techpolicy.press
That is the central tension now emerging in digital governance. Governments are likely to interpret genuine AI security risks as justification for even greater control over digital infrastructure, data, and technological ecosystems. Yet fragmentation and isolation may ultimately weaken security rather than strengthen it. The history of cybersecurity has repeatedly shown that resilience depends less on unilateral control than on collaboration: shared standards, coordinated defenses, information sharing, open research communities, and cooperation across governments, companies, technical experts, and civil society.
audiomass.co
AudioMass is a free, open source, web-based Audio and Waveform Editor. It runs entirely in the browser with no backend and no plugins required!
Tips: Please keep in mind that most key shortcuts rely on the Shift + key combo. (eg Shift+Z for undo, Shift+C copy, Shift+X cut... etc )
You can load any type of audio your browser supports and perform operations such as fade in, cut, trim, change the volume, and apply a plethora of audio effects.
Check out the codebase on Github
vaticannews.va
government regulation of the private companies that are driving the development of A.I.
protection and retraining for workers whose jobs are threatened
education to help students think critically about the technology
action to protect children from violent, hypersexualized or fake information online that is often generated by A.I.
safeguards to ensure that humans, not artificial intelligence, remain responsible for all decisions regarding the use of weapons.
github.com
Heretic is a tool that removes censorship (aka "safety alignment") from transformer-based language models without expensive post-training. It combines an advanced implementation of directional ablation, also known as "abliteration" (Arditi et al. 2024, Lai 2025 (1, 2)), with a TPE-based parameter optimizer powered by Optuna.
ghuneim.us
Four phases of the network, four sets of C-words. The five that describe how the network governs us are the same five that name the controls of the studio. Same words, different sides.
brennan.day
It doesn't need to be this way. https:// is not the only way to connect and interface with the Internet. Some that you may know are ftp:// for file transfers, mailto: for email composition, ssh:// for secure shell access, irc:// for Internet Relay Chat, or magnet: for peer-to-peer downloads. The majority of Internet browsers do not play nicely even with these protocols, handing them off to other applications.
arxiv.org
We aggregate 4.14M recipes from 11 sources spanning seven languages (English, Chinese, Russian, Vietnamese, Spanish, Turkish, Indonesian, German, and Indian-English) and normalise the raw ingredient strings to 1,790 canonical entries via an LLM-augmented pipeline. A 203,508-edge ingredient–ingredient NPMI graph and an 80,019-edge typed FlavorDB ingredient–compound graph (2,247 typed compound nodes across 15 categories) seed three Metapath2Vec variants that share architecture and hyperparameters and differ only in the random-walk schema: Cooc walks the co-occurrence graph only, Chem walks the typed compound metapaths only, and Core blends both via injected ingredient–ingredient walks at controlled mixing, placing each model at a distinct point on the chemistry-vs-recipe-context spectrum.
arxiv.org
Every agent interaction generates a next-state signal, namely the user reply, tool output, terminal or GUI state change that follows each action, yet no existing agentic RL system recovers it as a live, online learning source. We present OpenClaw-RL, a framework built on a simple observation: next-state signals are universal, and policy can learn from all of them simultaneously. Personal conversations, terminal executions, GUI interactions, SWE tasks, and tool-call traces are not separate training problems. They are all interactions that can be used to train the same policy in the same loop. Next-state signals encode two forms of information: evaluative signals, which indicate how well the action performed and are extracted as scalar rewards via a PRM judge; and directive signals, which indicate how the action should have been different and are recovered through Hindsight-Guided On-Policy Distillation (OPD). We extract textual hints from the next state, construct an enhanced teacher context, and provide token-level directional advantage supervision that is richer than any scalar reward. Due to the asynchronous design, the model serves live requests, the PRM judges ongoing interactions, and the trainer updates the policy at the same time, with zero coordination overhead between them. Applied to personal agents, OpenClaw-RL enables an agent to improve simply by being used, recovering conversational signals from user re-queries, corrections, and explicit feedback. Applied to general agents, the same infrastructure supports scalable RL across terminal, GUI, SWE, and tool-call settings, where we additionally demonstrate the utility of process rewards
github.com
ECHO is an environment cross-entropy hybrid objective, which trains terminal agents by combining policy-gradient RL with an on-policy cross-entropy loss for predicting environment tokens.
ECHO is implemented as an extension on top of SkyRL: SkyRL provides the core RL training stack, while this repo adds the terminal-agent integration, environment prediction loss, example configs, and a small SkyRL hook patch.
jonaharagon.com
Interconnected Local Meshes Distinct local groups being able to interconnect is a huge boon for content availability on the network, and the beauty is that all these network links in Reticulum automatically become redundant as more connections are made. A local mesh in Minneapolis could interconnect with a local mesh in Chicago over the internet, for example, but perhaps in the future some dedicated network operators are also able to establish a direct connection via microwave or LoRa between those cities. Connections may normally continue to traverse the internet at higher speeds, but in the event of an outage those alternate/ad-hoc paths can take over seamlessly, because they're all just paths on the same, single Reticulum network.
ft.com
One could plausibly argue that insider trading isn’t just a bug of prediction markets, it is an essential feature in what insiders insist is a mechanism for “truth”. And evidence of the Trump administration’s support for prediction markets is . . . manifold.
Even the head of the American Gaming Association has called Michael Selig — the Trump-appointed head of the CFTC and the self-appointed prediction market regulator — “quite frankly, a joke” for his championing of the industry. Not that the president gets the joke.
This is why the latest investigation is an intriguing development. Yes, this is just a solitary Google employee allegedly up to shenanigans with a very narrow type of bet — on what will top Google search rankings — but perhaps it is a sign of things to come?
research.panmureliberum.com
Strategy What if... the AI boom goes into reverse?Strategy Economics & ESG·14 May 2026