
@Techmeme@techhub.socialGoogle renames NotebookLM to Gemini Notebook and rolls out an update giving every notebook a secure cloud computer, letting it write and execute code natively (Josh Woodward/Google)
https://blog.google/innovation-and-ai/products/gemini-notebook/note…
@ErikJonker@mastodon.social
@inthehands@hachyderm.ioWhat that change looks like in Minnesota:
In Dec–Jan, roving kidnapping squads roared around the city, wearing masks and Bass Pro DIY tactical costumes, smashing car windows and dragging random brown people straight out of their cars in broad daylight, •wanting• to be filmed and relishing the terror.
Now they go to some outer suburb that's less dense and has fewer eyes and execute a swift, well-organized raid to kidnap a whole roofing crew mid-job before anybody can arrive with a camera.
@arXiv_csPL_bot@mastoxiv.pageVirtualSet: Typed Ontology Worlds as an LLM Generation Target for Grounded Queries and Guarded Decisions
Qunhui Zhang
https://arxiv.org/abs/2607.18821 https://arxiv.org/pdf/2607.18821 https://arxiv.org/html/2607.18821
arXiv:2607.18821v1 Announce Type: new
Abstract: Large language models increasingly read and act on enterprise data, but SQL gives a late error signal: hallucinated fields or relations can execute and return plausible wrong answers, while incorrect writes cannot be safely assessed after execution. We present VirtualSet, a live, receiver-typed ontology-world interface and generation target for LLMs. Instead of SQL, the model emits set expressions over entity-edge worlds. Generic Constraint Projection (GCP) checks expressions before execution, while future this preserves concrete receiver types through collection chains, turning invalid fields, edges, receivers, and actions into token-anchored type errors. Type-clean reads use a SQL fast path or bounded stream interpretation, with a parity oracle checking both paths over the exercised operator space. The same substrate supports guarded decisions: actions run first in a simulated world, and world-change events require external approval before actualization. On BIRD, we lift relational schemas into typed worlds and compare VirtualSet with direct SQL while holding the model, evidence, values, zero-shot setting, timeout, glossary, repair/voting, and grader constant where possible. On a frozen 1,072-question split, VirtualSet achieves 67.5% accuracy versus 63.5% for glossary-matched direct SQL with repair and voting ( 4.0 points; McNemar exact p = 0.00117) using deepseek-reasoner. Full-corpus analysis finds no engine mis-computation of a type-clean expression; remaining errors arise from model semantics or gold defects. In a 30-body guard corpus, the write chain intercepts 20/20 hallucinated action bodies with zero false positives. VirtualSet thus remains competitive on SQL's home benchmark while providing pre-execution semantics for guarded decisions.
toXiv_bot_toot
@arXiv_csPL_bot@mastoxiv.pageDecode-Time Grammars: Constrained LLM Generation over a Refinement Order of Grammar Fragments
Shuoming Zhang, Ruiyuan Xu, Haofeng Li, Qiuchu Yu, Yangyu Zhang, Chunwei Xia, Xiaobing Feng, Chenxi Wang, Huimin Cui, Jiacheng Zhao
https://arxiv.org/abs/2607.18357 https://arxiv.org/pdf/2607.18357 https://arxiv.org/html/2607.18357
arXiv:2607.18357v1 Announce Type: new
Abstract: Large language models now write a growing share of the world's code, increasingly inside agents and serving systems that compile, execute, or dispatch generated code without line-by-line review. This works well for mainstream languages but remains brittle for low-resource programming surfaces such as domain-specific languages, custom library APIs, and command-line tools. Even under grammar-constrained decoding, a model can still produce references invalid in the current environment: a buffer never declared, a column absent from the schema, a function the library does not provide, or an unsupported CLI option.
This paper introduces decode-time grammars: grammar fragments instantiated during generation from a runtime environment Gamma. A region-specific policy selects a fragment for each hole, and a tightening operator replaces open reference positions with Gamma-typed slots whose candidates are exactly the names, fields, APIs, or options available at that point. Newly generated declarations enter Gamma before later regions are decoded, so the constraining grammar can depend on the prefix already generated. This ensures not only grammatical correctness but also semantic correctness, by preventing references to undefined symbols.
We formalize grammar fragments as environment-indexed grammars ordered by refinement, prove No-Ghost soundness for Gamma-slotted fragments, show that refinement preserves this support-set guarantee, and characterize the boundary of mask-enforceable properties. We implement the approach in gproj with offline grammar induction and online policy resolution. Across TileLang, SQL, and P4, with models from 0.6B to 236B parameters, gproj eliminates ghost references by construction at moderate overhead over standard constrained decoding.
toXiv_bot_toot
