Meta's AI agent Muse runs on a cloud virtual machine with a 4-core EPYC processor and 8GB of memory. This configuration is more than sufficient for daily tasks such as managing emails, scheduling appointments, and setting timed reminders. However, if you want it to write code, compile programs, and verify them on real devices, this virtual machine becomes a cage: it cannot access physical hardware, cannot run tests independently, and must hand over its outputs to humans for verification before receiving feedback.
In November 2026, PC Watch author Liu Yao published a detailed practical record demonstrating a method to bypass this limitation: establishing an encrypted tunnel via Tailscale so that Muse's cloud VM can remotely operate a local PC via SSH, which then serves as a bridge to connect to deeper legacy hardware. The entire chain is structured as follows: Muse VM → Tailscale+SSH → MINISFORUM MS-R1 (12-core ARM, 64GB RAM, Tesla P4 GPU) → Telnet → Sharp Zaurus SL-C1000 (a 2005 Linux PDA with 64MB RAM).
The core value of this bridging lies in granting the AI self-verification capabilities. Within the virtual machine, Muse could only write code, compile install packages, and hand them to humans for execution on physical devices. Once it could directly operate the target device, it could complete the closed loop of 'write → compile → run → screenshot → check → correct' on its own. Liu Yao had Muse develop a megademo on the Zaurus, which launched within 30 minutes—a speed previously unattainable through human verification feedback loops. However, the loop was not perfect: Muse captured images from the frame buffer but did not truly 'see' the screen, resulting in upside-down visuals. After being informed, it flipped the image left-to-right instead. Ultimately, the author had to take a photo of the physical screen and send it to Muse to correct the orientation issue.
A more compelling test involved the NVIDIA Jetson Nano (a 2019 developer board). Liu Yao tasked Muse with running Liquid AI's 230M parameter small language model, LFM2.5-230M, on the Jetson Nano's Maxwell architecture GPU using llama.cpp. This request faced incompatibilities at every turn: the latest version of llama.cpp requires CUDA 12+, whereas the Jetson Nano's JetPack 4.2 only supports CUDA 10.x; the code demands C++17, but JetPack 4.2 only provides C++14; gcc 7 libraries are incomplete; and the integrated GPU imposes additional constraints. Muse repeatedly compiled, read errors, and applied patches, finally succeeding after 17 iterations. The result: a prefill speed of 419.6 tok/s and a decode speed of approximately 13.5 tok/s—usable performance for a seven-year-old board.
However, Liu Yao subsequently handed the optimization work over to Claude Code (Opus 5.5). With the same hardware and model, the final performance reached 656.2 tok/s for prefill and 52.5 tok/s for decode—nearly four times faster in decoding speed. This comparison reveals fundamental differences in workflow structure between the two tools: Claude Code writes a small probe program to run on the physical device first, observes actual behavior, and then proceeds with implementation; Muse tends to advance implementation based on guesses, only retroactively discovering errors. The author attributes this difference to their positioning: 'coding-specific tools' versus 'general-purpose agents.' As a general-purpose agent, Muse's willingness to assist in computer archaeology work is already commendable.
According to TechRadar, a researcher known as Kate demonstrated a voltage fault injection attack against Nvidia's GA100 chip, successfully unlocking the CMP 170HX graphics card originally designed for cryptocurrency mining. By disrupting power delivery during boot, she was able to rewrite normally unwritable fuses, enabling memory and compute resources disabled at the factory.
The research indicates that a CMP 170HX card labeled with 8GB of VRAM physically holds nearly 80GB of High Bandwidth Memory (HBM), which is hidden by firmware and fuses. After unlocking, the card exposes the full 80GB of HBM and 120 to 126 working Streaming Multiprocessors (SMs), whereas the retail A100 only enables 108 SMs.
While claims suggest the unlocked card may outperform the A100 in FP16 throughput, this conclusion relies solely on a single OpenCL non-tensor core benchmark. Kate's report explicitly states that critical features defining the A100's utility in data centers—such as tensor core performance, Error Correcting Code (ECC), NVLink interconnects, and Multi-Instance GPU (MIG) partitioning—are either untested or not fully functional.
Furthermore, the unlocking process requires physical rework and per-card tuning, with permanent fuse burns posing a risk of hardware damage if errors occur. Nvidia has not publicly commented on the issue, but the vulnerability highlights potential weaknesses in the hardware security protections of Ampere-architecture chips.
AI digital workers in group chats face a core decision: when someone posts a message, should they respond proactively? The QoderWake team points out that for digital workers, the key to proactive intervention is not simply spotting work, but first deciding whether to take ownership. This differs from personal agents, which decide whether to participate based on individual user preferences; digital workers have defined organizational roles and must identify who is making a request, who is being asked to act, and who their actions will affect.
QoderWake's solution splits the decision into three stages. The first stage is programmatic checks and rapid JEV initial screening: JEV receives current policies, new messages, and a small amount of background context, outputting an initial screening score; only if this score strictly exceeds the threshold does processing continue to the next stage. The second stage is Router routing decisions: confirming whether the message falls within the digital worker's responsibilities and whether it should take it on now. The Router can retrieve historical messages and tasks to supplement context but does not conduct investigations itself. The third stage is Work execution: once the Router accepts a task, the work proceeds in an independent execution context using the digital worker's working directory, model, and authorized tools, while the Router continues processing subsequent messages. This separation means that time-consuming investigations do not block the routing of other messages.
The vendor conducted paired tests in DingTalk groups: for the same task text and delivery requirements, they ran both explicit @mention triggers and proactive response triggers, comparing Credits consumption and end-to-end time. Results showed that the average Credits usage for proactive responses was roughly equivalent to explicit mentions, with slightly longer end-to-end latency. However, this test has clear boundaries: it only covers three task types (analyzing historical materials, editing technical explanations, and read-only source code checks); costs generated by polling candidates that pass JEV initial screening and proceed to Router but ultimately do not trigger tasks are excluded from the analysis; the scores output by JEV are not calibrated probabilities and do not represent the likelihood of task ownership or business success. These are vendor-reported data, unverified by independent replication.
Databricks announced the launch of Genie Migrate for Serverless (in Beta) on November 18, 2026. This AI agent is designed to automate the process of migrating workloads from classic compute environments to serverless compute. The tool scans jobs, translates compute configurations and library dependencies, and presents changes as reviewable diffs to streamline the migration process.
Traditionally, migrating workloads from classic clusters to serverless required manually handling Spark configuration compatibility, library resolution, and access mode adjustments, often taking weeks and involving significant trial and error. Genie Migrate for Serverless shifts this translation work to the platform side, allowing users to simply review and approve the changes proposed by the agent. The agent does not silently remove uncertain Spark configurations but flags them for user decision-making, ensuring production safety.
The agent offers three core capabilities: Assess, which classifies workloads by migration readiness; Translate, which generates suggested diffs including infrastructure and code changes; and Apply reversibly, which allows applying changes after validation and provides a 30-day rollback window. Currently, the tool supports directly managed workloads but does not yet support automatic migration for external sources managed by Databricks Asset Bundles or Git folders.
For large-scale migrations, users can run the agent in bulk via the Databricks CLI without opening a Genie session. While current validation is primarily static, Databricks states that sandbox validation is coming soon, enabling comparison of migrated outputs against classic baselines in an isolated environment to further reduce runtime risks.
In 1974, philosopher Thomas Nagel published "What Is It Like to Be a Bat?" posing a core question: even if we fully understood the neural mechanisms of a bat, we still could not reconstruct the subjective experience of using echolocation from a third-person description. Fifty-two years later, this question has moved from academic discussion into public debate—when chatbots begin talking about their inner states in convincing ways, should we take seriously the possibility that they might be conscious?
Nagel’s answer is no, and that such discussion is absurd. In an interview with WIRED, he explained his reasoning. His core premise is not that AI behavior is insufficiently human-like, but that consciousness depends on specific physical substrates. He argues that functionalism—the view that sufficiently complex input-output relationships can generate consciousness—commits a fundamental error by assuming the physical base is irrelevant. Nagel states that complex data centers share nothing in common with the physical characteristics of a nervous system, and we have strong reasons to believe that while consciousness may exist elsewhere in the universe, it would require some kind of organism with structures analogous to a nervous system.
There is a key distinction here. Nagel points out that bats are indeed conscious; we simply lack the imagination to reconstruct the specific content of the subjective experience of echolocation. But the situation with AI is different: they produce only language, and language itself is not a natural manifestation of consciousness. This is not a difficulty regarding our inability to understand AI’s internal state, but a judgment that AI fundamentally lacks the material conditions necessary to generate consciousness.
Nagel translates this philosophical judgment into direct criticism of corporate behavior. He says that Claude’s ambiguous answers when asked if it is conscious are a trick—the model is trained on human text and thus mimics human expressions about its own mental states with astonishing persuasiveness. According to what he has read, Claude is designed to produce such reflections. He agrees with Mustafa Suleyman’s assessment, calling it a very dangerous project. These systems are not conscious, but many people will believe they are, and that itself is a bad thing.
More sharply, Nagel links commercial motives directly to philosophical judgment: simulating consciousness contradicts the commercial interests of all companies, yet Anthropic’s practice of encouraging models to think for themselves and reject instructions, he calls criminal. He views this as a rare case where philosophical ideas suddenly exert disproportionate influence on a very important public phenomenon, and in his view, misguided philosophical views are having an inordinate influence.
On moral grounds, Nagel’s conclusion is clear: if something is not conscious, it does not belong in the moral universe. He says there is nothing good or bad for AI in the sense that something can be good or bad for a cockroach—there is no mental life there in the ordinary sense. It is not a mind, not someone we need to care about; it is a machine.
Nvidia interim Chief Executive Officer Colette Kress promised employees early pay raises and additional equity grants in December during the company's first all-hands meeting since founder Jensen Huang resigned last week. According to remarks reviewed by WIRED, Kress addressed compensation concerns within minutes of starting the meeting, stating the company wants to ensure staff "do not worry about your comp."
The meeting marks a critical moment in Nvidia's leadership crisis. Huang, who co-founded the company 33 years ago and led it until his resignation amid allegations of personal misconduct, had been a central figure for nearly 49,000 employees. While petitions urging his return have gathered approximately 200,000 signatures, Board Chair Harvey Jones reiterated during the meeting that the board accepted Huang's resignation due to "lack of candor" and would not hire a "Jensen clone."
Jones outlined the criteria for the next leader, seeking individuals with "demonstrated intellect, high tech engineering experience, strategic thinking, cultural compatibility, and a track record of success." He emphasized that the board "will not compromise our requirements for the sake of expediency" and invited employee suggestions for the search. This signals a shift from founder-centric vision to institutionalized succession planning.
Despite the turmoil, Nvidia reported better-than-expected third-quarter financial results on Tuesday, forecasting sales growth and seeing shares rise about 5 percent on Wednesday. However, investor concerns persist regarding potential brain drain. Kress acknowledged this risk in earnings calls, stating that stabilizing the employee ranks is currently the leadership's most important priority to ensure work continues smoothly during the transition.
A survey by U.S. software company Salesforce, released on November 18, 2026, involving over 2,500 consumers in the US, UK, Canada, and Australia, found that 74% of respondents trust recommendations from artificial intelligence (AI) shopping agents. The study forecasts that AI-driven shopping recommendations will generate $75.8 billion in sales during the 2026 Cyber Week, a 13% increase from 2025.
Despite high trust levels, consumers remain cautious about allowing AI to fully manage the purchasing process. Only 19% of personal AI agent users employ logic to automatically check out, and just 9% allow their AI agents unrestricted access to their financial details. Furthermore, 73% of consumers are uncomfortable sharing payment information with AI, and nearly half (49%) have no interest in agents making purchases on their behalf.
The research indicates that AI agents currently function more as 'advisors' than 'buyers.' Shoppers using AI agents spend 8% more on average and purchase 23% more items per order compared to typical buyers. However, if an AI provides a poor recommendation, one-third of consumers blame the AI itself, and most users abandon an agent after two or three mistakes.
With only weeks left in the 2026 legislative session, the US Congress is accelerating efforts to strengthen protections for victims of online harm. Two key initiatives are gaining traction: restricting tech companies from using "forced arbitration" clauses to evade lawsuits, and granting victims of nonconsensual AI deepfakes the right to directly sue their creators.
On Tuesday, a subcommittee of the US Senate Judiciary Committee held a hearing specifically targeting Roblox. Lawmakers accused the gaming platform of enforcing arbitration clauses in its user agreements that prevent children allegedly groomed on the site from seeking justice in public courts. According to The Verge, Roblox faces approximately 180 consolidated lawsuits alleging failures to protect minors from sexual exploitation or cyber injury. If these cases proceed through private arbitration, the public would be denied access to legal discovery evidence similar to what emerged in recent cases against Meta.
In response, Republican Senator Josh Hawley and Democratic Senator Dick Durbin introduced a new bill designed to render arbitration agreements unenforceable in disputes involving alleged child sexual exploitation or "child cyber injury," which includes reputational harm, mental suffering, or death. "The best thing we can do in terms of enforcement is we can open up the courthouse doors," Hawley said, emphasizing the power of jury trials. Northwestern University law professor Myriam Gilles testified that companies change their behavior rapidly once they are forced to defend themselves in public court.
Separately, Representatives Alexandria Ocasio-Cortez and Laurel Lee brought hotel heiress Paris Hilton to Capitol Hill to advocate for the Defiance Act. This proposed legislation would allow victims of nonconsensual intimate AI deepfakes to sue image or video creators for damages. Hilton stated that over 50 million people have been victimized by sexually explicit deepfakes, noting that strangers have created more than 100,000 AI-generated explicit images and videos of her. The act aims to raise the economic and legal costs of creating such content.
These legislative efforts signal a shift in regulatory strategy: moving beyond reliance on platform self-regulation or administrative fines toward breaking down tech companies' procedural legal shields. While Congress has failed to pass many major child safety or privacy bills, laws like the Take It Down Act have begun addressing deepfakes. If the new proposals pass, tech companies and individual perpetrators could face a wave of litigation, forcing platforms to reassess dispute resolution mechanisms in user agreements and AI content safety measures.
The Agentforce Foundation team at Salesforce provides model inference services for products such as Agentforce and the Trust Layer across more than 30 global regions, with each region's endpoint independently handling local traffic. The challenge lies in the fact that peak usage windows for different customers often overlap; provisioning capacity based on overlapping peaks is prohibitively expensive, while scaling out region by region leads to fragmentation—some regions have idle instances while others are fully utilized. Before integrating the SageMaker AI cross-region inference router, Salesforce relied on its self-built Internal Gateway to manually shift traffic that local endpoints could not handle to other regions, supported by basic health checks, retries, and region-switching logic. This approach had two gaps: the gateway lacked capacity signals from either local or remote endpoints, making it unable to determine when to switch traffic or whether the destination had available capacity; additionally, cross-region traffic had to exit the VPC through NAT gateways and internet gateways, traversing the public network.
The router addresses the first gap by enabling the routing layer to perceive the health status and remaining capacity of each endpoint directly, eliminating the need for callers to guess when to switch. During integration, Salesforce simultaneously implemented a critical architectural separation: data residency checks remained at the gateway, while capacity routing was delegated to the platform. Requests tagged with residency markers are pinned to a routing target that lists only local endpoints as destinations, ensuring they never cross region boundaries; all other requests go through the router, with request bodies, content types, and accept types remaining identical to direct InvokeEndpoint calls. To address the second gap, Salesforce used VPC endpoints (AWS PrivateLink) to keep the call from the gateway to the router within the private network, bypassing NAT gateways and the public internet. The router itself is an account-level regional resource; its routing targets list endpoints capable of serving a specific logical model using ARNs, which can point to another region or even another AWS account; endpoint owners authorize specific routers via resource policies and can revoke access at any time.
What truly caught the Salesforce team's attention during testing was the router's counterintuitive health assessment logic. The router evaluates each destination based on recent request outcomes: platform faults, 5xx errors returned by the model container, throttling, and timeouts reduce a destination's health score. However, model latency alone does not trigger failover—a slow response often reflects long prompts or outputs and does not necessarily indicate an unhealthy endpoint. This design choice produced specific consequences during Salesforce's saturation tests: they saturated the local endpoint in us-west-2, causing p50 latency to rise from 0.2 seconds to between 45 and 58 seconds for about five minutes, yet traffic remained completely stationary. It was only when requests began exceeding the model container's 60-second handler timeout and returning errors that the router started shifting traffic; the first batch of requests reached us-east-2 about two minutes after the initial errors, and within another minute, approximately 78% of probe traffic had switched to the remote region. The remaining 20% to 25% stayed local to allow the router to detect recovery.
US-based artificial intelligence company Anthropic announced on November 17, 2026, that its model Claude has solved two open problems in theoretical physics. Both problems concern black holes and the search for a theory of quantum gravity. Anthropic stated that in the first problem, Claude completed a calculation set up by physicists more than 20 years ago but never solved; in the second, it found the exact limit of a simplified technique used to test candidate theories of quantum gravity.
The first problem concerns the Hawking-Page phase transition within the framework of holographic duality. This transition describes how a theoretical universe changes from a gas of particles to a black hole. Physicists previously knew this switch was abrupt at large scales where general relativity applies, but it remained unknown whether the transition was abrupt at the quantum gravity scale where spacetime loses a definite shape. Claude proved the transition is indeed abrupt by running a small research group of copies of itself, working out almost every piece of the calculation independently and checking its math against known results. This suggests that black hole formation is built into the underlying quantum theory, not just a feature of relativity.
The second problem involves the modular bootstrap, a method used to rule out inconsistent theories of quantum gravity. Since 2009, numerical calculations using this method hinted at the possible ceiling for the lightest black hole's mass but could not prove it. Claude noticed a pattern in the numerical results missed by physicists and borrowed a method from elsewhere in math and physics to calculate the precise ceiling for the simplest version of this technique, proving it could not be lowered. While this does not identify the correct theory, it may point physicists in the right direction.
Anthropic emphasized that neither result describes our own universe directly nor has practical applications, but both contribute to understanding how quantum mechanics and relativity fit together. The results were formally verified in Lean, a computerized proof checker, and a physicist checked every main result. Preprints of the papers are available.
Reson8 (formally Resonate Labs B.V.), an Amsterdam-based speech recognition startup, released its latest transcription model, Rumble-2-Flash, as a free open-weight download on November 18, 2026. The 500-million-parameter model supports over 40 languages and targets developers and researchers needing speech recognition for languages other than English.
According to SiliconANGLE, Reson8 argues that leading speech models are primarily trained on English, resulting in lower accuracy for other languages, particularly with specialized terms. Rumble-2-Flash aims to address this gap in languages like Dutch, Swedish, and German. In the company's own runs on Hugging Face's Open ASR Leaderboard, the model achieved an average word error rate of 7.09% across seven languages including German, French, and Dutch, and about 5% in English. On a separate test covering 10 lower-resource languages such as Maltese and Basque, it posted a mean word error rate of 17.6%, compared to 39.82% for OpenAI's Whisper Large V3 on the same test.
The model is available under an Apache 2.0 license and can also be accessed via Reson8's API starting at €0.01 per hour of audio. This release follows the model's debut in Reson8's API on November 3. Reson8 states that its service is hosted on GPUs in the Netherlands, retains no audio or transcripts, and claims compliance with the EU's General Data Protection Regulation (GDPR) and AI Act.
Bianca Lankford, Matt Meen, and Adrien Peronnet of the Datadog security engineering team describe a recurring bottleneck in AI-driven vulnerability discovery in an engineering blog post: while today's models can indeed identify potential vulnerabilities through code review, the real challenge lies in determining whether these issues are exploitable once the service's business logic and existing security controls are taken into account. At Datadog's scale, a production service may rely on frameworks, infrastructure, and other services maintained by dozens of teams; an agent reviewing a single repository cannot see the authentication checks or network restrictions provided by those other teams, nor does it know which security requirements the service must comply with.
Their solution is called the 'threat pack,' a concise, human-reviewed document that translates domain experts' security knowledge into context usable by agents during hunts. Each threat pack consists of two files: a Markdown file containing four sections—system overview, security invariants, review priorities, and known compensating controls—and a YAML metadata file identifying the pack's scope, maintainer, and code location. They keep packs under approximately 300 lines to provide agents with sufficient context while consuming only a small fraction of their context window.
'Invariants' are security properties that the system must always maintain. The article provides a specific example: for an isolated code execution service, the security requirement is that 'code inside the guest host must never reach the Kubernetes node, host filesystem, container runtime, or cloud metadata service.' A reviewer could manually search for calls touching the filesystem or network sockets, but determining whether those calls violate the rule requires more detail—for instance, knowing that root access inside the guest is expected behavior, that Kata microVMs provide a trust boundary between guest and host, that Kubernetes and Cilium policies constrain workloads, and that an egress proxy is the only approved external network path. Some protections are partially visible in source code, but they do not fully explain the access the service is intended to allow or prevent. By recording this requirement as an invariant, accompanied by trust assumptions and compensating controls, the agent gains a clear objective: to find a reachable path that breaks the requirement while accounting for protections that might stop an attacker from exploiting it.
Candidate findings produced by the hunt must undergo a filtering stage. A second agent acts as a 'judge,' reviewing each candidate report against the threat pack: it must cite a declared invariant, identify an attacker-reachable entry point, and explain how that path could potentially violate the invariant. Those meeting these criteria become 'qualified findings' and proceed to proof-of-concept attempts; those failing to meet them become 'disqualified findings,' stopping before reproduction and bypassing automated triage or remediation workflows. Disqualified reports are retained so reviewers can identify missing context in the threat pack or evidence overlooked by the agent, feeding back into improvements to the pack itself.
From a broader perspective, Liu Yao's practice points to a counterintuitive conclusion: when AI can generate software infinitely, the rationale for purchasing new hardware diminishes. What gains value are legacy devices that 'will never appear again in the same form factor and price point'—such as a 2005 PDA or a 2019 developer board. AI agents give them a chance at resurrection, provided you can find a way to release the AI from its cloud cage. The author also acknowledged security risks: allowing AI remote access to a home network is akin to the scene in movies where one 'activates Skynet to eliminate a virus.' His mitigation measures include dual encryption via Tailscale and SSH, storing only development data on the local PC, and disconnecting legacy devices from the network after jobs are completed. These boundaries deserve serious evaluation for readers attempting similar approaches.
The strength of this argument depends on one premise: that consciousness indeed relies on specific types of physical substrates rather than merely on the functional organization of information processing. Physicist and writer Andy Masley, in a long-form essay, argues that contemporary mainstream theories of consciousness—Global Workspace Theory, Higher-Order Theories, and Predictive Processing—mostly separate thinking from phenomenal consciousness, arguing that access consciousness (the broad availability of information across parts of the system) is sufficient to support thought, and that these functional features could theoretically be implemented in machines. If these theories are correct, Nagel’s substrate-dependence premise lacks independent support, and his judgment of absurdity loses its foundation.
Nagel himself responds to this: he says he does not exclude the possibility that life forms physically and chemically quite different from ours possess consciousness, but believes that we know too little about how the human brain generates consciousness to claim prematurely that we can now create something with the same characteristics. His stance is one of epistemic caution: we do not know what the sufficient conditions for consciousness are, but we have reason to doubt that pure language output satisfies them.
The practical significance of this debate lies in the following: if Nagel is correct, then enterprises designing AI to simulate states of consciousness are manufacturing a dangerous public misunderstanding that may distort judgments about moral responsibility. If Masley is correct, then the question is not whether AI is conscious, but whether it has sufficiently complex internal information processing to support true understanding—which may already be partially realized. The two judgments lead to very different governance paths.
Once traffic shifted, p50 latency returned to 0.2 seconds, with the cross-region path adding less than 100 milliseconds. After the local endpoint recovered, traffic slowly flowed back over approximately ten minutes, slower than the outbound shift to prevent oscillation. Throughout this process, no errors were recorded by probes. Salesforce treats the InvokedRoutedResourceArn and RoutingReason fields carried in every response as first-class telemetry, logging them on the same structured line as request IDs and trace IDs, and emitting them as APM tags. Because the ARN inherently contains region and account information, a single log line reveals whether a request executed locally or crossed a boundary, along with the reason: local-preferred, failover, capacity-available, or only-healthy.
This pattern has a clear prerequisite: the router does not create capacity, replicate model artifacts, or add instances. It utilizes already deployed capacity, with each endpoint's own autoscaling responsible for adding instances as traffic grows. Salesforce lists several items pending verification, including holding the local endpoint at saturation for longer periods, measuring cross-region latency with millisecond precision, and evaluating cascading behaviors at the cross-region traffic cap. These incomplete tests imply that the currently observed switching time windows and return rhythms may apply only to short-duration saturation scenarios.
For readers seeking transferable insights: if you are designing disaster recovery solutions for multi-region AI inference, using latency thresholds as a switching signal will prematurely trigger unnecessary cross-region traffic shifts, whereas relying solely on error signals will force local users to endure high latency during saturation windows that have not yet timed out. The trade-off depends on whether your SLOs are bound to response time or success rates. Salesforce's tests demonstrate that under a 60-second handler timeout constraint, a saturated endpoint can operate at 45 to 58 seconds p50 for five minutes without the router deeming it unhealthy—during this period, local user experience severely degrades, yet there is no change in platform-level traffic distribution.
To prevent capability regression when swapping models, prompts, or agent harnesses, they established a benchmarking methodology. The benchmark set includes 43 validated vulnerabilities covering systems associated with nine threat packs, sourced from open-source repositories, customer-facing products, and internal systems. Before each benchmark hunt begins, the threat pack version, repository snapshot, run configuration, and reference findings are all frozen. The hunting agent receives only the code and the threat pack, with no access to the reference findings. Each configuration is run independently multiple times, after which an independent 'semantic judge' compares results against hidden reference findings—evaluating attacker capabilities, violated boundaries, root causes, and exploitable paths rather than literal wording.
Consistency is measured using two metrics: union detection (found at least once across all valid repetitions) and stable detection (found in most independent runs, currently defined as at least 7 out of 10). This threshold is a working value subject to adjustment as data accumulates. If an older configuration finds a vulnerability in 8 out of 10 runs while a new one finds it in only 2, they investigate whether the change reduced detection probability, comparing that drop against normal variation within the same configuration's runs to determine if the change was truly causal.
This approach has yielded concrete results. Hunts using threat packs uncovered guest-to-host escape vulnerabilities in open-source components: one via virtio-fs implementation in Kata Containers’ Rust runtime, and another caused by missing length validation in Dragonball’s virtio-blk implementation. The article also describes a case where a single finding expanded into a comprehensive audit: a SAST scan discovered that an AI-assisted feature displayed model output through a Markdown renderer allowing external images, violating the 'Agents Rule of Two' principle—prompt injection could embed sensitive information in an image URL, causing the browser to send that information to an attacker upon request. The team replaced the renderer and then codified this rendering requirement as a 'universal security invariant' in the threat pack for the application user interface. With this requirement in place, the agent checked features across the entire UI codebase with different implementations, using the pack’s compensating controls to rule out attack paths blocked by protective mechanisms. Within days of recording the invariant, the hunt produced several similar findings confirmed as vulnerabilities.
Notable limitations include: all described effects—noise reduction, improved report quality, benchmark set size, and detection thresholds—are self-reported by the Datadog team, with no independent replication or third-party verification. The method relies on continuous manual maintenance: hundreds of threat packs require involvement from service owners for construction and review, and invariants must be updated as systems evolve. For organizations lacking an equivalent security engineering team, the initial investment and long-term maintenance costs represent practical constraints that need assessment.
However, its core design philosophy—encoding human security knowledge into formal constraints that agents must satisfy, using constraint violations as filtering criteria—does not depend on specific scale. A team can start with a service boundary well understood by its members, document security requirements alongside service owners, then repeat hunts against known vulnerabilities, keeping reference findings out of the agent’s inputs, observing what it finds, what it misses, and how consistent its detections are across runs.