AI Briefing

Stanford just dropped its annual AI report card and the headline number is wild: generative AI hit 53% population adoption in three years, faster than the PC or the internet. But the uncomfortable subplot is that AI transparency is collapsing just as capabilities soar, with model disclosure scores plunging from 58 to 40. Meanwhile, Goldman Sachs put hard numbers on AI job losses for the first time, PwC found that 20% of companies are hoarding 74% of AI's economic value, and Google's TurboQuant algorithm might be about to reshape the economics of running AI entirely.

01ResearchStanford HAI ↗

Stanford's 2026 AI Index: Adoption Outpacing the PC and Internet, but Transparency Is in Freefall

Stanford HAI's annual AI Index landed at over 400 pages and paints a picture of an industry accelerating in capability while retreating on openness. Generative AI hit 53% population adoption within three years, faster than any previous technology wave. Coding benchmarks like SWE-bench Verified jumped from 60% to near-100% in a single year, and the best models now clear 50% on Humanity's Last Exam. But the Foundation Model Transparency Index dropped from 58 to 40, with 80 of the 95 most notable models launched without training code. The report also flags that China has nearly closed the gap with the US on model performance, software developer employment for ages 22-25 has fallen 20% since 2022, and documented AI incidents rose to 362. This is the most comprehensive annual snapshot of AI's state, and this year it reads like a warning: the technology is working, the guardrails are not keeping pace.

02IndustryFortune ↗

Goldman Sachs: AI Is Erasing 16,000 US Jobs Per Month, with Gen Z Taking the Hardest Hit

A Goldman Sachs US Daily note authored by economist Elsie Peng represents one of the most granular attempts yet to separate AI's substitution and augmentation effects on employment. The analysis finds AI substitution is wiping out roughly 25,000 jobs per month while augmentation adds back about 9,000, yielding a net loss of 16,000 jobs monthly over the past year. Gen Z workers are disproportionately affected because they're concentrated in the routine white-collar and administrative roles that AI automates most effectively. The wage impact is measurable too: a one standard-deviation increase in AI substitution exposure widens the entry-level-to-experienced wage gap by 3.3 percentage points. Perhaps most concerning, Goldman's analysis of four decades of labor data shows displaced workers struggle for up to 10 years after their initial job loss.

03IndustryPwC ↗

PwC Study: 20% of Companies Are Capturing 74% of All AI Economic Value

PwC surveyed 1,217 senior executives at companies with $1 billion or more in revenue across 25 sectors and found a stark divide: just one-fifth of organizations are capturing nearly three-quarters of AI's total economic gains. The differentiator isn't technology adoption but strategy. The top performers invest 2.5x more in AI than peers, deliver 7.2x better AI-driven financial performance, and are nearly twice as likely to use AI autonomously. Critically, they use AI for growth and business model reinvention rather than just cost-cutting. These leaders are increasing autonomous decision-making at 2.8x the rate of peers and are 2.6x more likely to report AI has helped them reinvent their business model. The gap is organizational, not sectoral, and it is widening.

04ResearchGoogle Research ↗

Google's TurboQuant Algorithm Compresses AI Memory 6x with Zero Accuracy Loss, Spooking Chip Stocks

Google researchers unveiled TurboQuant, an algorithm that compresses the KV cache in large language models to just 3 bits per value, cutting memory use by 6x and accelerating attention computation by up to 8x on NVIDIA H100 GPUs, all with zero accuracy loss. The technique requires no calibration data, no fine-tuning, and works on any transformer architecture. Accepted at ICLR 2026 for formal presentation on April 25 in Rio de Janeiro, TurboQuant has already rattled markets: memory chip stocks including Micron and Western Digital declined as analysts realized that if AI companies can compress their memory requirements through software alone, the insatiable demand for High Bandwidth Memory may be fundamentally tempered. Some have called this Google's 'DeepSeek moment,' a software breakthrough that could reshape the economics of AI infrastructure.