AI & Automation Briefing - November 1, 2026
New Benchmark Shows AI Agents Excel at Equity Research but Fail at Investment Judgment
A new paper from HuggingFace introduces Investor's Last Exam (ILE), a benchmark that tests AI agents on the full equity investment workflow using only information available at a specified decision date. The benchmark corpus spans 3.43 million versioned documents across 1.52TB and organizes 86 assignments around buy-side analyst tasks, with scoring tied to financial dependency graphs and reproducible calculations. Results show a sharp performance gap: Google's Astra scored 78.9 on research and evidence tasks but only 52.5 on forecasting and financial modeling. When ranked by realized investment outcomes, DeepSeek V4.1 Flash led with 40.3 versus Astra's 37.7. The finding confirms that current top models are strong at synthesizing historical evidence into analysis but weak at translating that analysis into forward-looking judgment. The time-cutoff mechanism also opens a path for domain-specific reinforcement learning, since historical timestamps can serve as a resamplable training environment using the same benchmark infrastructure.
Investor Confidence in Big Tech AI Spending Starts to Crack
Wall Street is growing skeptical of AI infrastructure spending just as Alphabet, Amazon, Meta, and Microsoft commit a combined $740 billion or more in capital expenditures for the year. Early in the AI buildout, investors treated large capex as a confidence signal. Now concerns about negative free cash flow and mounting debt are driving a market pullback. The Nasdaq dropped 2.9% in a single session, with Nvidia down 4%, Broadcom down 5%, and Qualcomm down 11%. Alphabet posted negative free cash flow of $9.1 billion in Q3, its second consecutive losing quarter, while spending $51.4 billion in capex that quarter alone. Its Search advertising business, about half of total revenue, grew only 12%, the slowest pace in five quarters, partly because its own AI Mode feature is reducing ad exposure. Amazon reported negative free cash flow of $42.4 billion over the trailing 12 months, reversing $14.8 billion in positive free cash flow from the prior year period, despite 39% growth at AWS. Meta posted negative free cash flow of $2.1 billion in Q3 and is guiding to up to $145 billion in capex for the year. Microsoft was the only major exception, recording $18.6 billion in positive free cash flow. For operators pitching AI automation to business clients, this is useful context: the ROI pressure hitting Big Tech is the same pressure clients face, making practical, measurable automation a stronger sell than broad AI investment narratives.
Reddit Moves to Logged-In Model as AI Search Kills Referral Traffic
Reddit reported that search referral traffic dropped by more than 1 million visits per week in October, a direct result of AI-powered search reducing click-throughs across the open web. In its Q3 earnings call, the company said it is transitioning to a primarily logged-in experience to protect its content from AI scraping, position itself for AI data-licensing deals, and improve advertiser targeting. U.S. daily active users fell from 53.2 million in Q2 to 50.7 million in Q3, down 2% year-over-year, and the company warned of a potential low-double-digit percentage decline next quarter if the trend holds. Despite the user erosion, revenue grew 53% year-over-year to $895 million, with U.S. revenue per daily active user rising 53% to $13.82, showing Reddit can monetize a shrinking audience more efficiently through logged-in signals. For operations and content teams running SEO-dependent workflows or audience-facing automations, Reddit's pivot is a clear signal that the open, search-driven distribution model is under structural pressure.