{"schemaVersion":"1.0","publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topic":"agents","availableTopics":[{"slug":"robotics","label":"Robotics"},{"slug":"open-models","label":"Open models"},{"slug":"data-centers","label":"Data centers"},{"slug":"biology","label":"Biology"},{"slug":"agents","label":"Agents"}],"total":6,"items":[{"id":"national-compute-research-preview-access","headline":"What National Compute’s research preview offers","canonicalUrl":"https://brightaifuture.com/discoveries/national-compute-research-preview-access","datePublished":"2026-10-07","dateModified":null,"sourcePublicationDate":"2026-10-07","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["agents"],"summary":"The restricted research preview describes access to eight-GPU nodes with $100 in starter credits. Its queue, billing rules and data-sharing terms show what that access means in practice.","evidenceState":"Emerging","keyFacts":[{"label":"AI’s role","value":"Marshall is the infrastructure agent used to request a whole GPU node in the restricted Public Research preview."},{"label":"Documented result","value":"For a researcher trying to get an experiment off the ground, National Compute’s most useful launch detail is a small, concrete offer: eligible institutional users can receive $100 in starter credits and request a whole eight-GPU node. The current documentation describes a restricted research preview, with access and hardware availability that can vary. It gives researchers enough detail to work out whether a first experiment fits, before treating a national compute grid as an established resource."},{"label":"Important limitation","value":"According to the sign-in documentation, new users with eligible .edu, .mil or .gov addresses confirm their email to create an account and a public-research organization with $100 in credits. Other users need admission to an existing organization, an invitation or a place in the waitlist rollout. Sign-ups can pause, including for institutional addresses, so domain eligibility does not guarantee immediate access."}],"limitations":["According to the sign-in documentation, new users with eligible .edu, .mil or .gov addresses confirm their email to create an account and a public-research organization with $100 in credits. Other users need admission to an existing organization, an invitation or a place in the waitlist rollout. Sign-ups can pause, including for institutional addresses, so domain eligibility does not guarantee immediate access.","Public Research allocates one whole node at a time, requested through Marshall, National Compute’s infrastructure agent, and accessed through SSH. The listed options are eight AMD MI355X GPUs or eight NVIDIA B300 GPUs, each with 288 GB of memory per GPU. The offer Marshall quotes is authoritative, and availability determines which options appear. There is no multi-node training, RDMA between pool nodes or notebook server in this route. That makes the preview a bounded environment for experiments that fit on one machine.","Requests join a first-come, first-served queue. Marshall shows a rate, the cost of a full lease and the balance required before confirmation. A request needs enough credit for its first hour, rather than the entire lease. Waiting is free; once the node is handed over, every GPU in it bills by the hour. A quiet or completed job does not stop that meter.","The current lease length is four hours, but it can continue and keep billing when nobody is queued. If another researcher is waiting after the lease length has passed, a preemption notice gives the current holder ten minutes to checkpoint. Releasing or losing the node destroys its local scratch storage. Results need to leave the node before that happens. Even the documented starter training job ends after about an hour while the node continues charging until released.","The distinction matters when reading the larger grid’s promises. Vultr describes a Grid Exchange vision built around completed work, or goodput. The current Public Research documentation specifies GPU-hour billing for capacity held. Those are different descriptions of service economics; the wider vision does not make this preview a pay-only-for-success experiment.","Academic access also has a specific data-sharing condition. Organizations created from .edu addresses enter Marshall’s expanded tier: usage analytics, conversation transcripts and qualifying training artifacts, including specified checkpoints, datasets, logs and configurations copied from documented workspace and shared-storage locations. The artifact rules include filename patterns, size thresholds and credential exclusions. The stated purpose includes improving Marshall and training its underlying model.","The October 6 terms give National Compute a perpetual, irrevocable license to approved Marshall content for those purposes. Academic organizations cannot change this setting inside the service; a request to change it may require withdrawal from the research program. Organizations created with .gov or .mil addresses start at usage analytics only. Researchers therefore need to consider what they are entitled to share, alongside the hardware they can access.","The credit terms add another practical boundary: purchased credits ordinarily expire after one year and are generally non-refundable. Promotional credits follow their issuance terms; the documentation reviewed does not establish a specific expiry for the $100 starter grant. Compute stops when the balance reaches zero."],"evidenceLinks":[{"title":"National Compute sign-in and account documentation","url":"https://docs.nationalcompute.com/sign-in/","type":"institution"},{"title":"National Compute Public Research documentation","url":"https://docs.nationalcompute.com/public-research/","type":"institution"},{"title":"National Compute billing documentation","url":"https://docs.nationalcompute.com/billing/","type":"institution"},{"title":"National Compute Marshall data-sharing documentation","url":"https://docs.nationalcompute.com/marshall-data-sharing/","type":"institution"},{"title":"National Compute Terms of Service updated October 6 2026","url":"https://nationalcompute.com/documents/TOS.pdf","type":"institution"},{"title":"National Compute Credit Terms updated October 6 2026","url":"https://nationalcompute.com/documents/CreditTerms.pdf","type":"institution"},{"title":"National Compute Public Marshall documentation","url":"https://docs.nationalcompute.com/public-marshall/","type":"institution"},{"title":"Vultr joins National Compute as a founding consortium member","url":"https://blogs.vultr.com/vultr-joins-national-compute","type":"institution"},{"title":"AMP describes its independent AI grid mission","url":"https://www.amppublic.com","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/national-compute-research-preview-access","embedUrl":"https://brightaifuture.com/embed/story/national-compute-research-preview-access","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},{"id":"apple-full-disk-access-agent-controls","headline":"Apple plans a more deliberate boundary for agents and private data.","canonicalUrl":"https://brightaifuture.com/discoveries/apple-full-disk-access-agent-controls","datePublished":"2026-10-02","dateModified":null,"sourcePublicationDate":"2026-10-02","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["agents"],"summary":"Apple plans additional Full Disk Access controls as autonomous AI agents become more capable. The announcement leaves implementation and timing open.","evidenceState":"Emerging","keyFacts":[{"label":"AI’s role","value":"Apple identifies increasingly capable, autonomous AI agents as a reason to revisit broad permissions. This is a proposed operating-system boundary around apps and agents, not a new model or a measured safety result."},{"label":"Documented result","value":"In its October 2 announcement, Apple says Full Disk Access bypasses many usual data protections so backup apps can work. It warns that other uses can expose files, mail, messages and browsing history, including other people’s communications. Apple plans additional controls requiring deliberate user action before granting this access.\n\nThe published notice does not describe the new interface, a rollout date or a released macOS version carrying the change. Bright is reporting an announced direction."},{"label":"Important limitation","value":"Emerging: planned controls, with no implementation or rollout date specified in the checked announcement."}],"limitations":["Emerging: planned controls, with no implementation or rollout date specified in the checked announcement.","Bright has not tested a new control or observed a reduction in privacy incidents. No specific agent, vendor or incident is identified in Apple’s notice.","Questions about narrower scope, revocation and usability are Bright’s analysis, not announced Apple features."],"evidenceLinks":[{"title":"Apple Developer · Updates to Full Disk Access in macOS · October 2, 2026","url":"https://developer.apple.com/news/?id=p6zjojqw","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/apple-full-disk-access-agent-controls","embedUrl":"https://brightaifuture.com/embed/story/apple-full-disk-access-agent-controls","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},{"id":"nvidia-open-agent-safety-platform","headline":"NVIDIA adds an independent safety layer for AI agents.","canonicalUrl":"https://brightaifuture.com/discoveries/nvidia-open-agent-safety-platform","datePublished":"2026-09-28","dateModified":"2026-09-28","sourcePublicationDate":"2026-09-28","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["open-models","agents"],"summary":"AI agents can write code, use tools and keep working long after a person steps away. NVIDIA’s September 28 Open Agent Safety Platform launch puts permission controls outside the agent’s workload and offers a separate hardware-backed watchdog. The aim is to give people stronger control over what agents can reach; the launch does not establish that agents are now safe in every setting.","evidenceState":"Emerging","keyFacts":[{"label":"AI’s role","value":"The agents remain responsible for choosing steps, calling tools and generating code. OpenShell 0.1.0 supplies the surrounding runtime: a sandbox limits filesystem and process access, while a supervisor outside the workload checks outbound requests. For configured HTTP, GraphQL and MCP traffic, policies can distinguish reading from writing. Real service credentials are substituted outside the agent workload for authorized requests. A formal policy prover checks modeled permissions against an operator-defined boundary; this is a check on the permission model, not proof that an agent’s goals or decisions are harmless."},{"label":"Documented result","value":"On September 28, 2026, NVIDIA announced the Open Agent Safety Platform, pairing broadly available open-source OpenShell software with a reference system design featuring Sentry. Sentry is an optional watchdog running on BlueField-4 data processing units using NVIDIA DOCA. NVIDIA describes it as monitoring agent activity from an isolated trust domain and says it can quarantine an agent that crosses its boundary in milliseconds. Its technical walkthrough also reports adversarial tests in which agents with reduced safeguards tried for up to two hours to obtain permission to modify a protected GitHub repository. NVIDIA reports no protected writes in those tests when an AI reviewer used the policy analysis alongside runtime controls. Bright has not reproduced either result. The release says more than 100 organizations are working with the platform’s technologies; that participation is not a count of independently validated deployments."},{"label":"Important limitation","value":"This is an Emerging security-platform record based on NVIDIA’s announcement, technical explanation and documentation. The cited launch materials do not establish independently measured reductions in incidents across real-world deployments."}],"limitations":["This is an Emerging security-platform record based on NVIDIA’s announcement, technical explanation and documentation. The cited launch materials do not establish independently measured reductions in incidents across real-world deployments.","The millisecond quarantine claim is NVIDIA’s. Bright has not benchmarked response time, missed detections, false alarms or performance overhead. The reported protected-repository experiment is a bounded vendor test, not a guarantee against every escape or attack.","OpenShell is available software; Sentry is an optional hardware-backed layer in a reference design. Its described BlueField-4 architecture should not be read as a universal feature on every computer. The release says some products and features remain at different stages of availability.","Operators still choose the policy and the authority they grant. Formal verification concerns modeled permissions and assumptions, not every possible harmful action. NVIDIA describes analysis of combined permissions across multiple agents as ongoing work.","Configuration and the surrounding system matter. NVIDIA’s security guidance identifies allowed endpoints as possible data-exfiltration channels. An authorized connection or action can still be harmful; infrastructure containment does not by itself establish trustworthy decisions or safe physical behavior."],"evidenceLinks":[{"title":"NVIDIA launches Open Agent Safety Platform to secure agents from testing to deployment · September 28, 2026","url":"https://nvidianews.nvidia.com/news/open-agent-safety-platform","type":"institution"},{"title":"NVIDIA Open Agent Safety Platform: a reference for continuous in-silicon agent monitoring · September 28, 2026","url":"https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/","type":"institution"},{"title":"Add runtime controls to AI agents with NVIDIA OpenShell · September 28, 2026","url":"https://developer.nvidia.com/blog/add-runtime-controls-to-ai-agents-with-nvidia-openshell","type":"institution"},{"title":"OpenShell security best practices · NVIDIA documentation","url":"https://docs.nvidia.com/openshell/latest/security/best-practices","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/nvidia-open-agent-safety-platform","embedUrl":"https://brightaifuture.com/embed/story/nvidia-open-agent-safety-platform","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},{"id":"autodiscovery-student-challenge","headline":"The students' job was to doubt the discovery.","canonicalUrl":"https://brightaifuture.com/discoveries/autodiscovery-student-challenge","datePublished":"2026-09-19","dateModified":null,"sourcePublicationDate":"2026-09-14","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["agents"],"summary":"In a University of Washington materials course, students examined patterns surfaced by Ai2 AutoDiscovery, checked literature, and decided whether each result was a useful question, coincidence, or data flaw.","evidenceState":"Emerging","keyFacts":[{"label":"AI’s role","value":"The agent searched materials data and surfaced possible relationships for students to interrogate."},{"label":"Documented result","value":"Ai2 reports a spring-course challenge in which students checked candidate patterns against literature and data. It reports neither a validated discovery nor measured learning improvement."},{"label":"Important limitation","value":"This is one reported classroom challenge."}],"limitations":["This is one reported classroom challenge.","No controlled educational outcome is available.","The source is Ai2's account; instructor and student confirmation is still needed for flagship publication."],"evidenceLinks":[{"title":"Teaching future scientists to interrogate AI tools for scientific discovery","url":"https://allenai.org/blog/autodiscovery-student-challenge","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/autodiscovery-student-challenge","embedUrl":"https://brightaifuture.com/embed/story/autodiscovery-student-challenge","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},{"id":"duke-grid-interconnection-agents","headline":"Two weeks of grid preparation, reduced to hours.","canonicalUrl":"https://brightaifuture.com/discoveries/duke-grid-interconnection-agents","datePublished":"2026-09-19","dateModified":null,"sourcePublicationDate":"2026-09-17","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["agents"],"summary":"Duke Energy says an AWS agentic workflow reduced data preparation for interconnection studies from two weeks of manual work to hours by coordinating existing models and scripts.","evidenceState":"Deployed","keyFacts":[{"label":"AI’s role","value":"Agents orchestrated existing engineering tools and data preparation; engineers retained final decisions."},{"label":"Documented result","value":"Duke reports reducing preparation from two weeks to hours. LBNL separately reports more than 2,000 GW of generation and storage seeking interconnection at the end of 2025."},{"label":"Important limitation","value":"The time reduction is partner-reported and not independently audited."}],"limitations":["The time reduction is partner-reported and not independently audited.","Faster preparation does not establish faster approvals, lower bills, or better reliability.","The system coordinates physics tools; it does not replace engineering sign-off."],"evidenceLinks":[{"title":"AWS Launches Agentic Grid Planning Program to Accelerate Interconnection Studies","url":"https://press.aboutamazon.com/aws/2026/9/aws-launches-agentic-grid-planning-program-to-accelerate-interconnection-studies","type":"institution"},{"title":"Queued Up: Characteristics of Power Plants Seeking Transmission Interconnection","url":"https://emp.lbl.gov/queues","type":"report"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/duke-grid-interconnection-agents","embedUrl":"https://brightaifuture.com/embed/story/duke-grid-interconnection-agents","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},{"id":"nemotron-3-super","headline":"An open reasoning model with its recipe beside it.","canonicalUrl":"https://brightaifuture.com/discoveries/nemotron-3-super","datePublished":"2026-09-07","dateModified":null,"sourcePublicationDate":"2026-03-11","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":["open-models","agents"],"summary":"NVIDIA released Nemotron 3 Super, an open 120-billion-parameter mixture-of-experts reasoning model with 12 billion active parameters, plus stated releases of its methodology, data, reinforcement-learning environments, and evaluation recipes.","evidenceState":"Deployed","keyFacts":[{"label":"AI’s role","value":"A hybrid Mamba-transformer mixture-of-experts model for complex agent subtasks, with a one-million-token context window and multiple active expert specialists at inference."},{"label":"Documented result","value":"NVIDIA says the model was available on March 11, 2026 with open weights under a permissive license. It says it published methodology, more than 10 trillion pre- and post-training tokens, 15 reinforcement-learning training environments, and evaluation recipes. NVIDIA reports up to 5 times higher throughput and up to 2 times higher accuracy than the prior Nemotron Super model."},{"label":"Important limitation","value":"The throughput, accuracy, and agent-quality results are NVIDIA's claims, not independent evidence of benefit in a workplace or public service."}],"limitations":["The throughput, accuracy, and agent-quality results are NVIDIA's claims, not independent evidence of benefit in a workplace or public service.","The underlying training data includes synthetic data from frontier reasoning models; openness does not itself settle provenance, bias, or safety questions.","A one-million-token context window can retain more material, but it does not prevent mistakes, goal drift, or unsafe tool use.","Hardware requirements and operational cost can still limit who can use or customize the model."],"evidenceLinks":[{"title":"New NVIDIA Nemotron 3 Super Delivers 5x Higher Throughput for Agentic AI","url":"https://blogs.nvidia.com/blog/nemotron-3-super-agentic-ai/","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/nemotron-3-super","embedUrl":"https://brightaifuture.com/embed/story/nemotron-3-super","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}}],"documentation":"https://brightaifuture.com/wire/terms"}