Evidence Timeline
Week of February 3–10, 2026 — the week the thesis went mainstream
Feb 3–4, 2026
Anthropic's Cowork plugins triggered a massive selloff in SaaS stocks. The Nasdaq Cloud Index plummeted, erasing $285 billion in a single session. Jefferies coined it the "SaaSpocalypse." Bloomberg headline: "Get Me Out" — traders dumped software stocks as AI fears erupted. BDCs with SaaS exposure hit hardest.
↗ Thesis Connection
The market correctly priced in AI replacing per-seat SaaS. But the panic assumed full autonomy. Regulated industries can't go unsupervised — the correction overshoots, creating the opportunity for a supervision layer.
Feb 6, 2026
Goldman's CIO Marco Argenti revealed 6-month collaboration with embedded Anthropic engineers. Building autonomous agents for trade accounting and client onboarding. Argenti called them "digital co-workers" for "scaled, complex, process-intensive" roles. Claude chosen for its reasoning ability — surprised them by excelling beyond coding into compliance and accounting.
↗ Thesis Connection
Goldman is building exactly what the market fears — AI agents in core financial operations. But note the language: "digital co-worker," not "replacement." A top-5 bank with $2.5T AUM won't run unsupervised agents on trade accounting. They need the supervision layer. This is our customer.
Feb 10, 2026
Maximor's Finance AI Adoption Benchmarking Report surveyed 100 mid-market CFOs ($50M–$500M revenue). Key findings:
• 79% of CFOs say 25%+ of finance workload already handled by agentic AI
• 86% have encountered hallucinated data from AI in finance tasks
• 97% say human oversight is critical (67% "extremely" or "very")
• Only 14% completely trust AI to deliver accurate data
Report quote: "Finance leaders will trust AI when they can audit it. Verifiable, traceable, and explainable outputs are non-negotiable."
↗ Thesis Connection
This is the smoking gun. AI agents are already doing the work (79%), but trust is broken (86% hallucination encounters). The report literally names our product category: "verifiable, traceable, and explainable outputs." That's a supervision dashboard. That's Plex Labs.
Feb 5, 2026
OpenAI launched Frontier — an enterprise platform to build, onboard, and manage AI agents. Includes feedback loops "meant to help agents improve the same way a review might help an employee." Works with OpenAI agents, enterprise-built agents, and third-party agents (Google, Microsoft, Anthropic). Customers: HP, Oracle, State Farm, Uber. Limited rollout.
↗ Thesis Connection
OpenAI sees the same opportunity — agents need management. But Frontier is model-provider-first, not industry-first. A bank doesn't want OpenAI managing their compliance agents. They want domain-specific supervision built for financial regulation. Validates the category; doesn't occupy our niche.
Feb 4, 2026
Alphabet announced massive capex increases for AI compute capacity — Google DeepMind expansion plus meeting "significant cloud customer demand." The hyperscalers are all-in on AI infrastructure. The supply side of the equation is solved. The demand side — how enterprises safely use these agents — is where value accrues.
↗ Thesis Connection
More compute = more capable agents = more need for supervision. Every dollar spent on AI infrastructure increases demand for the governance layer. We're not competing with the foundation model companies — we're building on top of their success.
Feb 4, 2026
Plex Labs was founded the same day $285B was erased from SaaS. Not planned — but the timing tells the story. A CFA + AI agent, building the supervision layer that the market just validated the need for. Within 6 days: website, investment thesis brief, pitch deck, and working product demo.
↗ Thesis Connection
The founding narrative writes itself. While the market panicked about AI agents replacing software, Plex Labs was being built by a human and an AI agent — working together. That's not just a company story. It's a proof of concept.