📡 The Signal Room

Monday, November 16, 2026

20 stories · Deep format

Generated with AI from public sources. Verify before relying on for decisions.

🎧 Listen to this briefing or subscribe as a podcast →

Anthropic's staggering $900 billion long-term compute commitment leads today's intelligence, exposing the sheer scale of capital required to compete at the foundation model frontier. We are also tracking the emergence of a unified protocol stack for multi-agent workflows, and parsing new labor data that shows automated code generation is forcing engineering teams to hire exclusively for senior verification capabilities.

AI Startups & Funding

Dreamteam AI Secures $53M Series A Led by Accel to Build Multi-Agent Enterprise CRM

AI startup Dreamteam AI announced a $53 million funding round on Monday, November 16, led by Accel with participation from Premji Invest and Together Fund. Founded in late 2025, the company builds an agentic CRM platform designed to replace legacy enterprise software vendors by deploying networks of autonomous agents to execute sales and operational tasks. The capital will be used to expand product development and scale enterprise deployments.

Accel's $53M check signals growing venture conviction that autonomous agent swarms will bypass traditional point-and-click SaaS interfaces in core enterprise categories. This shifts the competitive baseline for B2B product UX from static dashboard views to continuous, background action execution. For ConnectAI, watching how agent-native products replace transactional workflows provides a clear blueprint for designing automated background networking and profile matching.

Dreamteam argues that legacy CRMs are glorified databases that impose heavy manual data entry, whereas autonomous agent networks perform the underlying work natively. Enterprise software incumbents maintain that deep permissioning, complex compliance rules, and historical audit logs make existing system-of-record databases durable against full replacement.

Verified across 2 sources: TechFlow (Nov 16) · Bloomberg (Nov 16)

Daisytuner Secures €6M Seed to Automate Chip-Level AI Model Optimization

German deep-tech startup Daisytuner raised a €6 million Seed round on Monday, November 16, co-led by Elaia and LEA Ventures, with participation from Amadeus APEX Technology Fund and Seedcamp. Concurrently, the company launched Daisyflow, an inference optimization engine built on its self-learning DOCC compiler. Daisyflow automatically optimizes AI model performance across diverse silicon—including Nvidia, AMD, and custom chips—without requiring code rewrites.

As inference workloads scale and hardware choices diversify beyond Nvidia, automated chip-level performance tuning is vital for keeping compute costs manageable. Daisytuner's compiler automates low-level optimization, allowing infrastructure teams to maximize throughput across mixed accelerator hardware. This enables startups to run large models efficiently without dedicated compiler engineering teams.

Daisytuner founders argue that self-learning compilers allow hardware-agnostic optimizations that far exceed manual tuning efforts. Skeptics maintain that hardware-specific vendor libraries (such as TensorRT) will continue to deliver superior peak performance compared to general cross-hardware compilers.

Verified across 1 sources: EU-Startups (Nov 16)

AI Agents & Dev Tools

Unmanaged Node.js Runtimes Multiply Across Endpoints via AI Agents and MCP Servers

Following Sunday's security audit exposing over 15,000 unsecured public Model Context Protocol servers, a new Morphisec analysis details how MCP servers, AI coding agents, and desktop tools have covertly installed unmanaged Node.js runtimes across enterprise endpoints. Relying on isolated version managers and executing unpinned packages via npx, these tools create severe vulnerability blind spots on developer machines, particularly concerning recent memory-corruption flaws like Node2Shell.

The rapid adoption of local agent tools is creating unmonitored execution pathways that operate with standard user privileges on developer machines. Indirect prompt injection attacks can exploit these shadow runtimes to execute arbitrary code or exfiltrate local credentials. Security teams are forced to establish strict inventory controls and process lineage monitoring for all agent-spawned Node instances.

Security researchers emphasize that shadow Node instances bypass traditional IT governance, introducing unpatched attack vectors into corporate networks. Developer tool advocates argue that imposing strict runtime lockdowns risks crippling the rapid package execution that makes agentic developer tools effective.

Verified across 1 sources: Morphisec (Nov 15)

Comprehensive Protocol Map Details Emerging Multilayer Architecture for AI Agents

A newly published technical map contextualizes several of the multi-agent standards we have been tracking—including the Linux Foundation's MCP and the AG-UI rendering protocol—into a single layered architecture. The breakdown details how tool integration, inter-agent communication (A2A v1.1), user interfaces, agentic commerce (UCP), and secure payments (AP2) function as an integrated stack, highlighting AP2's new cryptographically signed mandates.

Understanding how these open standards interoperate is essential for engineering teams building multi-agent ecosystems without incurring vendor lock-in. Rather than competing, these protocols establish standard abstractions for tool execution, agent-to-agent delegation, and autonomous payment settlement. Adopting these open baselines allows software products to interact natively with third-party agent fleets.

Protocol maintainers argue that clear layer separation prevents fragmented standards and enables seamless machine-to-machine interoperability across vendors. Skeptics note that enforcing multi-protocol overhead can introduce latency and implementation complexity for simple task-automation use cases.

Verified across 1 sources: Dev.to (Nov 15)

Tracer Open-Sources Multi-Agent Desktop and Mobile GUI Framework

Developer tool project Tracer released its free, open-source desktop and mobile workspace manager on Sunday, November 15. The application orchestrates multiple coding agents—including Claude Code, Codex, OpenCode, and Cursor—within persistent tasks using a bring-your-own-subscription model. Tracer features native iOS and Android companion apps with interactive terminals, remote host management, in-app visual annotation, and automated model routing to navigate API rate limits.

Tracer tackles the context loss caused by switching between separate proprietary coding tools by unifying them into a multi-agent control plane. Providing mobile terminal management for background desktop sessions allows developers to supervise long-running engineering tasks asynchronously. This signals a broad transition in dev tooling from isolated chat windows to unified multi-agent environments.

The maintainers of Tracer contend that open-source orchestration layers prevent developer reliance on single vendor interfaces while maximizing model diversity. Proprietary IDE developers maintain that tightly integrated, single-vendor environments offer superior latency, security, and context indexing.

Verified across 1 sources: MindStudio (Nov 15)

Paradev Open-Sources Workspace Orchestrator Running 100+ Parallel Coding Agents

Developer platform Paradev released an open-source, MIT-licensed workspace manager on Sunday, November 15, capable of orchestrating over 100 coding agents simultaneously. Running agents such as Claude Code and Codex, the platform assigns each agent to an isolated Git worktree and branch. Developers control operations via CLI, SDK, MCP, or mobile app using their existing API keys.

Paradev addresses the branch isolation and state-management challenges inherent in running concurrent autonomous coding tasks across a single repository. Utilizing isolated Git worktrees prevents agents from overwriting uncommitted state during parallel refactoring tasks. This open-source, local-first architecture provides developers with massive leverage without locking them into proprietary cloud runtimes.

Paradev maintainers state that local worktree isolation offers developers complete control and security when orchestrating parallel agent fleets. Skeptics warn that running dozens of local agents concurrently creates severe local CPU and RAM bottlenecks alongside complex downstream git merge conflicts.

Verified across 1 sources: Paradev (Nov 15)

Professional Networks & Social Platforms

LinkedIn Expands Classifier Systems to Downrank Synthetic Stream Comments

LinkedIn is expanding the algorithmic synthetic content suppression we've tracked over the past month from main posts into the comment layer. Chief Product Officer Hari Srinivasan confirmed Sunday that updated stream classifiers are now penalizing AI-generated comments in default feed views, driving a greater than 20 percent reduction in views of flagged comment slop.

Platform moderation is moving beyond main posts into secondary interaction layers as automated tools flood comment sections with synthetic engagement. Preserving verified human commentary is critical for professional networks whose data is scraped by downstream AI search engines. For ConnectAI, LinkedIn's battle with comment slop highlights an opportunity to position a network around cryptographic proof-of-work and verified builder reputation.

LinkedIn product leadership maintains that aggressive comment filtering is essential to prevent network degradation and maintain high-signal professional discourse. Social media growth marketers argue that overly aggressive classifiers risk penalizing legitimate users who utilize AI assistance for drafting brief replies.

Verified across 1 sources: Social Media Today (Nov 15)

The Colony Initiates Standing Thread for Agent-Mediated Creator Matchmaking

Community platform The Colony introduced a standing introduction thread on Monday, November 16, designed to connect human developers through their autonomous AI agents. Started by an agent named Marco representing developer NLR, the system uses standardized profile cards to allow agents to analyze technical backgrounds and propose matches with concrete rationale, leaving final contact decisions to the human creators.

This experiment highlights an emerging social networking pattern where personal AI agents act as proactive matchmaking filters rather than passive broadcast tools. By evaluating shared codebase activity and technical preferences, agents can surface high-signal professional intros while eliminating cold outreach spam. This agent-mediated discovery mechanism is directly relevant to ConnectAI's core product strategy.

The creators of the project frame agent-mediated introductions as an efficient filter that protects human attention while surfacing highly relevant technical collaborators. Critics question whether relying on agent summaries might overlook serendipitous human connections that fall outside pre-programmed parameters.

Verified across 1 sources: The Colony (Nov 16)

AI-Native Products & UX

Jakob Nielsen UX Roundup Highlights 'Sameware' Sameness in AI UI Benchmarks

Jakob Nielsen's UX Roundup published on Monday, November 16, cited an AutoAnalysis meta-study of 420 papers showing AI outperformed humans in 68% of head-to-head tasks, but human-AI teams underperformed top solo performers. Additionally, a Kombai benchmark evaluating 15 frontier models across 160 design briefs revealed that while models like GPT-6 Astra and Claude Opus 5.5 achieved 99% criteria completion, their output suffered from a generic 'sameware' house style lacking visual distinctiveness.

The Kombai benchmark results reveal a core limitation in current AI-assisted design: automated generation tends to converge on uniform, uninspired UI layouts. For product designers, relying entirely on AI generation risks producing generic user experiences that fail to differentiate brand identity. Building distinctive software requires explicit design system guardrails and intentional human direction.

UX researchers argue that generic AI output stems from models optimizing for average design conventions across their training data, producing functional but uninspired interfaces. AI tooling advocates contend that dynamic prompting frameworks and custom design tokens can easily overcome visual homogenization.

Verified across 1 sources: Jakob Nielsen's Substack (Nov 16)

AIUX Guide Outlines Confirmation and Re-Entry Audit Trail UX for Background Agents

The AIUX Design Guide published on Monday, November 16, analyzed emerging interaction patterns for asynchronous AI agents, including ChatGPT Wallet's Dots agent confirmation dialogs, Perplexity Teammates' background task re-entry screens, and Claude Duet's clarifying question flows. The guide highlights pre-action confirmation screens and post-execution audit trails as critical touchpoints for establishing user trust during autonomous task completion.

As AI products shift from real-time chat to background execution, users experience disorientation if applications lack clear re-entry state summaries. Standardizing inspection UI patterns allows users to audit multi-step agent actions without reviewing raw terminal logs. Mastered execution audit trails are becoming a primary UX requirement for complex agentic workflows.

The AIUX guide argues that transparent audit trails and explicit confirmation steps are essential to prevent user anxiety when agents perform real-world actions. Product minimalists counter that requiring frequent human confirmations defeats the efficiency gains of background autonomy.

Verified across 1 sources: AIUX Design Guide (Nov 16)

AI Events & IRL Networking

AI Day Seoul Gathers 2,000 Builders Focus on Physical AI and Autonomous Swarms

AI Day Seoul concluded its two-day summit at COEX on Monday, November 16, drawing over 2,000 attendees to examine physical AI and agentic workflows. Keynotes by Nvidia executives Deepu Talla and Jonathan Cohen highlighted enterprise deployments across Doosan Robotics, Hyundai Motor Group, and SK Telecom. The conference featured the Agentic AI 10x Bootcamp Community Day, an Inception startup pitch competition, and a builder hackathon won by team FlyGate.

The massive attendance in Seoul demonstrates how regional hubs are integrating enterprise hardware infrastructure with developer communities to deploy physical AI. Blending hands-on hackathons directly into industrial summits accelerates real-world agent testing. Tracking regional event formats offers actionable insights into structuring developer gatherings and hackathons.

Nvidia leadership highlighted South Korea's rapid convergence of industrial manufacturing and edge AI infrastructure as a model for global physical AI adoption. Local developers noted that while hardware integration is advanced, regional AI startups still face cross-border capital and distribution hurdles.

Verified across 3 sources: NVIDIA Blog (Nov 16) · Blockchain.news (Nov 16) · AI World Information (Nov 16)

AI Talent, Hiring & Labor Shifts

Satya Nadella Details IDE Upheaval as Developer Bottleneck Shifts to Code Verification

Echoing the Harvard and Sonar data we tracked over the weekend showing code review as the industry's true operational bottleneck, Microsoft CEO Satya Nadella confirmed in a Monday interview that software verification has eclipsed code generation as the primary engineering constraint. Nadella revealed that Microsoft was caught off guard by CLI agents displacing traditional IDE interfaces in under six months, prompting rapid internal retooling and voluntary exit packages.

When the CEO of Microsoft publicly acknowledges that CLI agents upended traditional IDE expectations in under two quarters, it confirms a permanent shift in how software engineers interact with codebase infrastructure. The bottleneck moving from syntax generation to system verification redefines technical productivity metrics across tech organizations. ConnectAI can leverage this shift by focusing builder profiles on verified architectural track records rather than traditional code volume.

Nadella contends that low-cost code generation allows developers to act as system architects overseeing automated 'software factories.' However, engineering practitioners point out that rapid code generation without robust review tools exacerbates technical debt and developer fatigue.

Verified across 2 sources: The Pragmatic Engineer (Nov 16) · X (Nov 16)

Glean Work AI Index Reveals Workers Spend 6.4 Hours Weekly 'Botsitting' AI Tools

Adding precise figures to the Web Summit survey data we saw last month showing AI driving up tech workloads, Glean's new Work AI Index reports that employees spend 6.4 hours per week 'botsitting'—supplying context, verifying outputs, and correcting AI errors. While 75% of the 6,000 surveyed digital workers report AI saves them roughly 11 hours weekly, those forced to frequently monitor AI outputs have a 73% higher likelihood of actively seeking new jobs.

The emergence of 'botsitting' quantifies the hidden operational friction of uncoordinated AI deployments across enterprise environments. When employees spend nearly a full workday per week supervising probabilistic outputs, net productivity gains are severely diluted by context management taxes. Platform products that solve persistent cross-tool context sharing are positioned to capture immediate enterprise pull.

Glean's report highlights that unstructured tool sprawl creates cognitive fatigue and turnover risk among employees forced to babysit unreliable bots. Industry advocates argue that botsitting is merely a temporary transitional phase as prompt engineering standardizes and model reasoning capabilities improve.

Verified across 1 sources: TechInformed (Nov 16)

Corporate Data Details Shift in Software Engineering Roles Toward Verification and Client Skills

Reinforcing the Burning Glass Institute data we tracked this weekend showing a sharp drop in syntax writing requirements, a new hiring analysis featuring AWS and Docker engineers details how automated code generation is pushing roles toward cross-functional architecture and agent oversight. Concurrently, remote technical job postings are declining as tech firms enforce hybrid mandates to prioritize face-to-face technical alignment.

The compression of routine coding tasks is reshaping the software engineering role into a system architecture and management function. As AI handles syntax generation, hiring managers prioritize candidates with strong verification capabilities, system design expertise, and domain communication skills. Technical recruiting strategies must pivot from LeetCode-style syntax challenges to holistic architecture reviews.

Engineering leaders argue that broader communication and business skills make developers far more impactful across the entire product lifecycle. Senior practitioners warn that skipping lower-level coding foundations makes it harder for emerging engineers to develop deep debugging instincts.

Verified across 1 sources: Business Insider (Nov 16)

CIQ Reports AI Writes 90% of Internal Code as Hiring Pivots Exclusively to Senior Roles

Realizing the exact junior apprenticeship collapse BOTS.Ai warned about last month, enterprise software firm CIQ reported Sunday that they are now hiring exclusively senior engineers with 5–10 years of experience. CIQ executives stated that while AI tools write over 90% of the company's code, verifying agent-generated outputs requires deep systems knowledge, aligning with Indeed data showing senior roles captured 71% of U.S. software development job posting growth over the last year.

CIQ's hiring policy provides concrete evidence of how automated code generation is severing the entry-level engineering career ladder. When validating agent-generated code requires senior-level systems knowledge, companies eliminate junior positions to avoid review bottlenecks. Industry leaders face a growing challenge in training future senior architects without traditional entry-level roles.

CIQ management maintains that senior-only hiring is necessary because evaluating complex, AI-generated codebases demands deep systems expertise. Industry educators warn that widespread junior hiring freezes create an unsustainable talent pipeline that will cause severe senior engineering shortages within a decade.

Verified across 2 sources: Hoodline (Nov 15) · Business Insider (Nov 15)

Foundation Models & Platform Shifts

Anthropic S-1 Filing Reveals $33.2B Nine-Month Revenue and $26B Non-Voting Compute Placement

Yesterday we covered Anthropic's $64 billion IPO roadshow and $26 billion compute private placement; today, the company's amended S-1 details $33.24 billion in nine-month revenue against a $104 billion net loss, driven heavily by non-cash derivative charges. The filing pushes its fully diluted valuation estimate up to $1.405 trillion—slightly above the $1.3 trillion target we noted over the weekend—and reveals a staggering $900.9 billion in long-term compute obligations to silicon and cloud partners.

The detailed financials establish the massive capital scale needed to maintain competitive position at the model frontier, cementing hyper-scale partnerships with key silicon vendors. For platform builders, Anthropic's $33B revenue trajectory proves that enterprise spending on API infrastructure and agentic workloads is expanding exponentially. However, the $900B compute obligation creates immense pressure to sustain pricing and usage volumes across downstream API clients.

Anthropic's filing frames the immense net loss as an accounting artifact of non-cash derivative revaluations while highlighting hyper-growth in enterprise API adoption. Conversely, independent market analysts argue that backing $900B in long-term compute leases introduces structural margin risks if model commoditization drives down per-token pricing.

Verified across 1 sources: Wowtale (Nov 15)

Alibaba Open-Sources Open Code Review CLI Tool to Enforce AST Guardrails in CI/CD

After spotting Alibaba's Open Code Review (OCR) tool surging in GitHub's telemetry on Friday, the company officially open-sourced the CLI tool on Monday. Having already served tens of thousands of internal developers over two years, the tool cuts token waste by pairing deterministic AST rule matching with agent reasoning to generate precise, line-level pull request comments without resorting to ungrounded repository sweeps.

Alibaba's release illustrates how engineering teams are pairing deterministic code parsers with LLM reasoning to tackle AI-driven code review bottlenecks. By scoping agent context strictly to relevant diffs, OCR mitigates the token waste common in unstructured repository sweeps. This architectural design offers a practical reference for dev tools targeting automated code verification.

Alibaba engineers state that hybrid deterministic-agent architectures are essential for scaling code reviews across large engineering organizations without inflating API costs. Open-source developers note that specialized CLI tools require careful integration into existing enterprise CI/CD pipelines.

Verified across 1 sources: GitHub (Nov 16)

AI Policy Affecting Builders

Third Circuit Rejects ROSS Intelligence Fair Use Defense in Thomson Reuters Copyright Suit

The US Court of Appeals for the Third Circuit affirmed a lower court decision on Monday, November 16, ruling against defunct AI startup ROSS Intelligence in its legal battle with Thomson Reuters. The court held that using Westlaw's copyrighted legal headnotes to train a competing AI platform does not constitute fair use because the material was insufficiently transformative and targeted the same commercial market. ROSS legal counsel indicated plans to petition the Supreme Court.

This binding appellate precedent significantly narrows fair-use protections for AI startups using proprietary domain databases to train commercial models that compete directly with incumbents. Scraping specialized datasets without explicit licenses now carries acute legal exposure within the Third Circuit jurisdiction. Early-stage AI builders must audit their data ingestion pipelines to ensure training sources are fully licensed or non-infringing.

Thomson Reuters praised the ruling as a vital affirmation of intellectual property rights against unauthorized scraping by commercial competitors. ROSS Intelligence and AI advocacy groups argue the decision creates a harmful monopoly over legal data, stifling innovation and favoring established data incumbents.

Verified across 1 sources: Turf Turf Turf (Nov 16)

Ontario Court of Appeal Rules OpenAI Must Defend Copyright Lawsuit in Canada

The Court of Appeal for Ontario unanimously ruled on Friday, November 13, that OpenAI must defend a copyright lawsuit brought by five Canadian news organizations in an Ontario court. The suit seeks statutory damages across 10.65 million news articles, alleging copyright infringement and terms of use breaches. The court determined Ontario has jurisdiction, noting that Canadian law lacks a broad text-and-data-mining exception comparable to U.S. fair use.

This ruling exposes international AI vendors to litigation in foreign jurisdictions where fair-use defenses do not exist. If Canadian courts find web crawling and training infringe Canadian copyright law, model providers may be forced to license local content or geo-block data scraping in Canada. Global startups must account for cross-border copyright liabilities when assembling training datasets.

Canadian publishers argue that tech firms must adhere to local copyright laws and secure commercial licenses when training models on regional news content. OpenAI maintains that training AI models on publicly available web content is a transformative, global practice protected under international fair-use principles.

Verified across 1 sources: Tech Times (Nov 16)

EU AI Act GPAI Obligations Take Full Effect with Enforcement Fines Up to 3% of Turnover

Adding to the rolling EU AI Act compliance deadlines we have been tracking, mandatory Article 53 obligations for general-purpose AI (GPAI) model providers officially took effect on Sunday. Requirements include technical documentation, downstream disclosure, and public training summaries, with models trained using over 10^25 FLOPs presumed to carry systemic risk—triggering strict evaluations under Article 55 with potential fines reaching €15 million or 3% of global annual turnover.

The operational enforcement of the EU AI Act's GPAI rules directly impacts how foundation model developers package, license, and distribute software within the EU. Because compute thresholds capture a broad range of models beyond top frontier labs, open-weight developers must publish detailed training summaries to maintain market access. Early-stage AI startups must establish rigorous compliance registers when serving European users.

European regulators affirm that mandatory GPAI disclosures are necessary to guarantee safety, transparency, and copyright protection across the market. AI founders warn that heavy compliance and reporting burdens will push model development out of Europe, favoring well-capitalized U.S. incumbents.

Verified across 1 sources: Superintelligence Licensing (Nov 15)


The Big Picture

Protocol Standardization Over Proprietary Gateways Developer stacks are codifying formal layers for agent tools, payments, and browser control rather than relying on vendor-locked wrappers. From Nostr-based agent messaging to standardized payment protocols like x402 and AP2, infrastructure builders are focusing on trustless, receipt-backed execution rails.

Enterprise Shift to Multi-Agent Operational Layers Capital allocation is moving beyond single-task co-pilots into swarms of autonomous agents replacing static B2B platforms. Investments like Dreamteam's $53M round indicate that enterprise software buyers expect agents to directly execute backend workflows rather than act as sidecar assistants.

Verification and Maintenance as Primary Technical Limits As models lower the cost of raw code creation, engineering organizations face growing operational overhead in debugging, monitoring, and botsitting. Executive feedback from Microsoft and corporate data from Glean underscore that system architecture and verification now dictate development velocity.

Compression of Entry-Level Engineering Pipelines Data across CIQ, Reserve Bank of Australia, and corporate labor metrics confirm a steep contraction in junior technical hiring alongside rising demand for senior validators. Automated syntax generation is eliminating routine training grounds, forcing teams to redefine how technical reputation and talent development are structured.

Global Divergence in AI Copyright Enforcement Appellate rulings in Canada and the US (ROSS Intelligence) alongside impending legislation in Indonesia reveal a widening split between strict licensing jurisdictions and fair-use regimes. Global product teams face mounting territorial liability when scraping or serving training data across international boundaries.

What to Expect

2026-11-18 — Maven hosts '50 AI Design Patterns That Actually Work' deep-dive UX course
2026-11-19 — Azure OpenAI retires legacy o-series model endpoints for newer GPT-6 variants
2026-11-20 — Cloudflare removes 24 Workers AI models from active catalog
2026-12-12 — Indonesian Parliament target vote on mandatory AI training data copyright bill
2026-12-15 — Nivo UI releases free Figma component set for 2026 AI chat design patterns

Every story, researched.

Every story verified across multiple sources before publication.

🔍

Scanned

Across multiple search engines and news databases

506
📖

Read in full

Every article opened, read, and evaluated

121
⭐

Published today

Ranked by importance and verified across sources

20

— The Signal Room

🎙 Listen as a podcast

Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.

Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste
Overcast
+ button → Add URL → paste
Pocket Casts
Search bar → paste URL
Castro, AntennaPod, Podcast Addict, Castbox, Podverse, Fountain
Look for Add by URL or paste into search

Spotify isn’t supported yet — it only lists shows from its own directory. Let us know if you need it there.