TheBRRR Intelligence

The AI Trade, continuously underwritten

AI Trade Intelligence

One verified research stream feeding the daily briefing, catalyst radar and TheBRRR newsletter.

6
Research runs
24H
Front page

Run 3 ยท Running thesis log

Frontier Labs, Open Models & Software Disruption

Model capability, pricing, agents, enterprise adoption and software moat disruption.

This archive tracks verified developments, the specific change that earned coverage, affected exposures, causal mechanisms and the next observable evidence that would confirm or invalidate each thesis.

Run 3 ยท Models + Software/mixed

Production agents shift software value from user interfaces and seats toward execution, governed context, permissions and outcomes

Anthropic turned computer use, skills and files into a production agent stack

What happened: Anthropic made computer use, the Skills API and the Files API generally available, added a browser-use tool, multi-action turns, versioned custom skills, automatic file expiration, higher rate limits and 1 TB of organization storage. Skills and Files are also distributed through Microsoft Foundry.

Why it earned coverage: A production-platform release that packages procedural knowledge, durable documents and GUI/browser execution, rather than a benchmark-only model update.

Transmission: Reusable procedural packages and persistent files reduce bespoke integration work, while computer/browser use reaches software without APIs. This can move value from UI seats to workflow ownership, permissions, context, execution reliability and outcome pricing.

AnthropicMicrosoft FoundryCRMNOWADBEWDAYTEAMINTUvertical SaaSagent security and observability
Next receipt

Paid API growth; independent task success and cost; renewal/retention; production incident rates; Vertex availability; named customer outcomes; seat changes and displaced software spend.

Invalidation

Reliability, security or cost prevents production use; customers remain in pilots; systems of record capture the economics; or agents fail to produce retained, auditable outcomes.

Impact 4/5 ยท high โ€” Capabilities and availability are documented by Anthropic and its API references; commercial ROI remains unproven. confidence
Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

Frontier-lab value depends on converting capability into reliable external access while using the same systems to accelerate internal research; safety gates now affect monetization timing.

Anthropic turns frontier safety from disclosure into a model-release and internal-productivity gate

What happened: Anthropic's Aug. 14 Risk Report raised its high-stakes-misalignment assessment to low from very low, citing greater uncertainty after recent cyber-evaluation incidents. It disclosed heavy internal use of Mythos 5 and an unreleased Model 2 for coding, data generation and persistent agents; said Claude authors a large majority of code merged into production; estimated internal AI R&D is significantly faster but not yet 2x; and said current task-based evaluations have saturated. Anthropic has no current plan to release Model 2 externally.

Why it earned coverage: A primary company-wide report links rising capability uncertainty to an explicit external-release decision while documenting self-reported internal operating leverage.

Transmission: More capable agents increase internal R&D and coding leverage, but saturated evaluations and real-world incidents raise assurance costs, can delay external monetization and invite government release or access controls. Self-governance becomes investably binding when it changes actual model access or deployment.

AnthropicAmazonGooglefrontier labsAI coding agentscybersecuritymodel API revenueAI governance and insurance
Next receipt

Any Model 2 system card or release, external review of the August report, Anthropic roadmap deadlines in September, government cyber-capability rules, incident recurrence and disclosed enterprise adoption.

Invalidation

Independent audits show robust evaluation coverage, Model 2 is released without tighter controls or delay, internal productivity gains fail to persist, or incidents do not affect customer or regulatory behavior.

Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

Mission-critical software can retain strategic value when cash flow, backlog and workflow authority survive AI multiple compression.

Silver Lake's Workday talks test whether public SaaS pessimism has overshot private value

What happened: Reuters reported after the prior cutoff that Silver Lake is in talks to acquire Workday, which had a pre-report market value near $43 billion. Discussions are ongoing and may not produce a transaction. The latest market snapshot showed WDAY at $206.45, up 17.9%, with market capitalization near $52.5 billion; IGV rose 3.1%. Workday's Q1 FY27 primary results showed 14.3% subscription growth, $27.294 billion total subscription backlog, 31.8% non-GAAP operating margin, $616 million quarterly free cash flow and $4.353 billion cash and marketable securities.

Why it earned coverage: A major financial buyer is reportedly testing the value of a mission-critical software incumbent after sector multiple compression.

Transmission: Sticky HR/finance workflows, backlog, net cash and improving free cash flow can support leverage and private margin optimization even when public investors discount AI disruption.

WDAYSilver LakeIGVCRMNOWenterprise software private equitysoftware private credit
Next receipt

Workday or Silver Lake confirmation, 8-K or merger agreement, bid price and financing package, go-shop terms, regulatory review and next Workday earnings.

Invalidation

Talks terminate without a bid, financing terms prove uneconomic, the board rejects valuation, or Workday growth/retention deteriorates enough to break the leveraged cash-flow thesis.

Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

Enterprise AI adoption is broadening, but investable value depends on retained successful workflows, causal productivity and vendor economics rather than message volume.

CATCH-UP / MISSED PRIOR: Enterprise ChatGPT usage is broadening, but the telemetry does not prove ROI

What happened: Chatterji, Holtz, Rakholia, Tambe and Weeratunga linked ChatGPT Enterprise account records to worker roles, tasks and public-company characteristics through March 2026. At six months, the worker-level sample covers more than 1,500 organizations and more than 17 million messages. Adoption is concentrated among larger, more valuable and more R&D- and SG&A-intensive U.S. public companies; active use spans functions and seniority, with especially high intensity among early-career workers.

Why it earned coverage: Large-scale product telemetry is stronger adoption evidence than executive surveys and directly tests whether enterprise AI remains confined to technical or senior users.

Transmission: Broader and deeper usage expands the workflow surface and potential model/software demand and may favor vendors with enterprise distribution and governance. Without output, cost, renewal, revenue or headcount outcomes, message volume cannot establish productivity or margin capture.

OpenAIChatGPT EnterpriseMicrosoftenterprise software incumbentsAI-native workflow startupsR&D and SG&A laborearly-career knowledge workers
Next receipt

Renewal and cohort retention, cost per completed workflow, causal output-quality studies, public-company margin and headcount effects, vendor revenue attribution and non-OpenAI comparative telemetry.

Invalidation

Independent datasets show usage concentration reverses, broad use does not persist, enterprise renewals weaken, task quality is poor, or controlled outcomes show no economic value despite intensity.

Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

Model value is shifting from raw frontier scores toward price, distribution, routing and reliable agent execution.

CATCH-UP / MISSED PRIOR: Grok 4.6 moves the frontier contest toward price and distribution

What happened: SpaceXAI/xAI released Grok 4.6 on Aug. 12 for long-running agents and interactive work. Its primary page reports an Artificial Analysis Intelligence score of 61, equal to GPT-5.6 Sol Max and one point below Fable 5 Max, with API pricing starting at $2 per million input tokens and $6 per million output tokens. It launched across Cursor, Grok Build, xAI API, OpenRouter, Vercel and Cloudflare.

Why it earned coverage: The combination of frontier-like composite performance, lower list pricing and immediate developer distribution can compress model rent even without clear general superiority.

Transmission: Comparable aggregate capability at lower list price and broad distribution reduces switching friction and shifts value from a single model API toward agent harnesses, context, workflow ownership and routing.

SpaceXAI/xAIOpenAIAnthropicCursorOpenRouterCloudflareVercelagentic developer tools
Next receipt

Independent task-cost data, Cursor cohort usage after the 2x promotion, API token consumption, enterprise deployments and Grok 4.7 results.

Invalidation

Independent production evaluations show materially worse reliability, latency or token-adjusted task cost, or enterprise usage fails to persist after launch incentives.

Read permanent analysis โ†’
Run 3 ยท Models + Software/bullish

Workflow-control incumbents can monetize AI despite generic seat compression

ServiceNow and Atlassian keep narrowing the blanket SaaS bear case

What happened: Q2 subscription revenue rose 24.5% to $3.877B, cRPO rose 21% to $13.2B, total RPO reached $29B and management said AI ACV crossed $1B. Atlassian also reported about $1.766B Q4 revenue, +28%, cloud +31% and RPO about $4.8B, +44%, with earlier primary disclosures showing Rovo customers expanding ARR at roughly 2x non-Rovo customers.

Why it earned coverage: Two workflow incumbents paired growth durability with AI/control-plane receipts, forcing the software thesis away from blanket incumbent bearishness and toward dispersion.

Transmission: Agents need authority, context, permissions, compliance, workflow state and auditability; incumbents that own those control points can capture agent spend while generic seat-based apps compress.

NOWTEAMCRMenterprise workflow softwareAI control planesdeveloper workflow
Next receipt

AI ACV conversion, Rovo/AI-credit monetization, RPO/cRPO durability, organic cloud growth, retention, gross margin after inference and SBC discipline.

Invalidation

AI attach fails to convert into paid revenue, seats compress faster than usage expands, margins deteriorate or agent-native competitors bypass systems of record.

Read permanent analysis โ†’
Run 3 ยท Models + Software/bullish

AI creates software dispersion, not universal SaaS death

Atlassian breaks the lazy SaaS-death narrative

What happened: Atlassian reported fiscal Q4 revenue up 28%, RPO up 44% and a 36% operating margin.

Why it earned coverage: The results are direct counterevidence to a blanket software short and suggest systems of work may monetize AI while weaker seat-rental products compress.

Transmission: Incumbents with workflow control, proprietary context and enterprise distribution can make AI an expansion layer rather than a replacement event.

TEAMWorkflow incumbentsSeat-priced SaaS
Next receipt

FY27 paid AI attach, RPO durability, net retention and seat growth.

Invalidation

RPO slows, paid AI conversion disappoints or agent adoption produces net seat contraction.

Impact 4/5 ยท medium-high confidence
Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

AI consumption revenue is powerful but unusually reflexive

Datadog shows AI demand can cut both ways

What happened: Datadog reported Q2 revenue up 36%, but its guidance reflected lower usage from a major AI customer.

Why it earned coverage: Consumption pricing transmits both workload growth and customer optimization into revenue much faster than contracted seat models.

Transmission: AI customers can become the fastest-growing accounts and also optimize, renegotiate or reroute workloads unusually quickly.

DDOGConsumption softwareAI-native customers
Next receipt

Non-AI growth, customer concentration, usage stabilization and net expansion remain resilient.

Invalidation

AI usage stabilizes without pressure on guidance or net retention.

Impact 4/5 ยท medium confidence
Read permanent analysis โ†’
Run 3 ยท Models + Software/bullish

The enterprise agent moat is permission to act

OpenAI and Anthropic build the agent operations stack

What happened: OpenAI introduced Presence for enterprise voice/chat agents that can use systems, take approved actions and escalate. OpenAI also said enterprise is more than 40% of revenue and APIs process more than 15B tokens per minute. Anthropic's Claude Sonnet 5 is positioned as its most agentic Sonnet model, priced at $2/MTok input and $10/MTok output, with cyber safeguards enabled.

Why it earned coverage: Recent OpenAI and Anthropic enterprise/model releases.

Transmission: Enterprises buy reliability, permissions, governance and escalation systems, not just raw benchmark scores.

OpenAIAnthropicAgent infrastructureNarrow SaaS
Next receipt

Named customer case studies, renewal/seat expansion, support-labor metrics, incident rates.

Invalidation

Enterprise agent deployments fail to reduce cost or require too much human review to scale.

Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

Enterprise systems of record can defend the AI control plane through data, permissions, workflow authority and governance, but activity must convert into retained paid economics while human-seat value compresses.

CATCH-UP / MISSED PRIOR: Salesforce has a production-agent usage receipt, but only for survivors

What happened: Salesforce's Agentic Enterprise Index Second Edition reports that among businesses with Agentforce agents active in production every month from February 2025 through April 2026, average activated agents rose from 5 to 13, average unique skills per agent rose from 2 to 6, the action-to-output ratio grew at a reported 15% compound monthly rate, and Salesforce counted 734 million self-defined Agentic Work Units in April 2026. The primary methodology omits cohort size and excludes failed, paused and newly adopting organizations.

Why it earned coverage: Direct production telemetry on agents invoking business logic is stronger than a survey and directly tests whether an incumbent system of record can remain the action and governance plane for agents.

Transmission: Agents that update CRM records and invoke APIs increase the value of governed data, permissions and workflow ownership, allowing incumbents to defend the control plane even as human seats compress. A continuously active survivor cohort can scale while broader customers fail, and action counts can increase cost without improving outcomes. Durable value requires retained paid workflows, incremental revenue and successful-task economics above inference and implementation costs.

Salesforce and AgentforceCRMData 360 and SlackServiceNow and enterprise workflow incumbentsAI-native agent startupssystems integratorscybersecurity identity and agent-observability vendors
Next receipt

Cohort denominator and inclusion counts, failed and paused deployments, Agentforce ARR and organic contribution, RPO and NRR, customer spend expansion, inference-adjusted gross margin, seats and pricing, task success, human escalation and independent customer references.

Invalidation

Aggregate customer data show high abandonment, low renewals, stagnant Agentforce revenue, rising inference or implementation cost, poor task success or AI-native competitors displacing the incumbent control plane.

Impact 4/5 ยท medium-high for the reported cohort; low for market-wide extrapolation confidence
Read permanent analysis โ†’
Run 3 ยท Models + Software/bullish

Enterprise AI value is moving from chat assistance to delegated workflow execution

Codex usage shows enterprise AI shifting from assistance toward delegated execution

What happened: OpenAI reported that by May 80.6% of sampled individual Codex users had made at least one request estimated to exceed 30 minutes of human work; external organizational users generated about 63.3% of output tokens through Codex by June, while OpenAI's own internal rate was far higher.

Why it earned coverage: A large-scale usage study measures longer-horizon agent work rather than benchmark capability alone.

Transmission: Longer delegated tasks increase token/tool consumption and substitute for portions of human workflow, shifting value toward control planes and context-rich systems.

OpenAIagent orchestrationdeveloper toolsenterprise workflow softwareknowledge-work labor
Next receipt

Enterprise renewal/expansion, successful-task rates, cost per completed workflow and non-engineering adoption.

Invalidation

Organizational usage stalls, estimated long tasks fail economically, or agents remain confined to narrow coding workflows.

Impact 4/5 ยท medium-high confidence
Read permanent analysis โ†’
Run 3 ยท Models + Software/mixed

AI labor impact is showing first in task and entry-hiring compression, not broad unemployment

AI exposure shows a hiring signal before an unemployment shock

What happened: Anthropic found observed AI coverage remains a fraction of theoretical exposure; higher-exposure occupations have slightly weaker BLS projected growth, with tentative slower hiring for ages 22โ€“25, but no systematic unemployment increase in highly exposed occupations.

Why it earned coverage: Usage-linked labor evidence separates theoretical capability from observed deployment.

Transmission: Firms can reduce junior hiring or change task composition before layoffs appear in aggregate unemployment data.

entry-level knowledge workerscodingcustomer servicefinancial analysisenterprise hiringlabor-sensitive software seats
Next receipt

Updated CPS/JOLTS and occupational payroll data, ages 22โ€“25 hiring, wages and firm-level headcount per output.

Invalidation

Exposed occupations sustain equal or stronger early-career hiring after controlling for cycle and industry, with little task substitution.

Read permanent analysis โ†’